Phenom 9850 Black Edition released today!!

[GunterFalstaff]

as you all are probably aware... the new B3 revisions came out today (march 27th)
http://www.hothardware.com/articleimages/Item1127/avail_chart.png
and i was wondering who all plans on getting one and seeing how far they can go? i am really interested in how well these perform on air and water as i am definitely considering getting one =P. it seems the oc'ing trend on air with most critic websites is around 2.8-2.9, if these hit 3.0 and over with h2o... count me in! now ill have a reason to finally balls out and buy a water cooled system :)

[Rammsteiner]

Im waiting a little to get some more hardware. But Ill get one. And for less then 200Euro's in shops, very nice:yepp:

Hope it OC's nicely too.

X3's are released too

[SocketMan]

Time for all the guys with 790fx and 6400+, "waiting for B3s", to put their money where their mouth is.:p:
Mega-hurts await you.:D

[gurusan]

according to Anandtech reviews today even if you do manage to get them to 3.0-3.2ghz a similarly clocked dual core will outperform it in games.

[GunterFalstaff]

can you get a link? and if that is so... >.< i will probably have to wait for the 45nm phenoms :( . im currently sitting 24/7 at 3.5ghz with my 6400 and i do ALOT of gaming and lanparties and whatnot... i really dont want to get a downgrade =/

[gurusan]

AnandTech: Higher Clock Speeds, No TLB Issues and Better Pricing: The New Phenom (http://anandtech.com/cpuchipsets/showdoc.aspx?i=3272&p=1)

http://images.anandtech.com/graphs/phenom9850_032708031830/16767.png http://images.anandtech.com/graphs/phenom9850_032708031830/16768.png

example of a very gpu limited game (oblivion):
http://images.anandtech.com/graphs/phenom9850_032708031830/16769.png http://images.anandtech.com/graphs/phenom9850_032708031830/16770.png

[gurusan]

actually some other reviews came out and it looks like they are getting 3.0ghz stable...

[Suosaaski]

AnandTech: Higher Clock Speeds, No TLB Issues and Better Pricing: The New Phenom (http://anandtech.com/cpuchipsets/showdoc.aspx?i=3272&p=1)

http://images.anandtech.com/graphs/phenom9850_032708031830/16767.png http://images.anandtech.com/graphs/phenom9850_032708031830/16768.png

example of a very gpu limited game (oblivion):
http://images.anandtech.com/graphs/phenom9850_032708031830/16769.png http://images.anandtech.com/graphs/phenom9850_032708031830/16770.png

The graphs you posted also show up the effect of TLB-fix in those applications.

[tictac]

3GHz stable ? Nice :up:

[Undersea]

Most the reviews show OCed to 2.7 - 2.8.

One had it at 3.0g with 1.517 volts. Lots of volts!

I think I am holding on for 45nm. The game reviews show very little difference between X2's and X4 in FPS. :(

[gurusan]

tbh I'd rather have a 3600+ that clocks to 3.4-3.5ghz over the current phenoms.

[Rammsteiner]

tbh I'd rather have a 3600+ that clocks to 3.4-3.5ghz over the current phenoms.
I'd rather have a Phenom that clocks to 3.4~3.5Ghz over a same clocked Athlon X2's...

...

What's did for strange comparison:shakes:

[gurusan]

I'd rather have a Phenom that clocks to 3.4~3.5Ghz over a same clocked Athlon X2's...


Errm yeah of course but good luck clocking your phenom.

[madfaze]

arrrgghhhhh cant wait to have this....:D


anyway which you prefer triple core or the quad?

[BertM]

when nehalem comes out i will buy a phenom to.

[madfaze]

hopefully someone could give us a good batch to purchase.. :D

[BrokeDown]

45nm is what I'm waiting for. If AMD can't get it right by the end of this year, I'm giving up and going 790i/8400 or a quad 45nm intel

[madfaze]

isnt that 4 series vcards are also coming out this year?...

[Oese]

lol its obvious that a quad won't get better gaming scores then a dualcore unless it's clocked at comparable speeds..

same issue as with dualcore a few years before.. simply there is lack in software supporting multithreading..

so quad is quad and gaming is gaming... if you want a quad then a quad makes sense, if you want gaming, both quad and dualcore will be fine at around 3Ghz... Very nice if thats to be confirmed 3Ghz at air stable.. :D

[jas420221]

The graphs you posted also show up the effect of TLB-fix in those applications.Last I checked the TLB issue is moot with the B3. There isnt a 15% performance loss with the ability to overclock. So, I believe that statement is not correct.

[GunterFalstaff]

as far as lower gaming scores goes... thats expected... none of those games utilize 3+ cores yet, games only recently started integrating dual core optimizations, so as time passes quad core will become the standard for gaming. i dont think there will be enough of a difference between a 3.0ghz 9850 and my 3.5ghz 6400+ in games to outweigh the massive benefits of multitasking, (i have 3 monitors and i like to play movies and stuff while i game =P).

[B.E.E.F.]

Any word on the 45nm?

[Undersea]

Any word on the 45nm?

Second Half, late of 08' (Rumor)

[Oese]

as far as lower gaming scores goes... thats expected... none of those games utilize 3+ cores yet, games only recently started integrating dual core optimizations, so as time passes quad core will become the standard for gaming. i dont think there will be enough of a difference between a 3.0ghz 9850 and my 3.5ghz 6400+ in games to outweigh the massive benefits of multitasking, (i have 3 monitors and i like to play movies and stuff while i game =P).


i agree and additionally phenom is faster clock per clock compared to x2 as one can imagine from those figures so perhaps if 3ghz really could be stable or even more if you're lucky this would be very nice. let this b3 stepping mature a bit and we'll see...

[GunterFalstaff]

the 9950 is due for Q3, so i would say 45nm would probably be around Q4 this year of Q1 next... i think i might wait for 9950 for a more stable 3.0ghz clock.. but heck, if i wait that long i might as well wait for 45nm >.<

[Oese]

thats the problem..

I'll keep at x2 for a moment as i dont really need phenom (though it would be interesting to move to quadcore and especially phenom oc :love: soon..) but watching oc results and how well it goes under x64 environment is more sensible for me at the moment...


whatever, finally a bug free phenom:clap: :clap: :clap:

[Aussie FX]

The review sites aren't spending any time at all trying to overclock this cpu. All I have seen is the C2D form of overclocking, up the multi and see what gives.
Phenoms are a lot more subtle than that.
I believe these should hit 3ghz easily with a little bit of NB lovin.

@Gunterfalstaff, all new games coming out are now starting to be multithreaded, the Phenom will come to you. Look at Bioshock it is knocking off intels finest.
For ~$200 you can't go wrong imo.

[GunterFalstaff]

i personaly LOVE my 6400, runs soo smooth at 3.5ghz, i just hope that i dont "feel" any slower running a 3.0ghz'ish phenom... >.< if they are as easy to clock to 3.0 as you say i think i might very well buy one within the week tho =P i just need to find sum1 to take my 6400 off my hands =/

[Aussie FX]

i personaly LOVE my 6400, runs soo smooth at 3.5ghz, i just hope that i dont "feel" any slower running a 3.0ghz'ish phenom... >.< if they are as easy to clock to 3.0 as you say i think i might very well buy one within the week tho =P i just need to find sum1 to take my 6400 off my hands =/
I've still got my 6400 and it ran at 3.5 too. The Phenom feels so much faster it's not funny. In reality it's not but it's a bit like going from single to dualcore everything is just so much smoother. I'm sure you won't be disappointed.
What motherboard do you have, because I tried my Phenom in a Crosshair and it was a bit of a nightmare. AM2+ are definitely the way to go.
I'm not saying they are easy to overclock to 3ghz, I don't know I haven't tried one but the reviews I have seen they are upping the multi only and the lowest oc I have seen has been 2.8, so going on that if you spent the time 3ghz should be possible.

EDIT: I notice you have the DFI 790 so you will be sweet.

[GunterFalstaff]

well, do you think its possible to get a good stable OC to 3ghz on just air tho? im running my current 6400+ at 3.5ghz and it sits around 28 celcius idle (no cool and quiet) i have very good airflow and a lapped Asus arctic cooler (the 6400 is lapped as well and using AS5). if i got the 9850 i would probably lap it as well... with this current cooling setup, do you think i could get a stable 3.0? or do i need to go H2O?

[BrokeDown]

Well I forgot who did it but they got the 9850BE to 3.0Ghz stable; took almost 1.5v and I think it was on air too. I'm sure it would be possible, but for a 24/7 OC I would set at the highest stable clock on the lowest voltage, which looks to be about 2.8-2.9 Ghz.

With a good lap and H20 I'm sure 3.0Ghz would be capable on a slightly lower voltage.

Edit: It was Tech Report


2.7GHz/stock - pass
2.8GHz/stock - pass
2.9GHz/stock - no POST
2.9GHz/1.36V - pass
3.0GHz/1.36V - hang on Windows boot
3.0GHz/1.403V - BSOD on boot
3.0GHz/1.442V - BSOD on boot
3.0GHz/1.481V - pass
3.1GHz/1.481V - hang on boot
3.1GHz/1.519V - hang on boot

[Oliverda]

Well I forgot who did it but they got the 9850BE to 3.0Ghz stable; took almost 1.5v and I think it was on air too.

...with an MSI K9A2. :rolleyes: We should wait for a test with DFI LP UT 790FX-M2R. :D

[SocketMan]

I was taking the garbage out and someone left a working Northwood near the dumpster.The note on it said: "It works".
Think it was about 2.77Mhz or so, gigabyte board, anyway it over clocked to 3ghz. True story.

Now do you think I game on it because it's 400 Mhz higher then my Phenom?
Phenom brings in the HT3 for all the video bandwidth you might need specifically people with 3-4 GPU's .The cashes are faster then A64 too.
Clocks are nice but once you get into the high resolutions there is little difference that cpu clock will make.
I am thinking about getting one of the AM+ Opterons once they're out.
Has anyone (from the review sites) over clocked the North Bridge and/or HT ?
That makes huge difference for Ram/L3.

[Loser777]

Lol, all my friends did a week of suicide runs with Northwoods @ 1.8v when they got an A64. :P
3.0GHz is impressive enough on a Phenom, screw Yorkies...

[Sparky]

OK so I feel dumb, probably should have waited until the B3 before I got my 9600, now my 9600 is worth less and outperformed by a 100MHz slower 9550 :rolleyes: :doh:

[Lightman]

No single test in x64 environment... :shakes:
That means I will be guinea pig on that field :cool:
Ordering 9850BE as soon as they hit UK (except Overclockers - a bit too pricey:down: )!

[Undersea]

No single test in x64 environment... :shakes:
That means I will be guinea pig on that field :cool:
Ordering 9850BE as soon as they hit UK (except Overclockers - a bit too pricey:down: )!


I probably get one in coming weeks after I see some guinea pigs here. It will go into my Vista 64 machine :)

[BrokeDown]

I wouldn't mind a Phenom, but the big thing for me is the board to go with it. The Nv 780a boards look cool, but I haven't seen any for sale yet.

[SocketMan]

OK so I feel dumb, probably should have waited until the B3 before I got my 9600, now my 9600 is worth less and outperformed by a 100MHz slower 9550 :rolleyes: :doh:

On the bright side the price on B2's will probably drop so you can get another one;)

Just make sure your next AM2+ mobo has the TLB kill switch;) so you don't
loose the performance:)

I am sure you've seen this but anyway.......:

http://techreport.com/articles.x/14424/12

[Sparky]

Hmmm...
Anyone want a barely used, never overclocked 9600BE? :p:

[SocketMan]

To me when I see someone with a B2 Phenom it tells me 1 thing.
They got balls:cool:




One of the comments from Xbit's review

8. ZOMG....intels cheapest quad core only costs 35% more then AMDs...and on average performs 10-20% better real world apps.

Then of course theres the fact that their new "ground breaking" chip they put out every 3 months or so goes for $1500 or so, and performs marginally better than the almost identical chip that costs around $500.

But hey, at least with every new chip line...you gotta go and spend another $200-300 on a new motherboard that supports it...but it still will feel farmiliar since it's most likely the same socket design you're upgrading from. Maybe you'll even get to buy some new ram to support it since ram compatibility changes with that pesky FSB/NB thing.

2 years, 8 sockets and....how many pin layouts?

I'm very happy for you intel fan boys that brag about dropping several thousand bucks every 6 months or so to get marginal performance gains, no real design improvement outside of smaller die size, bigger slower cache and putting 64bit development at a standstill because intel can't figure that out quite yet.

But hey, AMD just managed to put out a board that will run vista, play Blu-ray, and even allow for passable gaming with no add in video card. Actually can build a vista system (board, cpu, 2 gigs of ram) for around $150. Not like intel had a big market share in that area or anything...oh wait..yeah they kind of did. But maybe they'll still compete with thier onboard graphic boards that can't run xp without an extra card.

Then again, what about the server market...intel server chips go for around $5k a piece, and put out 220-260w per chip...ah just like a cool spring breeze...of course with the cooling includes 8 high speed 120MM fans - 35mm deep and 6000 rpm. that nice heat output is coupled with the sound of a jet engine....and only with the need for a low power 1100w PSU.

Yes, they're faster. They're more expensive, they're seemingly plauged with driver problem and performance and compatibility issues....at least for awhile. But that's if you can get the new ones within the first 6 months of the "offical launch"

How bout you go do what was done in the old days, ya know when overclockers took all parameters into account.

Look at the speed of the processors.
Look at the Cache sizes.
Remember that the power consumption of the intel NB. should be factored in to even things up with AMDs on die memory controller in that power consumption of the cpu, ya know if you want to nit pick...or be accurate.
Look at memory speeds (notice that AMD still holds something like the top 30 slots in memory performance before intel makes it on the list, despite the often considerably slower speeds.

Now, go to tomshardware cpu charts. Look at the real benchmarks, not the theorhetical synthetic crap. Figure out the performance difference in each bench mark, then compare it to the corresponding piece of hardware that's benched.
Then factor in the price.

For example, back when the qx6850 was coming out with it's
$1500 price tag, it was benched against the Athlon 6000x2.
same clock speed,
intel chip has 75% more L2 cache,
50% more cores,
running 30% smaller die size
8% faster memory.

AMD still had like a 25% lead in memory performance despite it being slower, and real world app benchmarks the intel chip averaged 15-30% faster. The high end of the performance gain intel had was 45% in one bench. Intel quad, 2x4mb L2, 65nm die 800mhz ram..against the amd dual core, 2x1mb L2, 90nm 750mhz ram.

Average highend performance gain of 30% in real word apps.
Intel $1500
AMD $ 180
difference 88%
sad part, qx6850 had like a 20% high end performance gain over the $300 2.4ghz q6600 with other wise identical specs.

And now, the qx6850 still costing $1000, compared to the $200 Phenom has an avg performance gain of 10-20%, hitting a high at 35% in one bench.

If someone gave me an intel rig, sure i'd use it. But i'd much prefer to build 2 or 3 AMD rigs and over clock them to the breaking point and still be stable for the same cost of one intel rig.

lol it's just absurd that anyone can complain about AMD overcharging for a chip after doing the math, getting the pure clean hard facts about intels performance even compared with itself and the prices charged.

Intel has the lead, but it isn't as big as you amd flaming intel fanboys would like to belive, the performance in no way merits the price, and they don't lead in everything.

In fact, when it comes to innovation...they're way behind, look at the fact that it will have taken 6 years to release a chip with an IMC, and a couple more to figure out how to make a native quad core...

[Extelleron]

To me when I see someone with a B2 Phenom it tells me 1 thing.
They got balls:cool:




One of the comments from Xbit's review

8. ZOMG....intels cheapest quad core only costs 35&#37; more then AMDs...and on average performs 10-20% better real world apps.

Then of course theres the fact that their new "ground breaking" chip they put out every 3 months or so goes for $1500 or so, and performs marginally better than the almost identical chip that costs around $500.

But hey, at least with every new chip line...you gotta go and spend another $200-300 on a new motherboard that supports it...but it still will feel farmiliar since it's most likely the same socket design you're upgrading from. Maybe you'll even get to buy some new ram to support it since ram compatibility changes with that pesky FSB/NB thing.

2 years, 8 sockets and....how many pin layouts?

I'm very happy for you intel fan boys that brag about dropping several thousand bucks every 6 months or so to get marginal performance gains, no real design improvement outside of smaller die size, bigger slower cache and putting 64bit development at a standstill because intel can't figure that out quite yet.

But hey, AMD just managed to put out a board that will run vista, play Blu-ray, and even allow for passable gaming with no add in video card. Actually can build a vista system (board, cpu, 2 gigs of ram) for around $150. Not like intel had a big market share in that area or anything...oh wait..yeah they kind of did. But maybe they'll still compete with thier onboard graphic boards that can't run xp without an extra card.

Then again, what about the server market...intel server chips go for around $5k a piece, and put out 220-260w per chip...ah just like a cool spring breeze...of course with the cooling includes 8 high speed 120MM fans - 35mm deep and 6000 rpm. that nice heat output is coupled with the sound of a jet engine....and only with the need for a low power 1100w PSU.

Yes, they're faster. They're more expensive, they're seemingly plauged with driver problem and performance and compatibility issues....at least for awhile. But that's if you can get the new ones within the first 6 months of the "offical launch"

How bout you go do what was done in the old days, ya know when overclockers took all parameters into account.

Look at the speed of the processors.
Look at the Cache sizes.
Remember that the power consumption of the intel NB. should be factored in to even things up with AMDs on die memory controller in that power consumption of the cpu, ya know if you want to nit pick...or be accurate.
Look at memory speeds (notice that AMD still holds something like the top 30 slots in memory performance before intel makes it on the list, despite the often considerably slower speeds.

Now, go to tomshardware cpu charts. Look at the real benchmarks, not the theorhetical synthetic crap. Figure out the performance difference in each bench mark, then compare it to the corresponding piece of hardware that's benched.
Then factor in the price.

For example, back when the qx6850 was coming out with it's
$1500 price tag, it was benched against the Athlon 6000x2.
same clock speed,
intel chip has 75% more L2 cache,
50% more cores,
running 30% smaller die size
8% faster memory.

AMD still had like a 25% lead in memory performance despite it being slower, and real world app benchmarks the intel chip averaged 15-30% faster. The high end of the performance gain intel had was 45% in one bench. Intel quad, 2x4mb L2, 65nm die 800mhz ram..against the amd dual core, 2x1mb L2, 90nm 750mhz ram.

Average highend performance gain of 30% in real word apps.
Intel $1500
AMD $ 180
difference 88%
sad part, qx6850 had like a 20% high end performance gain over the $300 2.4ghz q6600 with other wise identical specs.

And now, the qx6850 still costing $1000, compared to the $200 Phenom has an avg performance gain of 10-20%, hitting a high at 35% in one bench.

If someone gave me an intel rig, sure i'd use it. But i'd much prefer to build 2 or 3 AMD rigs and over clock them to the breaking point and still be stable for the same cost of one intel rig.

lol it's just absurd that anyone can complain about AMD overcharging for a chip after doing the math, getting the pure clean hard facts about intels performance even compared with itself and the prices charged.

Intel has the lead, but it isn't as big as you amd flaming intel fanboys would like to belive, the performance in no way merits the price, and they don't lead in everything.

In fact, when it comes to innovation...they're way behind, look at the fact that it will have taken 6 years to release a chip with an IMC, and a couple more to figure out how to make a native quad core...


You make it sound as if the only two chips that can be compared here are a $1,000 QX6850 (which nobody in their right mind would buy with the QX9650 out, anyway. And I don't know where you get the idea that a QX6850 ever cost $1,500... it has always been $1,000) and a $230 Phenom 9850. And of course that couldn't be farther from the truth. You seem to forget that Intel has chips ranging from $40-$1,000. The $230 Phenom 9850 competes with Intel's Q6600, which has a 1000-unit pricing of $266. Its street price is actually quite lower than that; yesterday you could have bought a Q6600 for $180 at Frys, and the Q6600 has been available for $199 at Microcenter for quite a long time now.

When you compare the Q6600 and Phenom 9850... at stock settings, the Q6600 is faster in terms of performance and it consumes significantly less power. When overclocked, the chips aren't playing in the same ballpark... Q6600's hit 3.6GHz all the time, meanwhile Phenom 9850 seems to hit around ~2.9GHz from what we have seen so far.

I definately prefer AMD processors and have never built a system with an Intel CPU yet, but I cannot deny and anyone reasonable cannoty deny that Intel has a solid lead in terms of CPU performance for enthusiasts. At every single price range, you can find an Intel CPU that is just as good or better than an AMD CPU at the same price - the only possible exception is the Phenom X4 9550, which for $195 will be cheaper than the Q6600. But it's still slower than Intel's E8400 most of the time, and even in multithreaded apps, the 2.2GHz Phenom is not that much faster than a dual core Penryn @ 3GHz. When you throw overclocking and power consumption into the equation... once again, it's not even comparable.

As for innovation... innovation is great, but not that many people care about what is in a CPU core, they care about how it performs. Personally I find AMD CPUs more interesting than Intel's because of their unique features and arguably more advanced design, but if that advanced design doesn't net higher performance, then it is useless to anyone but the geeks who care about what is behind the IHS. AMD certainly has a better memory system, better scaling with native quad, and also better clock scaling (as long as both core + NB frequency is increased), but the problem is Intel's core logic is much more advanced than AMD's. The logic within Core 2 is state of the art, meanwhile Phenom is a revision of a 5 year old architecture. In one way it is a statement to how great the A64 was that AMD is able to compete with virtually the same architecture 5 years later...but that isn't going to earn AMD profit.

The biggest problem with AMD though, more than their archiecture, is the implementation of their ideas on actual silicon. AMD's process technology is simply nowhere near Intel's; this is to be expected of course, but it is really hurting AMD. AMD is hurting now with their 65nm process that offers lower frequencies than their 90nm process... hopefully 45nm will be better and Anandtech says that AMD appears confident with 45nm Phenom, but I'm not buying it until I see benchmarks and real products availabel for purchase. The problem with 45nm will of course be that intel is on HK + MG and AMD is using conventional materials. Certainly leakage on AMD's 45nm process will be higher than Intel's and AMD may run into power consumption problems preventing them from reaching higher frequencies.

Regardless, I've said enough. I just think you should realize how absurd the statements you are making are. I like AMD too, but pretending that they are doing well and comparing a $200 Phenom to a $1,000 Extreme Edition CPU is not going to get you anywhere.

[Loser777]

As for innovation... innovation is great, but not that many people care about what is in a CPU core, they care about how it performs.
Nice plagiarism of Mr. Paul Ottelini there... (Don't care if I spelled his name wrong)
I am buying an AMD CPU because I am a die-hard fan. The 9850s comparability with the Q6600 and the fact that Penryn is as Intel said themself "A transitional architecture for them to mature their 45nm process"-LGA 775 will die with Nehalem's launch, adds even more icing to the cake.
In Mid '09, I'll just laugh at all the people who bought a LGA 775 board during '08, only to have their formerly "high-end" QX9770 turn into a paper-weight by Nehalem. I'll cruise through the year with Bulldozer- hey even if it turns out to be another Core 2 Quad vs. Phenom scenario performance wise, at least I won't have to pay through the roof to get SIMILAR performance.

[Extelleron]

Nice plagiarism of Mr. Paul Ottelini there... (Don't care if I spelled his name wrong)
I am buying an AMD CPU because I am a die-hard fan. The 9850s comparability with the Q6600 and the fact that Penryn is as Intel said themself "A transitional architecture for them to mature their 45nm process"-LGA 775 will die with Nehalem's launch, adds even more icing to the cake.
In Mid '09, I'll just laugh at all the people who bought a LGA 775 board during '08, only to have their formerly "high-end" QX9770 turn into a paper-weight by Nehalem. I'll cruise through the year with Bulldozer- hey even if it turns out to be another Core 2 Quad vs. Phenom scenario performance wise, at least I won't have to pay through the roof to get SIMILAR performance.

I don't know if that is something that Mr. Otellini has said, but it is 100&#37; true. Very few people care about what is in their CPU, they care about: performance, power consumption, and in the case of enthusiasts, overclocking ability.

Anyone buying a QX9770 right now will be the people buying Nehalem when it launches in Q4 of this year, so your argument doesn't work so well there. Regular people who bought a Q6600 or Q9450 will simply upgrade as everyone does.

I don't think you can say that buying either AM2+ or LGA 775 is a better option... AM2+ will be supported longer, into 2009, but I doubt any of AMD's 45nm CPUs will offer superior performance to an overclocked QX9770. And Bulldozer is a new socket, as Anandtech's article from July '07 stated. Since Bulldozer is a fairly big change, it requires a new socket.

At this point I'm not sure Bulldozer is 2009 anymore either... I hope so, but AMD's roadmaps shown in their December '07 analyst day seemed to suggest Socket AM3 + 45nm "K10.5" would be the best '09 will have to offer, with Montreal core octal-core CPUs.

TBH I just see AMD becoming a different type of company in the future... I don't think they are going to challenge Intel for the performance crown again and I think that they will become a budget CPU manufacturer again. AMD just doesn't have the R&D money to combat Intel and at this point, I think Intel is unstoppable. They got really lazy with P4, but now they have committed themselves to a schedule of new core/new silicon every year, and AMD just can't keep up with that. AMD is a year behind on silicon technology and I don't think they will be able to launch a new core every 2 years with a limited R&D budget.

[SocketMan]

One of the comments from Xbit's review


Should have super sized it right away.:D

This is not a comment I have left at Xbits, just one of the few that didn't say the usual :Intel rocks,AMD sucks and so on...

http://xbitlabs.com/discussion/4815.html

As for my "gaming" opinion it's like this:
better to have an average cpu and exelent gpu than the other way around.

[retro77]

Wow, I wish Tom's Hardware readers were this mature about AMD.

All you see over there is Intel-fanboys "defending" the Intel name. I have been a long time AMD fan [ever since they fixed the overheating issues back in '97] and still using my current 939 3000+.

Next is a 5000+BE!


Thanks guys, I continue to read and enjoy.

[JumpingJack]

The graphs you posted also show up the effect of TLB-fix in those applications.

But not the B3, this was supposedly corrected.

[Suosaaski]

But not the B3, this was supposedly corrected.

My point exactly, just by comparing the results for 9550 and 9500.

[madfaze]

AMD fixes the Phenom X4 processor with the release of the 50-series, but launches the Phenom X3 processor series with B2 steppings that have the same issuses as the old Phenoms. Strange huh?


:down:

[cdawall]

Well I forgot who did it but they got the 9850BE to 3.0Ghz stable; took almost 1.5v and I think it was on air too. I'm sure it would be possible, but for a 24/7 OC I would set at the highest stable clock on the lowest voltage, which looks to be about 2.8-2.9 Ghz.

With a good lap and H20 I'm sure 3.0Ghz would be capable on a slightly lower voltage.

Edit: It was Tech Report


...with an MSI K9A2. :rolleyes: We should wait for a test with DFI LP UT 790FX-M2R. :D

hey notice that the guy only changed the multi so he left out an entire section of oc'ing

[JumpingJack]

My point exactly, just by comparing the results for 9550 and 9500.

Gotcha, missed the context.

[Jakalwarrior]

While you laugh at them spending money to upgrade their entire intel system while you just drop in a chip, they are lauging at you because even if they didnt upgrade at all, they would still be faster than you for another year. By the time AMD makes something with as big of a leap youll have to change boards too.

Its not that AMD keeps compatibility for ages, its that we arent going anywhere! and intel is speeding ahead so yes they do have to buy new boards all the time. I'm sure often times it could have been avoided with some extra work on compatibility but intel would rather push ahead and ignore that. More proffits, less wasted time that way. Maybe not the most moral... but considering you can buy an intel now and still be ahead of AMD in over a year......

[retro77]

Who cares if the system charts as being the fastest? What really, should, matter is how the system performs for you.

[madfaze]

amen!

[Jakalwarrior]

well most tests are meant to be an objective measure of how a system performs. If a system is bombing most normal tests, but for some reason it performs great for you... You need to figure up some metric to measure whatever other kind of speed it is that you are feeling. Otherwise its all in your head ;)

[Sparky]

suggestion:

Let's not turn this into YAAVIT (Yet Another AMD Vs Intel Thread).

[Morais]

Good performer cpu...too bad power consumption still high (higher than B2 :()

[tictac]

TLB patch reduce the B2 power consumption. B3 running at full blast ;) No wonder it have higher power consumption

[B.E.E.F.]

While you laugh at them spending money to upgrade their entire intel system while you just drop in a chip, they are lauging at you because even if they didnt upgrade at all, they would still be faster than you for another year.

To me its most important to have performance/$. I could care less Intel is coming out with the Core2Duo Xtra-Xtreme 10000 if its ridiculously expensive.

The new Phenoms seem to have much better real-life performance than the Q6600 at the same price point. :clap: :up:



I have some questions though. The Phenoms have a 125W TDP. Is that peak, or average? And that is thermal dissipation, not power consumption? I also heard that AMD TDP =/= Intel TDP because they changed the formula. Is this true?

[madfaze]

anyone has it yet?!

[jas420221]

To me its most important to have performance/$. I could care less Intel is coming out with the Core2Duo Xtra-Xtreme 10000 if its ridiculously expensive.

The new Phenoms seem to have much better real-life performance than the Q6600 at the same price point. :clap: :up:



I have some questions though. The Phenoms have a 125W TDP. Is that peak, or average? And that is thermal dissipation, not power consumption? I also heard that AMD TDP =/= Intel TDP because they changed the formula. Is this true?It would take a more expensive Phenom to equal the performance of the Q6600. Q6600's can be found for $199, but newegg has them for $249.99. Current Phenoms at newegg (not teh b3 ones), show $219 for a 2.3Ghz Phenom. Q6600 is faster at that price point by Pure speed (2.4 vs 2.3Ghz) and by clock for clock. ;)

[Brother Esau]

I don't care what the graphs say Phenom is a really nice CPU:up:

[madfaze]

intel fanboi...:down:

[jas420221]

intel fanboi...:down:Spam/Trolling posts... :down:












;) :clap: :rofl: :ROTF:

[SocketMan]

Let me see if I can get a point across:)

I'll try to keep it simple:p:
Intel has large cache everyone knows this.
What's more important is that it is "a shared" cache.
Now, someone runs some benchmarks 1 at a time if it's single threaded
then the "loaded" core gets access to that large (say 12Mb;) ) cache.
It helps quite a bit of course. I have not seen a website that would run 2,3 or more benchmarks at the same time. What do you think would happen when 4 cores running say 4 benchmarks (at the same time) will start competing
for that large, juicy cache?
I don't have a "current" Intel rig, so can not test, but, the logic tells me things would slow down ,to say the least.

With AMD it's different:the caches are "exclusive" for each core,so there wont be as much "cache competition" between the cores.
Having HT and not FSB would help as well;)
Now a million dollar question, does this make any sense or should I go see a doctor?:rolleyes: :p:

What I am getting at is when someone does very heavy multytasking: burning/converting/encoding etc and play a game on top of it ,will a Phenom based system be actually faster then Core2 ?

[retro77]

The benchmarks need to be re-written to use all cores then....

[K20]

And Bulldozer is a new socket, as Anandtech's article from July '07 stated. Since Bulldozer is a fairly big change, it requires a new socket.
The Anandtech article you mentioned referred to the server implementation of Bulldozer, the desktop version should use AM3 which is backwards/forwards compatible (to a degree) with AM2+.

AMD is a year behind on silicon technology and I don't think they will be able to launch a new core every 2 years with a limited R&D budget.
The article also mentioned 'Pipe' (Platform Innovation Progression) which is AMDs better sounding version of tick-tock where it also releases a new platform every 2 years rather than annually.


The benchmarks need to be re-written to use all cores then....

No, that's multithreading, the same data would still be in the cache. Reviewers need merely to run atleast 4 benchmarks at the same time.


Now the quoting's over with :clap: AMD :clap: . Not about the B3 but because of the return to X4, X3 etc. unfortunately they still have model numbers (please somebody tell me the connection between X4 98x0 and 2.5 GHz), I have a solution but in order for it to work I need to know if Hypertransport speed is locked to the Northbridge speed or if they're independent. I'm sure the answer is somewhere on this forum but I'd prefer not to spend the next week trawling through very long treads in order to find it.

[Oese]

NB and HT clocks are independent, but never should your ht run higher than your nb (or was it the other way round??)

good point with cache dedication.. and its common knowledge as well: without that large cache intels core microarchitecture would loose a lot performance.. no wonder super pi is that fast on a c2d, as all data fits in cache.. on amd its very memory dependent..

[aGeoM]

NB and HT clocks are independent, but never should your ht run higher than your nb (or was it the other way round??)...

Yes, HT<=NB speed.

I need start saving some $ for the 9850BE :D

[K20]

NB and HT clocks are independent, but never should your ht run higher than your nb (or was it the other way round??)

Thanks. Well I say thanks, now I need a new naming system :(. Unless it's the same case as core speed; AMD could set a processor to have its cores running at say 1GHz, 2.4GHz or 2.4GHz, 2.4GHz, 2.4GHz, 1.2GHz or even have each core running at a different speed but for arguably obvious reasons they don't.
So would AMD ever sell a processor with the Hypertransport link and NB running at different clocks?

[JumpingJack]

Let me see if I can get a point across:)

I'll try to keep it simple:p:
Intel has large cache everyone knows this.
What's more important is that it is "a shared" cache.
Now, someone runs some benchmarks 1 at a time if it's single threaded
then the "loaded" core gets access to that large (say 12Mb;) ) cache.
It helps quite a bit of course. I have not seen a website that would run 2,3 or more benchmarks at the same time. What do you think would happen when 4 cores running say 4 benchmarks (at the same time) will start competing
for that large, juicy cache?
I don't have a "current" Intel rig, so can not test, but, the logic tells me things would slow down ,to say the least.

With AMD it's different:the caches are "exclusive" for each core,so there wont be as much "cache competition" between the cores.
Having HT and not FSB would help as well;)
Now a million dollar question, does this make any sense or should I go see a doctor?:rolleyes: :p:

What I am getting at is when someone does very heavy multytasking: burning/converting/encoding etc and play a game on top of it ,will a Phenom based system be actually faster then Core2 ?

What you are referring to here is called cache thrashing, but it is not exclusive to shared caches... thrashing can occur in any case of shared resources where fair sharing is not appropriately applied.

There is one general feature defining the difference between inclusive and exclusvie caches, it has nothing to do with fair sharing rules. Copies of data between L1, L2, (and when L3 is present) L3 are guaranteed to be in all three caches in an inclusive design, where as in an exclusive design this is not true.

There are advantages and disadvantages to such methods for managing caches, one advantage is that in multicore systems when the caches are descrete with the lowest level shared, a miss in one cache means none of the caches have that data for an inclusive design (EDIT-- clarify -- down to the lower levels) ... the processor then knows to go directly to main memory to retrieve the data/code. In an exclusive cache design, like Phenom's, a miss means all caches must be snooped to find the most relevant version of the data (in fact, it is this snooping and retrieving that is at the heart of the TLB errata that causes the race condition). In the exclusive case, cycles are burned while all discrete cache pools are snooped.

Switching back to your point ... Intel's shared L2 cache dynamically allocates the shared resource between the two cores depending on loading, so Intel has implemented some form of fair sharing rules to avoid cache thrashing, furthermore, thrashing is completely avoided in fully associated caches, but is very prone in direct mapped caches. Problem with direct map caches is they are much more resource sucking to manage, so a nice compromise is to use n-set associative caches. Though it does not guarantee cache thrashing will not occur, it does diminish the odds that it will.


Finally, to answer you question ... here is a a bench run in multitasking between B2 Phenom and Q6600:
http://www.hexus.net/content/item.php?item=10427&page=12

Your logic is not incorrect, things will slow down just not as much as you think... and only so much so that the cache pool utilization negates the effectiveness a large cache provides in removing bus traffic. However, Intel will only suffer if the cache is so overwhelmed such that the bus traffic also overwhelms the available BW... this is not such a huge problem in DT... but one reason why AMD still shows competitive performance in server with high BW demanding apps.

Jack

[JumpingJack]

TLB patch reduce the B2 power consumption. B3 running at full blast ;) No wonder it have higher power consumption

Yes, the B2 patch effectively turns off the TLB. Though the TLB bug was majorly unfortunate for AMD at this time, for people really interested in CPU and how they work, provides oodles of good info... running software with TLB on/off really gives you some insight into how certain programs function, this is a good benefit of B2 Phenom in my opinion... and B2 is perfectly stable for 99.99&#37; of DT workloads you will throw at it so I also have a great folder with the TLB patch turned off :) .. and I only say 0.01% simply because I would assume there are some rare circumstances where the bug may occur.. I have not observed it yet.

http://www.soe.ucsc.edu/~renau/docs/isca07.pdf

Fascinating article, table 2 ... total power, the DTLB1 and 0 combine to consume/dissipate about 8% of the total CPU power (estimated from the thermal modeling/measured results).

[Jethro]

Excited the new B-3's and triples are coming out. Cant believe how much energy folks waste arguing points of no relevence to reality. Such sadly misplaced energy, what a shame. Sick of folks feeling compelled to piss on every single AMD thread i must say. Makes me wanna stay AMD even more.

I have NO reason to upgrade except im getting good money for my current setup and as such wouldnt mind playing with phenom for fun. Bound to be at least as fast as my current rig + a few more folding pnts at idle. So now to decide between B3 quad or a B2 triple...hmmm Really leaning towards b-3's clock speed potential i suppose.

[Oese]

Thanks. Well I say thanks, now I need a new naming system :(. Unless it's the same case as core speed; AMD could set a processor to have its cores running at say 1GHz, 2.4GHz or 2.4GHz, 2.4GHz, 2.4GHz, 1.2GHz or even have each core running at a different speed but for arguably obvious reasons they don't.
So would AMD ever sell a processor with the Hypertransport link and NB running at different clocks?

no but its needed for oc attempts to have the ability to adjust both of them....

[Loser777]

If a new 9850 BE can match a Q6600 or a QX6850, at least somewhat, I'd be satisfied. Hell, I bought my X2 4200+ during the first round of price cuts..., while I could have bought an E6300 or E6400 system.
I stopped buying Intel parts, when one day: I went to a friend's house, and used his comp... I was liek OMG UR COMP IS SO FAST WTF... what processor u usin? He said he was on a 1.6GHz Sempron... My 2.66GHz Celeron... couldn't compare.

[keithlm]

What do you think would happen when 4 cores running say 4 benchmarks (at the same time) will start competing
for that large, juicy cache?


I also have the same question as you... I want to see some of the single threaded benchmarks that get cited a lot get run while 3 copies of Prime95 are being run on the "unused" processors. (Or maybe 2 copies of Prime95 if the game or benchmark is multi-threaded.)

Many people will cite multi-threaded benchmarks. However most multi-threaded benchmarks have threads that "play well" together... so benchmark results of one multi-threaded application will NOT be the same as results from multiple applications that might be in "contention" for the same system resources.

Perhaps my example of Prime95 should be modified... to find some different applications that will purposely stress different parts of the system. What happens if one application uses all the memory bandwidth... while another does a lot of disk activity and perhaps a third to ensure the bus is saturated. In those conditions how will a benchmark run? (And the situation needs to be easily reproduced. How about a virus scan that continually scans a HUGE compressed file.)

This type of test would not really be just testing the CPU. It would also be testing the motherboard's chipset and other system components.

But a benchmark that would do these things could show much more than just running F.E.A.R. at 3 different resolutions and then declaring that the CPU that runs at 85 FPS is clearly the winner over the one the CPU that ran at 79 FPS. (A car might easily go 140 mph while another might only be able to go 110 mph. But if the car going in 110 mph is in complete control and the one going 140 mph is basically just a missile with no control.. what is "better"? Perhaps the mph "measurement" used as a benchmark in this situation doesn't tell the entire story. What if that car can't go 140mph if someone turns on the radio?)

I personally believe that more complicated benchmarking methodologies will definitely show that taking a CPU and adding a ton of cache to get more "speed" really doesn't succeed. (And I also disagree with all the people that advocate that "more cache is always better.")

[Extelleron]

I also have the same question as you... I want to see some of the single threaded benchmarks that get cited a lot get run while 3 copies of Prime95 are being run on the "unused" processors. (Or maybe 2 copies of Prime95 if the game or benchmark is multi-threaded.)

Many people will cite multi-threaded benchmarks. However most multi-threaded benchmarks have threads that "play well" together... so benchmark results of one multi-threaded application will NOT be the same as results from multiple applications that might be in "contention" for the same system resources.

Perhaps my example of Prime95 should be modified... to find some different applications that will purposely stress different parts of the system. What happens if one application uses all the memory bandwidth... can another do a lot of disk activity and perhaps a third to ensure the bus is saturated. In those conditions how will a benchmark run? (And the situation needs to be easily reproduced. How about a virus scan that continually scans a HUGE compressed file.)

This type of test would not really be just testing the CPU. It would also be testing the motherboard's chipset and other system components.

But a benchmark that would do these things could show much more than just running F.E.A.R. at 3 different resolutions and then declaring that the CPU that runs at 85 FPS is clearly the winner over the one the CPU that ran at 79 FPS.

I personally believe that more complicated benchmarking methodologies will definitely show that taking a CPU and adding a ton of cache to get more "speed" really doesn't succeed. (And I also disagree with all the people that advocate that "more cache is always better.")

More cache is always better for the consumer, unless cache latency is increased which with Intel's current CPUs it is not. The only disadvantage to having more cache is increased die size, which results in higher cost for the manufacturer but is irrelevant to the consumer.

The reason that Intel CPUs have more cache is simple - AMD's on-die memory controller allows for faster access to memory, which means that high-speed memory on the CPU die is less effective (although still important). But Intel CPUs see a bigger benefit from increased cache, and their manufacturing advantage allows them to continually put more cache on die than AMD.

I don't know exactly what you are trying to get at...there are programs that test for certain parts of the system, such as memory bandwidth. Those are the kind of programs where AMD has an advantage because of the on die memory controller, although Phenom reduces that advantage because of its slow L3 cache increasing latency. Yet that kind of advantage does not translate into real performance, as we can see with A64 vs C2D. Any memory system advantage that Phenom might have is negated by the slow L3 + Intel having 3X the cache on die.

[JumpingJack]

I personally believe that more complicated benchmarking methodologies will definitely show that taking a CPU and adding a ton of cache to get more "speed" really doesn't succeed. (And I also disagree with all the people that advocate that "more cache is always better.")

The data exists.... several sites have cache information, several academic journal publish this type fo data.... it would productive if you spent some time learning about cache technology, the purpose and uses of such, you hypothesis is fundamentally incorrect.

Adding cache will always decrease the miss rate (or increase hit rate, same thing), less misses means more performance ... the amount of speed up, if you will, decrease geometrically as cache size increases depending on the size of the working set -- i.e. the magnitude of the benefit falls off, such that 256 KB to 512 KB gives a higher percentage better performance than going from say 1 Meg to 2 meg. But it is always true, all things being equal more cache is better for hit or miss rate. Where this falls apart is latency increases with cache size, to there is a trade off where increasing cache becomes dead weight this is only true when the efficiency falls below that of a call to main memory, and this then is the case where all things are not equal. To reach that you would need probaly gigs of cache.

Architects can also improve hit/miss rates via organizational algorithms, as well as implementing advanced prefetching.

http://lwn.net/Articles/252125/ Midway down shows effectively cache miss rate for cache size.... 16 Meg is still an improvement.

http://www-hydra.stanford.edu/publications/ISCA94.pdf
http://www.ai.mit.edu/projects/oxygen/oxygen-book2001/Section3-Systems/Sys-06.pdf
http://csg.csail.mit.edu/pubs/memos/Memo-430/memo-430.pdf

jack

[keithlm]

More cache is always better for the consumer....

...I don't know exactly what you are trying to get at...


I respect your right to your opinion. I do not believe you to be correct. Adding more cache to get speed is not always the best practice. You are correct if you value single applications and don't really run many applications at one time. (In which case a single core CPU is probably good enough.)

BOTTOM LINE: I'll take a more efficiently designed small cache over a huge cache that is not as efficient.

===

What I am pointing out is that currently benchmarks are run in artificial "perfect" conditions. What happens when these "perfect" conditions are not so perfect anymore? What happens to you in the middle of a game if your virus scanner kicks off. How much do you suffer? What if you need to run a database server, a web server, a job server, and a on-demand virus scan program. And you also want to be able to actually use the machine to do other things at the same time.

Current benchmarks do not tell the entire story since they don't really tell us how the system performs in real world situations.

BTW: This is something I realized when I had an "epiphany" a long time ago while running 2 copies of Prime95 and I kicked of NVidia's "Cascades" demo. I figured it would bog down my dual core system in a major way. Most people, including me at the time, don't realize that the fairly boring and unremarkable "cascades" demo does all of it's processing on the GPU. When you realize that that demo becomes mind boggling. Although not really related... this did make me start questioning current benchmarking practices.

But most people are more interested in arguing semantics than actually discussing benchmarking methods that might reveal different aspects of various CPU.

[Lightman]

Looking thorough various reviews I've noticed that most single/dual threaded applications scores are lower than expected.
Explanation is simple - C'n'Q!
Every review is benching with it switched ON. That is AMD guideline of course.
This unfortunately is decreasing Phenom single/dual thread performance by 5&#37;-30% (from it's results with C'n'Q OFF). Worst case scenarios are gaming and badly coded single-threaded raytracers/encoders.
Reason is simple - independent core clocking.
This technology is simply too new for todays operating systems as they can't properly manage and distribute tasks across available cores.

[keithlm]

Adding cache will always decrease the miss rate (or increase hit rate, same thing), less misses means more performance ... the amount of speed up, if you will,

Theoretically you are correct. When working with a small number of applications or benchmarks more cache will "always" be beneficial.

If you concurrently run enough applications then the system WILL lose the benefit of the cache. It is not a question of "if" this happens it is a question of "at what point" will this happen.

Current benchmarks do not take this into account. (Imagine that...)


EDIT: This is what happens all the time with this particular discussion. People are so determined to convince me that more cache is "always better" that the focus is lost. The focus in this case is on WHAT HAPPENS to the benchmarks if they are run in a real world situation that could possibly remove all benefit of that cache. What then? What would those conditions reveal?

[JumpingJack]

Theoretically you are correct. When working with a small number of applications or benchmarks more cache will "always" be beneficial.

If you concurrently run enough applications then the system WILL lose the benefit of the cache. It is not a question of "if" this happens it is a question of "at what point" will this happen.

Current benchmarks do not take this into account. (Imagine that...)

Doesn't matter to the degree that code is code, contextual threads run and are fetched into cache with one thread or many threads.... the only trick is memory management, keeping it coherent.

Designing a benchmark to behave differently than that is simply rigging the test so to speak, it would not represent how it would be used.

Hexus did the test correctly and got the correct results
http://www.hexus.net/content/item.php?item=10427&page=12
This is more valid than what you are describing ... since it represents actual usage under available code base.

[keithlm]

Hexus did the test correctly and got the correct results

That is EXACTLY what I am describing. But we need MORE testing done like this. (But with more applications run to create desired situations. Cache starved, IO bottleneck etc.)

But the Hexus tests are definitely going in the right direction. We just need some kind of "standard" methodology that can be used.

[JumpingJack]

That is EXACTLY what I am describing. But we need MORE testing done like this. (But with more applications run to create desired situations. Cache starved, IO bottleneck etc.)

But the Hexus tests are definitely going in the right direction. We just need some kind of "standard" methodology that can be used.

Ok, we are coming to consensus ... I agree, there is only one or two sites that really probe this correctly and they only do it on rare occasions.

Jack

[KTE]

9650 isn't OEM only, you should be able to get them from retailers after 8th April. I know I can.

B3 might have been announced yet but I see no sign of them about here. Typically we get them after US, so I'm still waiting. Prices have been dropped today for Phenoms, the B3 to release are even lower retailed on opening than the older. You can see here (http://img81.imageshack.us/img81/9624/phenompricingxw2.jpg) prices are only listed after 7th April.

9850 like the rest of the Familly 10h CPUs, undoubtedly, is scaling higher than the clock frequency rise (http://www.legitreviews.com/article/682/13/), and that's due to the NB and MEM efficiency as the CPU speed rises mainly. Same if you compare any higher oc to stock on many memory intensive applications. Quite obvious just by running EVEREST single threaded memory/cache bandwidth tests.

suggestion:

Let's not turn this into YAAVIT (Yet Another AMD Vs Intel Thread).Yea man, exactly. :up:

[Lightman]

I saw one e-tailer asking only £144 (tax. included) for 9850BE!
Of course not in stock yet - available from 7/4 :)

That would be very right price for it. Anyway I will try to get first shipment of 9850BE from Scan as they are good to dealt with and not very expensive (£157).

[keithlm]

Ok, we are coming to consensus ... I agree, there is only one or two sites that really probe this correctly and they only do it on rare occasions.

What I personally want to see for myself is whether there actually IS a point where one design fails and the other succeeds OR if they both lose performance equally as adverse or "extreme" benchmarking conditions affect them.

If they both lose performance at equal rates but neither reach a point where performance drops quickly then that means that neither design is "better" or "worse". At that point then the only important thing to consider is the speed of the benchmarks. (And the theory that adding cache is always better would be correct.)

However, if one design reaches a point where performance drops dramatically during this "extreme" benchmarking and the other does not... that tells us that one design is better than the other and we might need to investigate further and possibly question the validity of the currently popular benchmarks which would have failed to reveal this situation.

Personally I would put my money on the AMD chip and predict that the Intel chip will lose performance in these conditions first: And I believe this will be directly because they depend so much on their cache for performance.

BTW: I believe this enough that I'll be buying an X4-9850 as soon as they are commonly available in a week or two. (And I'll run some extreme benchmarks as I have proposed to see if I can verify my hypothesis.)

[Jethro]

Lightman do you mean april the7th?

[JumpingJack]

What I personally want to see for myself is whether there actually IS a point where one design fails and the other succeeds OR if they both lose performance equally as adverse or "extreme" benchmarking conditions affect them.

If they both lose performance at equal rates but neither reach a point where performance drops quickly then that means that neither design is "better" or "worse". At that point then the only important thing to consider is the speed of the benchmarks. (And the theory that adding cache is always better would be correct.)

However, if one design reaches a point where performance drops dramatically during this "extreme" benchmarking and the other does not... that tells us that one design is better than the other and we might need to investigate further and possibly question the validity of the currently popular benchmarks which would have failed to reveal this situation.

Personally I would put my money on the AMD chip and predict that the Intel chip will lose performance in these conditions first: And I believe this will be directly because they depend so much on their cache for performance.

BTW: I believe this enough that I'll be buying an X4-9850 as soon as they are commonly available in a week or two. (And I'll run some extreme benchmarks as I have proposed to see if I can verify my hypothesis.)

It will be app dependent. You have that data already. FP_Rate from SPEC shows AMD pulling ahead signifcantly, INT_Rate from spec (where the cache is not overwhelmed) Intel lays it down.

Cache is really nothing more than moving the data closer to the core, Intel will show superior performance all the way up to the cache cannot satisfy the core without excessive hit rates such that the penalty to going over a slower bus causes sufficient drop to be beating by AMD.

AMD has smaller cache pools precisely due to the IMC, where they pay less penalty when needing to fetch from memory. AMD implementation of on IMC was novel, innovative and such... but make no mistake, it was also to save die size to compete with Intel while consistently staying 6 months to a year behind in the litho node. AMD touts how great they BW is, but fails to mention it is to make up for a smaller cache pool. (EDIT: I should be careful with the wording here... cache size is a design decision based on several factors, the IMC enable equivalent performance for poportionately small cache pools).

My money would be on Intel for Int heavy apps, AMD for high BW with large working sets (as seen in FP_rate) as reported by SPEC2006. The data exists to answer the questions you are asking... SPEC2006_rate (not base) is essentially this test and there are plenty of other studies.

EDIT: If doing the lit searches and such is too time consuming or you prefer the HW geek site type data:
http://it.anandtech.com/IT/showdoc.aspx?i=3099&p=7
http://it.anandtech.com/IT/showdoc.aspx?i=3099&p=8
http://it.anandtech.com/IT/showdoc.aspx?i=3099&p=9



Jack

[K20]

Prices have been dropped today for Phenoms, the B3 to release are even lower retailed on opening than the older. You can see here (http://img81.imageshack.us/img81/9624/phenompricingxw2.jpg) prices are only listed after 7th April.
Curious, the link Legit Reviews has links to a processor pricing page that's effective as of 27th March. The prices on that page are identical to the ones they claim are effective as of 7th April.

Lightman do you mean april the7th?
He does.
A better way to think of it might be the 7th of April.


In regards to this extreme/more accurate benchmarking could somebody please provide a list of suitable applications to be used for this.

[Nedjo]

Lightman do you mean april the7th?
yes! in Europe day is in first place, and month in the second... naturaly ;)

[Jethro]

Awesome news! Looks like i waited just long enough =) I'll be ordering one or the other as soon as they show up here! :cheers:

[SocketMan]

Isn't there supposed to be an "opteron" for the AM+ socket?
Maybe at a later date.:confused:

Best thing to do imo, for the "AMD VS Intel" compassions is to wait till AMD is on 45nm and Intel has their IMC on their "Nehalem".
Thanks for the links Jack.:up:

[JumpingJack]

Isn't there supposed to be an "opteron" for the AM+ socket?
Maybe at a later date.:confused:

Best thing to do imo, for the "AMD VS Intel" compassions is to wait till AMD is on 45nm and Intel has their IMC on their "Nehalem".
Thanks for the links Jack.:up:

Yep.. this will be fun to see.... I doubt it will quelch The Debate though. ;)

[purecain]

come on amd........ bring out a chip that overclocks and i can start building amd based pc's again....
its all here.... i have to say its not all bad, but in the face of the competition you cant help but will amd to do more......
http://news.firingsquad.com/hardware/amd_phenom_9850_black_edition_review/

[KTE]

Nope, as I said a while back, first delivery of B3 here is supposed to be 8th to the retailers, nothing earlier. Most places will have this delayed to 11th, I'll tell you in advance. AFAIK, for US the first delivery is the 2nd, nothing earlier. Pricing is effective from the 7th. You can see pricing here: http://www.amd.com/us-en/Corporate/VirtualPressRoom/0,,51_104_609,00.html?redir=CPPR01

This is how it compares on official tones:

http://img149.imageshack.us/img149/2070/pricingci5.jpg

Very noteworthy that US/CAN prices are not our UK/EU prices at all. In fact, the EU market pays usually x1.667 what US pays for the same product, and there is a larger gap here between the CPUs in pricing because of the widespread stronger currency value.

Yesterday many places have B2 Phenom dropped by $20 here, quite suddenly.
2 weeks ago they dropped the Q6600 G0 prices too.

Here's some gathering of reviews (some already posted):
http://www.pcper.com/article.php?aid=539&type=expert&pid=1
http://www.legitreviews.com/article/682/1/
http://techreport.com/articles.x/14424
http://www.xbitlabs.com/articles/cpu/display/phenom-x4-9850.html
http://www.sharkyextreme.com/hardware/cpu/article.php/3737046
http://www.hothardware.com/Articles/AMD_Phenom_X4_9850_B3_Revision/
http://www.firingsquad.com/hardware/amd_phenom_9850_black_edition_review/
http://www.hardocp.com/article.html?art=MTQ3OSwsLGhlbnRodXNpYXN0
http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=3272
http://www.maximumpc.com/article/amd_s_new_x4_quad_core
http://reviews.cnet.com/processors/amd-phenom-x4-9850/4505-3086_7-32908353.html
http://www.tomshardware.co.uk/Phenom-B3-Stepping-AMD,review-30476.html (hardly a review)

Needless to say many have outliers and "weird" results in between them but I haven't the time to waste on them if nothing is done about it by the authors, I'll have mine to look to instead of them. ;)

[GenTarkin]

What I am pointing out is that currently benchmarks are run in artificial "perfect" conditions. What happens when these "perfect" conditions are not so perfect anymore? What happens to you in the middle of a game if your virus scanner kicks off. How much do you suffer? What if you need to run a database server, a web server, a job server, and a on-demand virus scan program. And you also want to be able to actually use the machine to do other things at the same time.

Current benchmarks do not tell the entire story since they don't really tell us how the system performs in real world situations.

BTW: This is something I realized when I had an "epiphany" a long time ago while running 2 copies of Prime95 and I kicked of NVidia's "Cascades" demo. I figured it would bog down my dual core system in a major way. Most people, including me at the time, don't realize that the fairly boring and unremarkable "cascades" demo does all of it's processing on the GPU. When you realize that that demo becomes mind boggling. Although not really related... this did make me start questioning current benchmarking practices.

But most people are more interested in arguing semantics than actually discussing benchmarking methods that might reveal different aspects of various CPU


You know, after reading through this section of the thread and about this type of benchmarking...I always wondered why benches werent done like this more often on multi core CPU's....cuz like I think this truely shows how well a multicore cpu can cope with a system used by a person in which that system is doing various things at the same time of which can be interrupted often or have different priorities placed on the processes...or foreground and background switching of games and encoding programs and the like.
That sort of thing is just not compared anymore .... and like was pointed out in this thread... I think should be compared in reviews.
It says a lot for a system that can keep doing everything quickly and smoothly vs a system that can only handle a few things quickly and smoothly.

I dont know if in this type of benchmark would truely bring to light the quality / effectiveness of the architecture being reviewed or whatnot...but I still think its important to review differences of architectures in this manner....

[keiths]

e8400 $183.. If only.

[Jethro]

Anand says even the triples will be B3's within a month or so! That would be huge for budget clockers. $150 b3 triple! DROOL

[GenTarkin]

one question here...regarding the Core2 architecture vs AMD's

the core2 is a 4 execution core right?
Is amd's k10 still a 3 execution core?

If thats true...then AMD is really really close in terms of performance for only a 3 execution core.

Cuz IIRC this is one of the MAJOR things that gives core2 its massive IPC advantage over anything before it.

If AMD somehow made a 4 execution core, It would own the crap out of the core2 I would think....

[JumpingJack]

one question here...regarding the Core2 architecture vs AMD's

the core2 is a 4 execution core right?
Is amd's k10 still a 3 execution core?

If thats true...then AMD is really really close in terms of performance for only a 3 execution core.

Cuz IIRC this is one of the MAJOR things that gives core2 its massive IPC advantage over anything before it.

If AMD somehow made a 4 execution core, It would own the crap out of the core2 I would think....

Your terminology is a bit off and your understanding of how it works and what it means is not quite correct.

Core 2 is a 4 issue core, meaning it can dispatch 4 instructions per clock, with macrofusion (only capable in 32-bit mode) it can effectively issue as many as 5 instructions per clock.

K8 and K10 is a 3 issue core, same meaning, 3 instructions per clock. Going to a 4-issue core would not guarantee AMD would 'kick the crap out of it'.

Your misunderstanding is that the width != actual instructions issued over time. Think of it as burst mode.

Instruction level parallelism is not perfect, while Core 2 can issue up to 4 instructions per clock, rarely does it sustain this number of instructions dispatched into the execution units. This does not mean it was a waste, C2D can sustain more instructions in flight than K10 can simply because of the width. Likewise, a 3 issue core is not going to sustain 3 instructions per clock. This is because the parallelism that can be extracted from a singular stream of code posses enough interdependency such that in some cases only 2 instructions can go, in other cases only 1 or maybe 3. Time average IPC is actrually much less than the actual issue width. In fact IPC varies greatly with application and can run between as little as 0.8 instructions per clock on average to as many as 3 or 4 depending on width.

So, what Intel did was widen the core and added some logic to help with ILP, such as allowing loads to move ahead of stores and expanding IPC that way... AMD did not widen the core but they expanded the execution window to look for deeper reordering in the reorder buffer which increases the chances of finding instructions deeper in the reorder queue that can go through in parallel.

Because C2D is so wide, it has been commented that it is a much better candidate for SMT than what was netburst:


Core is wide enough that I can see hyperthreading returning to Intel's desktop and server processors fairly quickly. There's no question that hyperthreading is a good way to counter the wasteful effects of memory latency, and its addition to Core will yield even more performance per watt.

http://arstechnica.com/articles/paedia/cpu/core.ars/8

The upshot is you cannot make a generalized statement about IPC by simply taking the width of the core at face value.

http://www.ece.rochester.edu/~albonesi/wced03/papers/ekman.pdf (interesting paper in that they compare model and compare varying number of cores with issue widths such that Core x width = 16, and look at IPC overall cores for multithreaded software)

http://rsim.cs.uiuc.edu/~sasanka/UIUCTechRep.pdf

http://www.cs.princeton.edu/~skadron/papers/tradeoffs.pdf (this paper studies cache size, execution window size and such, and but the key here is to look at the IPC... very few apps get much above 2 issued instructions per clock).

http://www.ece.gatech.edu/academic/courses/fall2006/ece6100/Lectures/Module&#37;2014%20-%20Multicore/multicore.processor.examples.pdf Sorta summarizing key architectural details between Intel and AMD processors.

Jack

[Oldguy932]

I think I'm going to go for the 9750 95W edition if I can find one. Hopefully these B3's take well to the cold, I'll be really disapointed if I get the similar results to my B2's.

[GunterFalstaff]

this thread has veered way off course into an intel vs amd thread... why?? i started this thread specifically asking who plans on getting one and for discussion on the new B3 phenoms, i run an AMD system and ONLY an AMD system not because they are the best architecture/performance chips... but because i enjoy supporting a smaller company and i enjoy going against the grain as far as gaming rigs go... EVERYONE at my lanparties (roughly 20ish people) has an intel dual/quad system with an 8800 (all versions) video card, i am basically the only guy with an amd (or ati) system... can we please just discuss the new b3 cpu performance? Im frankly just not interested in cpu's that are incompatible with my system... thanks guys

[Jethro]

AMEN!

There really isnt any reason why we cant have fun with these chips w/o comparing em constantly, its so silly.

Im hoping to coax one to 3ghz myself =) Oldguy932 you appear to have plenty of experience clocking these guys, can u tell me if buying the unlocked cpu is necessary to see say 3gig? Provided i buy good ram and a decent clocking board? Also what are some of the clocks you have been able to achieve with the b2 silicon?

[Extelleron]

There's two things I'm wondering about with the Phenoms for those of you who have them...

For one, how easy it to overclock the NB on the Black Edition and what kind of frequency can you hit with it? Techreport reached 2.2GHz on the NB but I don't know if that was the max. Can the NB multiplier easily be changed in every board's BIOS or AOD?

And has anyone ever tried overclocking the HyperTransport frequency? I know with K8's it is supposed to run at a set frequency, 2000MHz, but K10's and the AM2+ motherboards are supposed to be able to run 5200MHz HT. I know from what I've read on here that HT speed can never exceed NB speed, so I suppose that means you would need a 2.6GHz NB to hit that frequency. I also know that Phenoms seem to be unable to hit HT (base) speeds of more than 260-270MHz, but I would assume there is a multiplier in the BIOS to adjust the HT multiplier (like with K8's, where you always have to adjust the multiplier down so HT won't exceed 2000MHz).

[JumpingJack]

There's two things I'm wondering about with the Phenoms for those of you who have them...


KTE is the man to ask.... and he has documented a great deal in this thread:
http://www.xtremesystems.org/forums/showthread.php?t=167447

This thread has more info on the Phenom than what you will find most anywhere on the net.

Jack

[SocketMan]

There's two things I'm wondering about with the Phenoms for those of you who have them...

For one, how easy it to overclock the NB on the Black Edition and what kind of frequency can you hit with it? Techreport reached 2.2GHz on the NB but I don't know if that was the max. Can the NB multiplier easily be changed in every board's BIOS or AOD?

And has anyone ever tried overclocking the HyperTransport frequency? I know with K8's it is supposed to run at a set frequency, 2000MHz, but K10's and the AM2+ motherboards are supposed to be able to run 5200MHz HT. I know from what I've read on here that HT speed can never exceed NB speed, so I suppose that means you would need a 2.6GHz NB to hit that frequency. I also know that Phenoms seem to be unable to hit HT (base) speeds of more than 260-270MHz, but I would assume there is a multiplier in the BIOS to adjust the HT multiplier (like with K8's, where you always have to adjust the multiplier down so HT won't exceed 2000MHz).

Jack is right, KTE is the man:yepp:
One thing to keep in mind is that KTE does not run a multigpu setup(s), and
thats were you'll see the difference in regards to HT frequency. I run my NB at 2400 Mhz (all the time) set in BIOS. It is not (on my Asus m3a) changeable with AOD in windows. To run HT @ 5200Mt/2.6Ghz you will need a voltage increase for the NB (it needs to be @ 2.6 as well)
I just run some quick Everest benchmarks and 3d Mark2006 with HT @ 1200 and 2400 (I can change the HT multy with AOD) there was almost no difference in the results:)

[freeloader]

this thread has veered way off course into an intel vs amd thread... why?? i started this thread specifically asking who plans on getting one and for discussion on the new B3 phenoms, i run an AMD system and ONLY an AMD system not because they are the best architecture/performance chips... but because i enjoy supporting a smaller company and i enjoy going against the grain as far as gaming rigs go... EVERYONE at my lanparties (roughly 20ish people) has an intel dual/quad system with an 8800 (all versions) video card, i am basically the only guy with an amd (or ati) system... can we please just discuss the new b3 cpu performance? Im frankly just not interested in cpu's that are incompatible with my system... thanks guys

I'm so damn tempted to purchase one from www.ncix.com this week. I have a 5000+ Black Edition that does 3.4ghz and I'm not sure if I want to part with it for the Phenom! :( 4 cores at 2.8ghz or 2 cores at 3.4.....hmmmm

[msimax]

do it do it:D

[freeloader]

Does anyone know where to purchase a 9850 in Canada, that actually has them in stock? LMK...

[quik_2_win]

Looks like Ben Sun from http://www.motherboards.org/ submitted a review for the 9850 BE. He posted numbers (images actually) @3.0 and 3.2 Ghz. -although he claimed 3.2 was unstable...pretty decent results regardless.
http://www.motherboards.org/reviews/hardware/1762_3.html

[Extelleron]

I'm so damn tempted to purchase one from www.ncix.com this week. I have a 5000+ Black Edition that does 3.4ghz and I'm not sure if I want to part with it for the Phenom! :( 4 cores at 2.8ghz or 2 cores at 3.4.....hmmmm

Well Phenom is ~20% faster clock-for-clock so that means it will probably be just as fast as your X2 5000+ BE, and 9850's seem to be able to hit 3GHz.

I wouldn't buy it from that site though... they want $292 CAD, when 1000ku is supposed to be $235 US. Judging from pricing on current Phenoms, the retail price will probably lower than the 1ku price as well.

[GunterFalstaff]

thats good, he did it on air too... good to hear! =)... i was worried i would have to bust out some water cooling to get to 3ghz (i know i would ruin my computer somehow installing one >.< lol ).

[freeloader]

Well Phenom is ~20% faster clock-for-clock so that means it will probably be just as fast as your X2 5000+ BE, and 9850's seem to be able to hit 3GHz.

I wouldn't buy it from that site though... they want $292 CAD, when 1000ku is supposed to be $235 US. Judging from pricing on current Phenoms, the retail price will probably lower than the 1ku price as well.

I've seen them for $260 CDN, but of course no availability. :(

[Brother Esau]

I'm so damn tempted to purchase one from www.ncix.com this week. I have a 5000+ Black Edition that does 3.4ghz and I'm not sure if I want to part with it for the Phenom! :( 4 cores at 2.8ghz or 2 cores at 3.4.....hmmmm

Actually Mike the Phenom is a Really nice CPU and pretty damn awesome at gaming just like my 6400+BE is that I just sold and I have no regrets about that although I really enjoyed it :yepp:

[freeloader]

Actually Mike the Phenom is a Really nice CPU and pretty damn awesome at gaming just like my 6400+BE is that I just sold and I have no regrets about that although I really enjoyed it :yepp:

Hey Campbell, how's it going? :)

My AMD system is currently a 5000+ BE at 3.3ghz. It's decent. I'm looking forward to the 9850 as it's the B3 stepping I've been waiting for. Just hate having to wait for it!

[SocketMan]

Have you noticed the HTT?!?:shocked:

Also 1.328V is very low (comparing to B2's) for 3ghz,
the very few B2 Phenoms that made 3Ghz needed much more then that.
I think april 7th is the B (3) day;)
We'll just have to suffer and save $$$ till then.



I could not believe how easy this kit H20 kit was to install
if you wanna try W/C this one is great, was on sale for about 120 I think

[GunterFalstaff]

i haven't even thought about touching pre-built kits because of all the bashing on them in these forums... i thought that you pretty much had to build one from the bottom up in order to have a good one? if that is indeed a good kit i might actually consider it =P, thanks socket =)

[tictac]

Looks like Ben Sun from http://www.motherboards.org/ submitted a review for the 9850 BE. He posted numbers (images actually) @3.0 and 3.2 Ghz. -although he claimed 3.2 was unstable...pretty decent results regardless.
http://www.motherboards.org/reviews/hardware/1762_3.html

that's a good result :up:

Now i'm tempted to buy some :hehe:
Wonder how high it will play with water or subzero cooling :idea:

[Brother Esau]

Hey Campbell, how's it going? :)

My AMD system is currently a 5000+ BE at 3.3ghz. It's decent. I'm looking forward to the 9850 as it's the B3 stepping I've been waiting for. Just hate having to wait for it!

Kinda been shaky past week Mike my Cat Marshall that I have had since I was 20 years old got really sick and almost died:( He's a good boy and my best friend and he's also 16 years old and has some major issues now with his health but I have been blessed and Marshall has recovered and done a complete turn around compared to 3 days ago when He was so sick I thought our time was up together:( It was quite scary and I had him on IV at the Vet and Emergency Animal Hospital for 1 1/2 days after they had to remove fluid from his lungs.

But he's good now and god willing the Big Man will watch over us and all will continue to be and do well;)

[GIZER]

i really want a 9850 but that will cost me quite a few bucks cuz i don't want to use it with my Asus mobo.

Question: is DDR2 800 enough for Phenom? or do i have to get 1066??

[Brother Esau]

Heres a Good Review on the New chip .........AMD Phenom 9850 X4 Black Edition CPU Review :: (http://www.motherboards.org/reviews/hardware/1762_1.html)

[Oese]

yeah i really think 3,2ghz will be possible with phenoms soon, at least under water...

[KTE]

Haven't had time to read the thread sorry (forgive), just too busy but here's a nice review. Alwaysthe best this site: http://translate.google.com/translate?hl=en&sl=de&u=http://www.computerbase.de/artikel/hardware/prozessoren/2008/test_elf_prozessoren_athlon_phenom/6/&sa=X&oi=translate&resnum=1&ct=result&prev=/search%3Fq%3Dhttp://www.computerbase.de/artikel/hardware/prozessoren/2008/test_elf_prozessoren_athlon_phenom/6/%26hl%3Den%26client%3Dfirefox-a%26rls%3Dorg.mozilla:en-US:official%26hs%3Dhug

[Undersea]

Haven't had time to read the thread sorry (forgive), just too busy but here's a nice review. Alwaysthe best this site: http://translate.google.com/translate?hl=en&sl=de&u=http://www.computerbase.de/artikel/hardware/prozessoren/2008/test_elf_prozessoren_athlon_phenom/6/&sa=X&oi=translate&resnum=1&ct=result&prev=/search%3Fq%3Dhttp://www.computerbase.de/artikel/hardware/prozessoren/2008/test_elf_prozessoren_athlon_phenom/6/%26hl%3Den%26client%3Dfirefox-a%26rls%3Dorg.mozilla:en-US:official%26hs%3Dhug

Wow,

Great link!

[Oldguy932]

I did get 3ghz with my regular 9600 on air, but I have the vapo setup so I can't go back and get pics of it. The htt doesn't run nearly as stable as high as it did under cold, and the processor hated any voltage above 1.325 set and also didn't get too far past 2.9. My 9500 and 9600 BE both hit 2.8ghz, but I finally experienced the overclock loss on the BE because I used to just run it at 2.6 and now it freezes after about 15 min at 2.6. I'll post my findings whenever I get the new one, it gives me something to do other than sleep through my last month or two of high school.

[SocketMan]

i haven't even thought about touching pre-built kits because of all the bashing on them in these forums... i thought that you pretty much had to build one from the bottom up in order to have a good one? if that is indeed a good kit i might actually consider it =P, thanks socket =)


http://www.vr-zone.com/?i=5142
http://www.tomshardware.com/2007/10/05/comparing_water_coolers/index.html
http://www.cluboc.net/reviews/super_cooling/swiftech/H20-120Compact/page3.asp

It wont be as good as some custom ones but:

1.It's super easy to set up (I am on the lazy side)
2. It's cheaper then custom (I am on a budget)
3. It's small (compact)
4. It get's the job done, the temperature on my opty 165 dropped over 10c,from Scythe Ninja (that's a big,big cooler)
Got mine here:
http://www.ncix.com/products/index.php?sku=25710&vpn=H20%2D120%2DCOMPACT&manufacture=Swiftech
BUT ON SALE: think it was 109 or 119 Cdn, thats a much better price.

[SocketMan]

i really want a 9850 but that will cost me quite a few bucks cuz i don't want to use it with my Asus mobo.

Question: is DDR2 800 enough for Phenom? or do i have to get 1066??


Yea it's enough, even ddr2-400(pc2-3200) will work
I'd say DDR2-800 might even be a better choice,as far as compatibility goes;)


These Opterons (that I've mentioned earlier) are not listed on AMD or almost anywhere else, yet seems like they (will) exist?
OS1352WBJ4BGHBOX
http://www.shopblt.com/cgi-bin/shop/shop.cgi?action=thispage&thispage=011003000502_BQ44194P.shtml&order_id=!ORDERID!

[SocketMan]

Haven't had time to read the thread sorry (forgive), just too busy but here's a nice review. Alwaysthe best this site: http://translate.google.com/translate?hl=en&sl=de&u=http://www.computerbase.de/artikel/hardware/prozessoren/2008/test_elf_prozessoren_athlon_phenom/6/&sa=X&oi=translate&resnum=1&ct=result&prev=/search%3Fq%3Dhttp://www.computerbase.de/artikel/hardware/prozessoren/2008/test_elf_prozessoren_athlon_phenom/6/%26hl%3Den%26client%3Dfirefox-a%26rls%3Dorg.mozilla:en-US:official%26hs%3Dhug


The charts are excellent, the % changes when you highlight a cpu in relations to other cpu's
Wow they went all out,using tons of benchmarks!
But forget about trying to read anything there:
Nevertheless teaches this last test Intel fear, the integrated memory controller of AMD can demonstrate his skill.:rofl:

[MotF Bane]

Haven't had time to read the thread sorry (forgive), just too busy but here's a nice review. Alwaysthe best this site: http://translate.google.com/translate?hl=en&sl=de&u=http://www.computerbase.de/artikel/hardware/prozessoren/2008/test_elf_prozessoren_athlon_phenom/6/&sa=X&oi=translate&resnum=1&ct=result&prev=/search%3Fq%3Dhttp://www.computerbase.de/artikel/hardware/prozessoren/2008/test_elf_prozessoren_athlon_phenom/6/%26hl%3Den%26client%3Dfirefox-a%26rls%3Dorg.mozilla:en-US:official%26hs%3Dhug

Interesting, a 9950.

[d412k5t412]

I think I have some bad news about the 9850. Not sure if its been covered yet, but I purchased my 9850 Black Edition from compsource. I just asked for an updated ETA on when its going to be shipped out. They told me not until 5/2! When i recently heard 4/7. Bad news:(

I hope they're wrong. I can't wait another month!!!! Anyone heard anything about it being delayed? If they're wrong I might just go somewhere else.

[Marvin_The_Martian]

I think http://www.xtremesystems.org/forums/showthread.php?t=183025 needs to be cross linked. Tony posted some (unstable) 3.5 clocks right here :D

[Hemi]

The egg has the 9550 in stock.

[Loser777]

I think http://www.xtremesystems.org/forums/showthread.php?t=183025 needs to be cross linked. Tony posted some (unstable) 3.5 clocks right here :D
Unstable?
He hasn't posted prime results, but it ran 3dm @ Cinebench fine...