Phenom and 3DMark

[macci]

Already posted at AMD section but thought it might fit better under this section. I think its pretty interesting.

We all know the clock situation :D

but how about at the same clocks? Here is an initial comparison:


Here is a bit better benchmark (ran under Vista32bit without tweaking) -- single card 3DMark05! :D

http://img.techpowerup.org/071203/3dm05_vista_19610_3gsinglecard.png

I have a reference score with 3GHz Yorkfield using the same GFX clocks and its 300 points less (ran under WinXP32bit though)...

Here is another Yorkfield comparison score at 3GHZ (this one is with OVCKed FSB & RAM) (by Sampsa);
http://plaza.fi/s/f/editor/images/yorkfield_x38_ddr_05.png

Pretty close match, eh? :D

[NeoForce]

Very good!
I hope AMD could return positions in TOP CPU benchmarks

[G-UNIT91]

If so Phenom will be much better than the Core 2 Q....

[dpa]

the fact that Yorkies benches at 4700Mhz on air... and Phenom struggles to do 3000Mhz is an important fact - and weird....

[ZoLKoRn]

yah it look very good and some result from i benched by Phenom 9600@2.875GHz and HD3870 CF default clock :D

[VictorWang]

problem is that....is there any 5Ghz+ phenom since Yorks almost going for 6Ghz 3D @.@

[RyderOCZ]

This is not a thread about overclocking potential

2 Quad Core CPU's running the same clock speed, with results using the same graphics card. Everyone was concerned about K10 performance, here is an example, plain and simple.

Yorkie even has an advantage in ram speed 1142 vs 1000 at the same timings.

[emoners]

interesting... :)

noob question, can the AMD Overdrive be used in other procs?:p:

[donitsi]

This is not a thread about overclocking potential

2 Quad Core CPU's running the same clock speed, with results using the same graphics card. Everyone was concerned about K10 performance, here is an example, plain and simple.

Yorkie even has an advantage in ram speed 1142 vs 1000 at the same timings.


+1 :up:
yes that's absolutely how it goes

[BeardyMan]

This is not a thread about overclocking potential

2 Quad Core CPU's running the same clock speed, with results using the same graphics card. Everyone was concerned about K10 performance, here is an example, plain and simple.

Yorkie even has an advantage in ram speed 1142 vs 1000 at the same timings.

Yup can't say it better, this should ease down on the hardcore intel fans :D

Thx for posting his macci :up:

[Athens[2004]]

Interest info ,

the sad thing is that no1 else except macci have ever seen 3ghz. . . so far as i know the average clocks are betwn 2.5-2.7 [?stable?] . PLUS the comparism done with an unreleased cpu.... [enginering sample aka the 3gigs comes only with 12 multi... right now.]

[donitsi]

;2604936']Interest info ,

the sad thing is that no1 else except macci have ever seen 3ghz. . . so far as i know the average clocks are betwn 2.5-2.7 [?stable?] . PLUS the comparism done with an unreleased cpu.... [enginering sample aka the 3gigs comes only with 12 multi... right now.]

They have seen 3g, but not with 2500mhz NB cloks ;)
I knew it was the achilles heel

[HousERaT]

Nice to know that clock for clock performance is similar, maybe even a tad better for AMD. This will be great for the normal community at large. Not gonna mean much of anything to an overclocker. I hope this ensures that AMD will be around for awhile. The consumer needs them around to give Intel some form of competition.

[demek]

what about feauture test? they are done on amd system and not on intel system.is this make any difference on score?

[PhilDoc]

Feature tests make no difference in scores. There should be a unlocked Phenom comming out and this should allow some 3G overclocking. Hopefully, they'll get the NB issues worked out and well see some of this from the locked versions.

[dpa]

why dont you do a 3dmark 2001 compare Macci..? This is only 3dmark 2005.., give us some other benches :)

[macci]

why dont you do a 3dmark 2001 compare Macci..? This is only 3dmark 2005.., give us some other benches :)
3DMark2001 is all about Cache speed - and it doesn't look pretty (as you can guess) :D
Same goes to SuperPi. They didn't have SuperPi or 3Dmark2001SE in mind when they built this chip...

As far as 3DMark06 goes: as long as both systems are running under Vista it should also be a pretty close match - I'll run some 06 later.

[jas420221]

Nice SYNTHETIC results... I hope that translates to real world FPS!!!

[nachthymnen]

Intrested to the '06 results:D :up:

[metro.cl]

i gues this doesnt have the microcode update for L3 cache issue.

Nice to see a comparison but just like intel and Super Pi, you cant only use 1 benchmark to compare.

[Zannema]

Intrested to the '06 results:D :up:
:yepp: :yepp: :yepp: :yepp:

[Bigchrome]

Would have thought the Yorksfield score would be better than that.
Can you try CF and see which wins then please?

[hipro5]

Ha, ha, ha......Sampsa just "lowered" his York score to much you.... :D :p:

Try it with a 2900XT that I have to compare at 860/1016... ;)

[BeardyMan]

Ha, ha, ha......Sampsa just "lowered" his York score to much you.... :D :p:

Try it with a 2900XT that I have to compare at 860/1016... ;)


Sure :D

right now Phenom > Yorkfield clock per clock baby :D

[teat]

New procesor for AMD is good only in 3D Mark ? :| what they heven't got results in super pi or c2q?

[ZoLKoRn]

New procesor for AMD is good only in 3D Mark ? :| what they heven't got results in super pi or c2q?

and 3D games too :D

[teat]

hm... Phenom is go in chipset NF4? :p:

[Morais]

Nice...NOW SHOW US THE QUADFIRE RESULTS !!! :D :D

[GazC]

I wonder what the comparison results would be like using Nvidia cards?

[IvanAndreevich]

Sure :D

right now Phenom > Yorkfield clock per clock baby :D

In your wildest dreams ;) I'd love for that to be the case. As it stands, AMD is pretty much out of consideration when it comes to overclocking.

[railer]

All respect to Macci. But dude works for DAMIT. What do you expect............ Time and time again it's been shown that Phenom looses clock for clock. And not like you can expect Phenom to do 4g+ on air. For OC'ers Phenom holds no value

[Harshal]

Nice to see you post, Macci
and good to know Phenom is not a DUD overall :D
one second thought, 3-4 years of R&D and only this???

[DStealth]

Ha, ha, ha......Sampsa just "lowered" his York score to much you.... :D :p:

Try it with a 2900XT that I have to compare at 860/1016... ;)


Ye thinking the same - seems AMD HT make 2 slower card to communicate each other better through the bus - ist's called chipset. That doesn't make even close, the highly overclocked Phenom seen to time VS stock York-ie, in CPU performance.
I would like to see all 3Dmark with Single 2900 1ghz/1ghz with 3ghz AMD compared to 4500mhz York-e with same voltage ;) Ye we know even year and a half Core2duo can outprerform AMD x4 in multithreading tasks when overclocked, not to talk about 3dmark and quad ;)
Come on guys these processors are like comparing apples to oranges ;)

[dinos22]

looks good there macci

[BeardyMan]

All respect to Macci. But dude works for DAMIT. What do you expect............ Time and time again it's been shown that Phenom looses clock for clock. And not like you can expect Phenom to do 4g+ on air. For OC'ers Phenom holds no value

and how come macci proved the opposite?

3ghz yorkfield with 1100mhz mem looses against a 3ghz phenom with 1000mem.

Selective reading is what fboys are good at it seems :D

[nachthymnen]

Have you read this guys?
http://techreport.com/discussions.x/13724

[Skratch]

Am I the only one that noticed that it only beat the core in the fairy benchmark?

Its easy to cherry pick a bench to favor a cpu,just like core kills in 01/pi why dont you run the same setup in 3d06 and see the out come.

We also dont know what drivers were used on the intel setup.

[Sampsa]

What a nice fight in AMD vs Intel discussion thread once again ;)

Macci, looks like our small tweaking battle made some people very frustrated. I'll push the Yorkfield system over 19610 in 3DMark05 tomorrow :)

Here are couple of things to consider when comparing the results:

QX9650 price: $999
Phenom 9600: $283

QX9650 scaling with air cooling: ~4,5 GHz
Phenom 9600 scaling with air cooling: ~3 GHz

3DMark is synthetic benchmark and in real world applications you will see different results, don't take 3DMark scores so serious. K10 architecture has some nice potential but with current clock speeds, scaling and L3 TLB issues Phenom has definitely been a disappointment. Hopefully new stepping will improve things in Q1/2008 and we will see close to 4 GHz results with better cooling.

[jas420221]

What a nice fight in AMD vs Intel discussion thread once again ;)

Macci, looks like our small tweaking battle made some people very frustrated. I'll push the Yorkfield system over 19610 in 3DMark05 tomorrow :)

Here are couple of things to consider when comparing the results:

QX9650 price: $999
Phenom 9600: $283

QX9650 scaling with air cooling: ~4,5 GHz
Phenom 9600 scaling with air cooling: ~3 GHz

3DMark is synthetic benchmark and in real world applications you will see different results, don't take 3DMark scores so serious. K10 architecture has some nice potential but with current clock speeds, scaling and L3 TLB issues Phenom has definitely been a disappointment. Hopefully new stepping will improve things in Q1/2008 and we will see close to 4 GHz results with better cooling.QFT...

Im sure once the rest of the Quad Penryn lineup comes out the pricing issue wont be anywhere near the difference it is now. A (ficticious) 2.3Ghz Penryn for $300 o/c to 3Ghz would look pretty tastey next to AMDs offering. :yepp:

Your last paragraph is spot on! :up:

[KTE]

Very nice work there. ;) A good comparison (if you leave out oc potential and expense). $$$ where it comes to it so we need a Q9450 compared.

[dinos22]

hehe Sampsa you are inviting fanbois now to completely derail the thread

[KeZzZu]

here's some cpu score testing with phenom 9500 @ 2.7 ghz nb @ near 2ghz
http://img508.imageshack.us/img508/9721/3dmarkqs8.jpg

[hipro5]

Heh, heh, heh.......I'm STILL waiting for macci to bench a 2k5 AND a 2k6 @ 3000MHz with an 2900XT @ 860/1016..... :p:

@ Sampsa......IF we get a E8400 which will be cheep enough for the 2k5 and play with it at 500MHz FSB, then come and tell me about prices and performance in this particular bench... ;) :p:

[BeardyMan]

Take a look at both results again , i did notice something ;) :D

[hipro5]

Take a look at both results again , i did notice something ;) :D

What? - what?..... :D

[dinos22]

single channel RAM on AMD rig???

[EnJoY]

single channel RAM on AMD rig???

No, it's 64bit+64bit unganged mode iirc.

[knightwolf654]

All respect to Macci. But dude works for DAMIT. What do you expect............ Time and time again it's been shown that Phenom looses clock for clock. And not like you can expect Phenom to do 4g+ on air. For OC'ers Phenom holds no value

:rofl: :ROTF:
fanboyism at its best :D

its not that you dont see many of them its more so it takes alot of time to o/c such a complicated cpu. macci was just the first one to it :)

[dinos22]

No, it's 64bit+64bit unganged mode iirc.

haven't heard of that before eek

[alexio]

And not like you can expect Phenom to do 4g+ on air. For OC'ers Phenom holds no value
I have to agree here and since this is XS....

K10 at it's current price point isn't a bad architecture for specific server users but as a desktop architecture for enthousiasts it's disapointing. I can't help feeling K10 is just an updated K8 architecture updated in ways not too interesting for desktop users. A quad core K8 wouldn't offer much less value than Phenom for 99% of desktop users. AMD has had enough time and funds to make something better than K10, so in my opinion they have failed.

[macci]

Heh, heh, heh.......I'm STILL waiting for macci to bench a 2k5 AND a 2k6 @ 3000MHz with an 2900XT @ 860/1016.....
whats the ref. score George? Right now I dont even have a 2900XT (512MB). but please do show the target :D I might be able to find a 2900XT (512MB)...

[hipro5]

whats the ref. score George? Right now I dont even have a 2900XT (512MB). but please do show the target :D I might be able to find a 2900XT (512MB)...

:D You first..... :p: 2k6 is ABOVE 13k for sure....and 2k5 above 20k.... :) Yes a 512MB one at 860/1016...:)

Ahhh......I see you "quoted" me without bold lettering and I don't like the letters now..... :( :D

[KTE]

Haha lol!

Anyway 3840 2.64GHz CPU score w/2k6 under XP 32-bit w/stock 2600XT. Which GPU you have matters on your total and CPU score, saw it with C2D/C2Q time and time again.

[BeardyMan]

What? - what?..... :D

Nothing special :D

You can see that in the system tests the yorkfield gets better numbers by more then a few fps, yet it fails in GT2 while both are running on same GFX card clocks :D

AMD the better optimized platform for 3D? i guess so :D

[Sampsa]

@ Sampsa......IF we get a E8400 which will be cheep enough for the 2k5 and play with it at 500MHz FSB, then come and tell me about prices and performance in this particular bench... ;) :p

I see no point to derail this any further and start to speculate what is coming in a couple of months from which manufacturer. Simple point of this thread was showing that Penryn and K10 architecture perform very similar in 3DMark05, nothing else :)

PS. 6.5x multiplier didn't seem to work with Yorkie and Maximus Formula so no improved result from me, macci stays ahead with Phenom :)

[bananax]

nvm!

[railer]

No need to get personal. We get the point of this thread. At same clock these two cpu's perform somewhat similar. 5% here or there do not make huge difference. What makes the difference is how far you can take them. Something tells me that a cheap Wolfdale @4.3 on air will be much more beneficial for gaming and 3D then Phenom @3g. ;)
I hope with time Phenom would scale way past 3G mark on air. Time will tell.....

[Movieman]

My apologies gentleman for the rude comments that were posted here by that "gentleman"
I've removed the posts, cleaned up the ones that related to them and asked that he be removed from XS..

[Hot'n'Chilly]

Great thread to whoever created it, hopefully this means that future Phenom chips will be fairly competitive. Which is good for everyone (competition = cheaper chips)

[mpilchfamily]

Why was a benchmark that is primarily for video cards used to compare results of 2 CPUs? Using 3D Mark 05 or any other 3D Mark benchmark tool is pointless and proves nothing. So why not run a few benchmarks that are more CPU dependent and run as many different benches as you can? Only then will we really see a good comparison and know for sure if they are a match clock for clock. So for now this is kind of a pointless thread till more benchmarks are demonstrated here.

Look i like AMD and i hope that there new lineup of CPUs is as good or better then Intel. We need more stiff competition between the 2 to promote faster development of new technology and to keep prices down. In a year or 2 when i do my next build i hope to go with AMD again. But only if they are at the best price per performance ration at that point in time.

[Movieman]

Gentlemen:
Please be respectfull of the XL's that come here and bring us information.
They have earned that respect over time with their accomplishments.
I said Please, but it's a demand more than a request.

[ElMoIsEviL]

Nothing I said was dis-respectful,

I see no one can question Macci on Xtreme Systems. I believe in Democracy and Republicanism.

I did not insult, simply called someone up on their behaviour which is not in line with what we, the overclocking community, preach.

It's about Ethics and a Code of Conduct as well as Transparency. These rules have governed our kin since the very beginning. Our distrust for the Media and large corporations has been our driving force in trying to reach high performance on cheaper products.

You of all people should know this Movieman seeing as you hail from New Hampshire.

[Movieman]

Nothing I said was dis-respectful,

I see no one can question Macci on Xtreme Systems. I believe in Democracy and Republicanism.

I did not insult, simply called someone up on their behaviour which is not in line with what we, the overclocking community, preach.

It's about Ethics and a Code of Conduct as well as Transparency. These rules have governed our kin since the very beginning. Our distrust for the Media and large corporations has been our driving force in trying to reach high performance on cheaper products.

You of all people should know this Movieman seeing as you hail from New Hampshire.

Parts of your post I considered rude to an XL here and I still do.
As I said above, these people have earned our respect over time and should be treated with respect.
I am not saying you don't have the right to post facts that counter anything they say but you should do so without comments dirrected at the person.
Now, I would like this ended and the discussion back on track.
Thank you.

[Morais]

macci, do you have any idea when the drivers for Tri-Quad crossfire will be released?

[sirheck]

I want to see 3dmark 06 results.
Its a lot more cpu intensive than 05.

[ElMoIsEviL]

I want to see 3dmark 06 results.
Its a lot more cpu intensive than 05.

And how's about 3DMark 03, PC Mark suite and actual game scores, like Quake 4, Enemy Territory, Doom3, Half Life 2, Crysis.

Hell, how's about Application tests, and maybe even Cinebench and SuperPi.

Something conrete.

[RyderOCZ]

Here is a thread with a a lot of K10 benchmarks: http://www.xtremesystems.org/forums/showthread.php?t=167820
Done by someone who doesn't work for AMD.

Now I am sure you will have no trouble finding Yorkfield examples here at XS to compare to :)

[Movieman]

Thread closed pending an Admin taking action to stop the attacks on one of our XL's name.
Thanks for your understanding..

[Movieman]

Problem child has been dealt with..thread open..

[mstp2009]

From the sidelines of this argument, I have just one question for everyone here:

Why are we comparing CPUs using a GPU benchmark (especially when the CPU sub-score is not given)?

Seems like a moot point if you ask me. :shrug:

[sirheck]

From the sidelines of this argument, I have just one question for everyone here:

Why are we comparing CPUs using a GPU benchmark (especially when the CPU sub-score is not given)?

Seems like a moot point if you ask me. :shrug:


EXactly.

[KTE]

Honestly. Because those interested and comparing are trying to see how well a specific GPU does paired with a specific CPU in a benchmark.

If GPU's/RAM/Chipset/OS is pretty much on par, the difference goes to show for the CPU in the bench. Pretty simple. That's the reason of the thread.

If you guys complaining want to just compare a CPU then 3DMark is the last app to do so and that's already been done too many times on this forum elswhere anyway. But those who compete in 3DMark scores all the time will be interested in regards Phenom/Core 2/Yorkfield in 2k1/2k5/2k6 and that's the reason and main audience of this thread.

[Lightman]

From the sidelines of this argument, I have just one question for everyone here:

Why are we comparing CPUs using a GPU benchmark (especially when the CPU sub-score is not given)?

Seems like a moot point if you ask me. :shrug:

Because this is 3D section? Because you can't compare CPU in games without GPU involvement?

Can't you all understand that BOTH Phenom and Yorkfield machines was running exactly same GFX card at the same clocks on the same driver?
What is different in that test from testing games on same GFX card with same settings and drivers?

Geez!!

I found this test informative and interesting. I can make my own conclusions from it, if someone can't then should stay away from this thread.

Thanks Macci! :)

[dinos22]

Dave

why ban ElMoIsEviL

what was so unreasonable in his posts that are up at the moment :confused:

There is nothing wrong with asking someone to provide some more information as one test is just not enough to really determine performance on new CPUs.....It's not unreasonable to ask that of macci considering he is Mr. AMD for us here and he should certainly need to provide more information. :shrug:. Being an employee of AMD he should be more transparent than a common user i think.

Dino

[mstp2009]

Honestly. Because those interested and comparing are trying to see how well a specific GPU does paired with a specific CPU in a benchmark.

If GPU's/RAM/Chipset/OS is pretty much on par, the difference goes to show for the CPU in the bench. Pretty simple. That's the reason of the thread.

If you guys complaining want to just compare a CPU then 3DMark is the last app to do so and that's already been done too many times on this forum elswhere anyway. But those who compete in 3DMark scores all the time will be interested in regards Phenom/Core 2/Yorkfield in 2k1/2k5/2k6 and that's the reason and main audience of this thread.

If that is the case, then why is the CPU score not given.

It seems that would eliminate a lot more of the controversy.

[mstp2009]

Because this is 3D section? Because you can't compare CPU in games without GPU involvement?

Can't you all understand that BOTH Phenom and Yorkfield machines was running exactly same GFX card at the same clocks on the same driver?
What is different in that test from testing games on same GFX card with same settings and drivers?

Geez!!

I found this test informative and interesting. I can make my own conclusions from it, if someone can't then should stay away from this thread.

Thanks Macci! :)

So basically, in a GPU benchmark, we are supposed to take home that this video card is NOT bound by the CPU.

Gee, that's . . . . pretty much expected in a GPU benchmark.

[KTE]

You can ask Sami for the CPU score. Nothing wrong with that at all and I don't mind seeing it too... :D But keep in mind platform does make a difference with benchmark results, sometimes a few hundred points easily.

[macci]

Why are we comparing CPUs using a GPU benchmark (especially when the CPU sub-score is not given)?
This is XS 3DMark section, here we compare 3DMark results, 3DMark05 is one of these benchmarks.
CPU score is not included in the overall score. And in case of 3DMark05 the CPU score is not considered as a stable or realistic benchmark anyways. 3DMark06 CPU score is much more relevant score. I will have that number up too (yorkfield should have the higher score in that one though).

I am sure you didnt know this but 3DMark05 is actually almost 100% CPU (not GPU) benchmark with these latest Gen GFX cards. THere is only 1 test (GT3, Canyon Flight) that is mainly GFX bound (in single adapter scenario, depending on CPU performance level). Under CrossFire or SLI its is 100% CPU benchmark.

And since the GFX configuration in the 1st post is EXACTLY the same we are just comparing CPU/Memory/Latency performance.

This question was already explained in the other thread (http://www.xtremesystems.org/forums/showpost.php?p=2605152&postcount=119).

To give a very clear example:

3G + 1GFX card = 20k
3G + 2GFX cards = 21.5k
5G + 1GFX card = 26k

its very easy to see where the bottleneck is

[arisythila]

Dave

why ban ElMoIsEviL

what was so unreasonable in his posts that are up at the moment :confused:

There is nothing wrong with asking someone to provide some more information as one test is just not enough to really determine performance on new CPUs.....It's not unreasonable to ask that of macci considering he is Mr. AMD for us here and he should certainly need to provide more information. :shrug:. Being an employee of AMD he should be more transparent than a common user i think.

Dino

Dino, I dont think you can question XIP's, XL's here. You usally get banned for being "rude" to them when you question their methods.


Meh.

[macci]

Again, this one is about Phenom/SPider platform and 3DMark.

its not about Phenom and Office'97 or SuperPi or CineBench. I am sure that there are various other threads for the other benchmarks.

this one is for the 3DMark guys (you know who you are), that's why its posted in the 3D section

Now, let's continue the 3GHz challenge! :D

George mentioned he has a 20k+ score with 3G Yorkfield and 2900XT at 860/1006. I will do my best to match that (first I need to get a 2900XT card though, should not be an impossible task) =)

[dinos22]

(first I need to get a 2900XT card though, should not be an impossible task) =)

:rofl: ROFL

[zakelwe]

Macci, nice work with the overclocking software, it seems widely praised. :toast:.

I would guess that these scores had a bit more "macci magic" added as well though...

Regards

Andy

[BeardyMan]

Again, this one is about Phenom/SPider platform and 3DMark.

its not about Phenom and Office'97 or SuperPi or CineBench. I am sure that there are various other threads for the other benchmarks.

this one is for the 3DMark guys (you know who you are), that's why its posted in the 3D section

Now, let's continue the 3GHz challenge! :D

George mentioned he has a 20k+ score with 3G Yorkfield and 2900XT at 860/1006. I will do my best to match that (first I need to get a 2900XT card though, should not be an impossible task) =)

yes go and beat George, he really really want's it :D

[BadNizze]

I´n not one of the "well known" overklocker her at XS or any other place for that mather, BUT as macci pointed out even I can se where the bottle neck is in 05...

Here is my Phenoms CPU score in 3Dmark06 for refernce!
http://i30.photobucket.com/albums/c338/Swe_Eric/06max.gif

I dont mean any harm to ANY of you guys, you are well known and respected by me and others. But cant you see the numbers...

[ryanmartini]

I´n not one of the "well known" overklocker her at XS or any other place for that mather, BUT as macci pointed out even I can se where the bottle neck is in 05...

Here is my Phenoms CPU score in 3Dmark06 for refernce!
http://i30.photobucket.com/albums/c338/Swe_Eric/06max.gif

I dont mean any harm to ANY of you guys, you are well known and respected by me and others. But cant you see the numbers...

Nice cpu score there for that clock! Oh if the Phenom would just clock! :( We would be seeing a different market in the oc crowd.

[MexxT]

I´n not one of the "well known" overklocker her at XS or any other place for that mather, BUT as macci pointed out even I can se where the bottle neck is in 05...

Here is my Phenoms CPU score in 3Dmark06 for refernce!
http://i30.photobucket.com/albums/c338/Swe_Eric/06max.gif

I dont mean any harm to ANY of you guys, you are well known and respected by me and others. But cant you see the numbers...

NO way Phenom is faster than Yorkfield Clock for clock!

when I saw your post I directly shut down IE7, clocked my CPU back to 2749MHz (to compair with you) and shutted down winamp + MSN Messenger. Nothing more.

I ran 3Dmark 06 and my CPU score:
3859 Marks

you keep in mind:
- Untweaked
- windows Vista (way more load for the CPU, XP - vista is about 100 points differance )
- Verry slow untweaked memory
- Older chipset (P965 P5B deluxe)

If i used XP and used the same memory timings + speed and a bios clock to 2750MHz I would definitely beat That phenom with a Q6600 !

Screenshot :
http://www.techzine.nl/f/g/35530php1CNCWS.jpg

Compair url:
http://service.futuremark.com/orb/projectdetails.jsp?projectType=14&projectId=4132994

[KeZzZu]

Yet you forget that we cant run nb faster than 2-2.25 ghz depends on cpu and mobo, we arent sure what causes it, but theres lot of tweaking for amd systems too eg i see theres one simple tweak what werent used on that run mentioning badnizzes screenie :) could be lot of more unused tweaks, ill try run with vista and all tweaks what i know. But lower nb speed.

[MexxT]

Yet you forget that we cant run nb faster than 2-2.25 ghz depends on cpu and mobo, we arent sure what causes it, but theres lot of tweaking for amd systems too eg i see theres one simple tweak what werent used on that run mentioning badnizzes screenie :) could be lot of more unused tweaks, ill try run with vista and all tweaks what i know. But lower nb speed.

If i had used XP i would score like 3950 or something with the same settings.
imagine a Yorkfield ;)

And i'm not familionair (how do you write that word?) with the Northbridge speed you are talking about.

Btw, my vista is untweaked. just: boot -> clock back -> run 3Dmark
And I Think you forgot my CPU is ''just'' a Q6600 ;)

[KeZzZu]

well there is magic uber boost on yorkfield thought it performs better clock to clock against older c2q, but you could crank your ram up to 500 with same latencys and then compare ^^

btw k10 nb speed affects on l3 mem speed. if im not mistaken :)

[MexxT]

well there is magic uber boost on yorkfield thought it performs better clock to clock against older c2q, but you could crank your ram up to 500 with same latencys and then compare ^^

btw k10 nb speed affects on l3 mem speed. if im not mistaken :)

I'll try tomorrow.
Don't know if my OCZ Flex XLC PC6400 (CL3) wil make it to 500MHz CL4.. Hope so!

I use 500 - 5-5-5-15 24/7

[Simps]

LOL @ ppl with bad atitude on this thread.

When Macci had the time to play / bench with highend hardware, he was ALWAYS n.1 in every benchmark.
He is probably the best player this game has ever seen. Show some respect.

You can desagree, and post your opinion, but don't post :banana::banana::banana::banana: here...

[]'s
Simps

[KTE]

MexxT: Your Q6600 CPU score is 40 odd points lower than Phenom 9500 at 2640MHz in XP 32-bit (untweaked OS). For reference: http://img70.imageshack.us/img70/6020/2k6pk4.jpg

I'll run it at higher MHz later.

Setup for the run attached.

[Simps]

@ KTE

C2Q @ 2.66Ghz / 266FSB on VISTA x86
CPU SCORE = 3938
http://service.futuremark.com/compare?3dm06=4119271

PS: on XP PRO my CPU score would be on 4300+...

[]'s
Simps

[KTE]

OK. There's thousands of you with G0 Q6600 but only a few here with Phenoms.

Try XP Pro 2k6 with Q6600 at 3Ghz or stock and post the score please.
I'd like to see XP Pro Q6600 score. Because Phenom didn't take that big a hit of 300-400 points from XP to Vista.

Name your GPU. Card makes a difference on your CPU score. That's why macci's score has good relevance since he kept the same card for both.
I have Vista Ultimate but too many problems with Vista so I'm not even going there.

[Simps]

@ KTE

C2Q @ 2.66Ghz / 266FSB (VGA is a GTX SLI stock)
CPU SCORE = 4234 on XP PRO
http://service.futuremark.com/compare?3dm06=2608911

This is old score from Jul/07, on 680i. That is from a friend.

I just don't have XP PRO instaled here... But I have benched there, and I get almost 4400 CPU score with 2.66Ghz / 266FSB on 680i (VGA is single Ultra)

It is sad, but true. At least with C2Q, the CPU score is about 300-400 points below on Vista :(

[]'s
Simps

[KTE]

I'll try Q6600 with the same GPU/RAM timings as the Phenom in XP Pro (same build) as soon as I get some things sorted out to compare better.
By the looks of it, it seems the C2 does worse in Vista only.

[STEvil]

Guys, York vs Phenom cpu performance discussion needs to go to a different thread.

This is Phenom vs. Intel quadcore 3d performance.

[FUGGER]

This is the 3D section and to have any results is good but to have results by Macci has high value here. Anyone may post any 3D results and they shouldnt be attacked for showing strongest bench, we get it and happy to see where the platform is at.

Showing disrespect will get you banned here, its not an ego.

[Praz]

Guys, York vs Phenom cpu performance discussion needs to go to a different thread.

This is Phenom vs. Intel quadcore 3d performance.
Then hopefully there will be some 3d scores posted. The entire thread contains two maybe three actual scores.

[Zytek_Fan]

Thanks for posting the results everybody! :D
Hopefully once B3 stepping is out we can see some killer performance :up:

[KTE]

Thanks for deleting the most stupid ad hominem post. :)

I might try getting the GPU in question and check the 2k6 scores when I do with Phenom (if they tie in with macci's). Can do for Q6600 too but not Yorkfield. I doubt it.

[BadNizze]

This is the 3D section and to have any results is good but to have results by Macci has high value here. Anyone may post any 3D results and they shouldnt be attacked for showing strongest bench, we get it and happy to see where the platform is at.

Showing disrespect will get you banned here, its not an ego.


Mayby time to delete all posts exept Maccis first post? Me and other have gotten this thread on the wrong foot so to say, I´ did not mean to start an new AMD Vs intel sand box dissusion. Again sorry macci! I hope you macci can maka e new thread or continue in this because I and others are very intrested in youre setup\results! Over and out!

[knightwolf654]

NO way Phenom is faster than Yorkfield Clock for clock!

when I saw your post I directly shut down IE7, clocked my CPU back to 2749MHz (to compair with you) and shutted down winamp + MSN Messenger. Nothing more.

I ran 3Dmark 06 and my CPU score:
3859 Marks

you keep in mind:
- Untweaked
- windows Vista (way more load for the CPU, XP - vista is about 100 points differance )
- Verry slow untweaked memory
- Older chipset (P965 P5B deluxe)

If i used XP and used the same memory timings + speed and a bios clock to 2750MHz I would definitely beat That phenom with a Q6600 !

Screenshot :
http://www.techzine.nl/f/g/35530php1CNCWS.jpg

Compair url:
http://service.futuremark.com/orb/projectdetails.jsp?projectType=14&projectId=4132994

he he that made me smile :)
i know this is 3d but i have just one question. is their is a bigger point difference with intel going from xp to vista then phenom?

[MexxT]

@ knightwolf654, I compared with a friend of mine. He uses a Q6600 @ 3,285Ghz and windows XP Pro score: 5188 . I use a Q6600 @ 3,6GHz with Vista, and my score: 5211 :x

Huge differance!

Oh, and I ran 3Dmark again, with Vista, and Bios clock + compairable memory speed.

http://www.techzine.nl/f/g/35530phpU0A99m.jpg

Q6600 is A LOT faster..

My advantage is:
-8MHz more memspeed.

My Disadvantage is:
- 5MHz less corespeed
- Cl5-5-5-15 VS CL4-4-4-12
- Vista VS XP

So, it looks clear to me :)

[KeZzZu]

haha nice! 10 more points in cpu test, eh but vista in use :/

[MexxT]

haha nice! 10 more points in cpu test, eh but vista in use :/

Vista = Crap with benchmarks ;)

[Simps]

If you search ORB for the highest vista score (3DMARK06), it will show something arround 20K LOL!

[]'s
Simps

[MexxT]

If you search ORB for the highest vista score (3DMARK06), it will show something arround 20K LOL!

[]'s
Simps

Now search XP ;)

[macci]

:D You first..... :p: 2k6 is ABOVE 13k for sure....and 2k5 above 20k.... :) Yes a 512MB one at 860/1016...:)
Ok, I found a 512MB 2900XT.

so, how much over 20k? :D

20k seems to be quite easy with 860/990 setting already - didnt need to push NB / RAM at all. Actually this was the first run I did with this config so far.

first run at 20k (http://img.techpowerup.org/071213/20015014.png)

edit: OK, just to be fair - here is a better run with higher NB / MEM

http://img.techpowerup.org/071213/20429.png

more details:
Vista ultimate 32bit, Catalyst 7.11 driver

[linhvndiy]

Wow,:D. Waiting Mr Hipro5.:D
Thanks Macci.

[mongoled]

Wow,:D. Waiting Mr Hipro5.:D
Thanks Macci.http://www.improvresourcecenter.com/mb/images/smilies/popcorn.gif
This looks like its going to turn into a fun thread!

Should be renamed

'3Ghz challenge of the titans, Macci V Hipro'

:D:D

[hipro5]

Heh, heh, heh.......In a minute..... :D

......and the 2k6 one?..... :p:

[Oliver]

Going to be really cool :).

Keep pushing it guys.

[gOJDO]

@macci Why don't you run the default benchmark(Game and CPU, without feature)?

It would be easier for us to compare the score with many others. Also, since the thread is about 3D Mark, it would be nice to run the rest of the 3D Marks.

[ahmad]

Heh, heh, heh.......In a minute..... :D

......and the 2k6 one?..... :p:

You are evil. :p:

[BeardyMan]

lol it seems like Hipro can't cutt it at same GPU, CPU en MEM clocks :D

that minute was two hours ago

[Sampsa]

Yes, looks like a minute is very long in Hellas ;)

[dinos22]

LOL hipro
you just got owned by macci dude

killer result there man wow :D

i hear a mate is running a phenom 9900@3Ghz ..... he'll be testing the coldbugs soon :) should be interesting

[BadNizze]

Very nice Macci! You put a smile in my face ;)

[knightwolf654]

macci it be freakin sweet if you could do a write up on how to overclock phenoms. :)

[purecain]

thanks for posting the phenom results, really interesting. seems AMD are on the right track at least... keep em coming....

[massman]

Macci, has you already tested for signs of coldbug?

[macci]

@macci Why don't you run the default benchmark(Game and CPU, without feature)?

It would be easier for us to compare the score with many others. Also, since the thread is about 3D Mark, it would be nice to run the rest of the 3D Marks.
people usually run ONLY the Game tests for 3Dmark05. that is considered as default by many. the CPU Score in 05 doesnt affect your overall score and as mentioned is not a stable or realistic benchmark anyways.
so the numbers you see in the screenshots are very easy to compare with others.

the reason I ran the feature test (mainly the fillrate) is to just show the theoretical GFX speed. Fill rate only depends on your GFX power (certain clock speeds can produce only certain amount of fillrate). it can be used as a verification of your GFX clocks (or GFX power in general).

Single fillrate = Memory bandwidth of the card (e.g. memory clocks and memory timings affect this one)

multi fillrate = GPU bandwidth (e.g. GPU clocks, this is a test that is only affected by the GPU clocks, so its easy to compare and verify that the other guy was running the same GPU clock)

(now if someone is paying a very close attention - comparing the 20k score vs the 20429 score you'll see that the fillrates are very different! I dont know what went wrong there!! :D , need to retest both settings again today, it seems that only the fillrate test is screwed since the Pixel shader test is the same, GT3 is different too but that could be because of higher PCIe clock and higher overall system performance in the 20429 score)

[itznfb]

such pointless arguing. the thread shows me that summer '08 may be a fun time :D

[macci]

massman, with the production parts: NO
with earlier parts: YES
it would boot up fine at -80C

I have a plan for the weekend ;)

[KeZzZu]

massman, with the production parts: NO
with earlier parts: YES
it would boot up fine at -80C

I have a plan for the weekend ;)
OT/
ermm you mean CPU or gpu's? if cpu, retails suffer from coldbug?
and little task/question: would you write little quide to tweak/oc phenom to amd section. becose there aren't many phenomist's around :( so were doing slow in learning how to :D / OT

Very nice score Macci :clap:

[massman]

massman, with the production parts: NO
with earlier parts: YES
it would boot up fine at -80C

I have a plan for the weekend ;)

You're kidding right? :D

With the right steppings, we should be able to use dice without having too much problems? If so, I want one :p:

[BeardyMan]

You're kidding right? :D

With the right steppings, we should be able to use dice without having too much problems? If so, I want one :p:

If that's the case full LN2 control :)

[Sampsa]

Ok, I found a 512MB 2900XT.

so, how much over 20k? :D

20k seems to be quite easy with 860/990 setting already - didnt need to push NB / RAM at all. Actually this was the first run I did with this config so far.

first run at 20k (http://img.techpowerup.org/071213/20015014.png)

edit: OK, just to be fair - here is a better run with higher NB / MEM

http://img.techpowerup.org/071213/20429.png

more details:
Vista ultimate 32bit, Catalyst 7.11 driver

macci,

http://plaza.fi/s/f/editor/images/qx9650_r600_3dmark05_20448.png

QX9650
Asus P5E3 WS Professional (Intel X38)
ATI Radeon HD 2900 XT (860/999)
OCZ PC3-12800 Platinum @ 1712 MHz 7-6-6-18 1T

Catalyst 7.11
Windows XP SP2

;)

[BeardyMan]

Lol Ddr3 ;) :d

[massman]

Sampsa: DDR2 vs DDR3 and XP vs Vista :)

[hipro5]

I´n not one of the "well known" overklocker her at XS or any other place for that mather, BUT as macci pointed out even I can se where the bottle neck is in 05...

Here is my Phenoms CPU score in 3Dmark06 for refernce!
http://i30.photobucket.com/albums/c338/Swe_Eric/06max.gif

I dont mean any harm to ANY of you guys, you are well known and respected by me and others. But cant you see the numbers...

Actually a nowadays York at 2750MHz does this..... :)

http://www.hipro-tech.com/images/hipro5/1_ASUS_P5E64_WS_Pro/2k6_2750_CPU_score.png


Ok, I found a 512MB 2900XT.

so, how much over 20k? :D

20k seems to be quite easy with 860/990 setting already - didnt need to push NB / RAM at all. Actually this was the first run I did with this config so far.

first run at 20k (http://img.techpowerup.org/071213/20015014.png)

edit: OK, just to be fair - here is a better run with higher NB / MEM

http://img.techpowerup.org/071213/20429.png

more details:
Vista ultimate 32bit, Catalyst 7.11 driver


Sorry for the delay but I totally FORGOT it coz I was benching the PCMark2005 today..... :D

http://www.hipro-tech.com/images/hipro5/ASUS_Blitz_Extreme/York_Retail/PCMark05_27183_5508MHz.png

Here you go...... :)

http://www.hipro-tech.com/images/hipro5/1_ASUS_P5E64_WS_Pro/2k5_3000MHz_20522_SINGLE_2900XT_860_1000_1.png

http://www.hipro-tech.com/images/hipro5/1_ASUS_P5E64_WS_Pro/2k5_3000MHz_20522_SINGLE_2900XT_860_1000_2.png

....and 2k6.....where's YOUR 2k6?.....I can't see it anywhere?..... :D :p:

http://www.hipro-tech.com/images/hipro5/1_ASUS_P5E64_WS_Pro/2k6_3000MHz_14020_SINGLE_2900XT_860_1000_1.png

http://www.hipro-tech.com/images/hipro5/1_ASUS_P5E64_WS_Pro/2k6_3000MHz_14020_SINGLE_2900XT_860_1000_2.png


@ Massman please DON'T compare systems each other by saying DDR2/DDR3 or so.....IF AMD rig whould have DDR3, the guys whould bench with DDR3.....Does it have?.... :D
Everyone benchies with what he has.... :) ;)

[dinos22]

massman, with the production parts: NO
with earlier parts: YES
it would boot up fine at -80C

I have a plan for the weekend ;)

what
no way
amd CPUs that can boot up at -80C......i gotta see that

[BeardyMan]

Hipro, you are not comparing fair :)
i'm afraid you can't stand it that Phenom is better lol ;)

[dinos22]

these guys are right George

Apples for apples ;)

btw that PCMARK SCORE IS EEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEK

[hipro5]

Hipro, you are not comparing fair :)
i'm afraid you can't stand it that Phenom is better lol ;)

:confused: What do you mean?......macci with AMD "gave what he got" and I with Intel "gave what I got"....Where is the problem with that?.... :confused:
I MUST DOWNCLOCK my rams or use DDR2?.....What for?.....THIS is the new technology now......Do you want to go back and get a Pentium4 to bench instead?.... :D

[Praz]

:confused: What do you mean?......macci with AMD "gave what he got" and I with Intel "gave what I got"....Where is the problem with that?.... :confused:
I MUST DOWNCLOCK my rams or use DDR2?.....What for?.....THIS is the new technology now......Do you want to go back and get a Pentium4 to bench instead?.... :D
I agree. This thread is showcasing AMD's latest and greatest. So to keep it fair it should be compared to last year's technology?

[Syz`]

I've emptied this as it is innapropriate and not needed, carry on benching. ;)

[GazC]

^ Oh dear.

[macci]

George,

don't downclock anything or switch to DDR2 - thanks for the ref. score! now I have a new target again (which is not too far either) :D

I think my GT3 is too low for the clocks (I remember it stutters in the beginning of GT3). Let me try WinXP. I think I can get it to over 20.5k too.

in any case its still a VERY close match.

I'll work on that 06 number during the weekend.

speaking of latest technology: I think you should try with Vista ;) (for both 05 / 06)

[Simps]

Owned! Lol

[dpa]

this is a very close match.. but Macci runs very high NB clocks, nobody here on the forum has managed so high clock on either cpu og NB?

anyways, George CAN run the setup at 4.7Ghz on air, while macci is stuck at 3Ghz on air.. enough said at xs.org? :D

[Praz]

speaking of latest technology: I think you should try with Vista ;) (for both 05 / 06)
Definitely. I've seen as much as 600 points difference in 06 between Vista and XP at 3600MHz.

[tictac]

Owned! ;)

[KTE]

very nice... I'm reading. :D

Quit trying to distract us with that PCMark score... George. :p: :eek:


this is a very close match.. but Macci runs very high NB clocks, nobody here on the forum has managed so high clock on either cpu og NB?
People run NB at 2.25GHz for sub 2.9GHz. I run NB 2.16GHz at 2.64GHz CPU. Less than 300MHz difference and he's at higher clocks. But look at the FSB George ran for 3GHz. Is that stock QX9650? Stock is 333MHz and he is running 428.7MHz. 379MHz difference. But Sami is 22MHz above stock on HT. ref (200MHz) while George is about 95MHz above.

And RAM is not fair at all. DDR3 is better than DDR2 for bench -> all benchers incl. Geroge himself say this and that's why they use it. We've all seen your posts. :D

Again that 2750Mhz George compared to Badnizze had 458MHz FSB. That's insane FSB for quads. Try ~333MHz and you have a proper comparison.

[BeardyMan]

:confused: What do you mean?......macci with AMD "gave what he got" and I with Intel "gave what I got"....Where is the problem with that?.... :confused:
I MUST DOWNCLOCK my rams or use DDR2?.....What for?.....THIS is the new technology now......Do you want to go back and get a Pentium4 to bench instead?.... :D

The point is that you have the possibility to compare fair yet you prefer to do it otherwise. You wanne know who is faster clock per clock in this bench yet you use a factor that influence GT1 and GT3 for sure.

[massman]

@ Massman please DON'T compare systems each other by saying DDR2/DDR3 or so.....IF AMD rig whould have DDR3, the guys whould bench with DDR3.....Does it have?.... :D
Everyone benchies with what he has.... :) ;)

You're right if one would compare what cpu is better for extreme benching purpose ... but then, you would choose Intel just because you can clock it at 5,8G using ln² ;).

To find out what cpu is better clock per clock, you need to use the same hardware/software :) .

[Sampsa]

massman, CPU can't do much alone since it needs a chipset and memory to work so the performance is a sum of these three components. Right now the best these two companies can offer:

AMD = Phenom (K10 architecture) + AMD 790 FX + DDR2
Intel = Yorkfield (Penryn) + Intel X38 + DDR3

[massman]

You may be right, Sampsa, though I'd rather see results with both systems running on DDR2 to compare both chips just as I'd rather see XP vs XP instead of Vista vs XP comparisons :)

[BeardyMan]

You may be right, Sampsa, though I'd rather see results with both systems running on DDR2 to compare both chips just as I'd rather see XP vs XP instead of Vista vs XP comparisons :)

Indeed, is it me or does HIPRO, Sampsa know what the result is when using DDR2 at the same speeds...:D ;)

[SF3D]

I can bench 05 tonight with following setup:
Yorkfield 333x9 + X38 + DDR2 + 2900Pro 860/980
Operating system is Vista 32-bit.

Let's see what the score will be...

[kwalker]

I would be interested in seeing yorkfield vs phenom folding SMP for a half hour each both at stock settings but then again this would be the wrong section.
cant blame a guy for asking :D

[misteroadster]

Really interresting thread, i'm really happy for the good results from the phenom, 2005 is so cpu dependant. 3DM2005 is a lot more cpu dependant than new games as Crysis with quality setting settings just under 60Fps.

It would be interresting to see some numbers under XP too, particulary 3DM2001 :)

Just any question : Phenom should have 6Mo L3 and 4X1Mo L2 in 2008 ?


Thanks for the numbers macci, i hope a lot since you reappeared :clap:


Edit : Ok Vista Roxx :D

[hipro5]

George,

don't downclock anything or switch to DDR2 - thanks for the ref. score! now I have a new target again (which is not too far either) :D

I think my GT3 is too low for the clocks (I remember it stutters in the beginning of GT3). Let me try WinXP. I think I can get it to over 20.5k too.

in any case its still a VERY close match.

I'll work on that 06 number during the weekend.

speaking of latest technology: I think you should try with Vista ;) (for both 05 / 06)

You filthy pron macci...... :D
You didn't tell me that 2k5 on Windows VISTA shines!.....AHA!.....THAT'S why you have benched it on VISTA right?...... :D :p:

I ALSO have stutters in the beginning of GT3 with XP OR VISTA....You should fix the drivers a bit I think..... :D

With my previus setting on rams, I got 20694 and with more tight settings, I got the below...... ;)
ALSO my VGA rams are LOWER than yours.... :p:

Take a look at the Performance Level you INTEL guys..... ;)

http://www.hipro-tech.com/images/hipro5/ASUS_Blitz_Extreme/York_Retail/2k5_20802_3000MHz_2900XT_860_990_1.png

http://www.hipro-tech.com/images/hipro5/ASUS_Blitz_Extreme/York_Retail/2k5_20802_3000MHz_2900XT_860_990_2.png

http://www.hipro-tech.com/images/hipro5/ASUS_Blitz_Extreme/York_Retail/2k5_20802_3000MHz_2900XT_860_990_3.png


The point is that you have the possibility to compare fair yet you prefer to do it otherwise. You wanne know who is faster clock per clock in this bench yet you use a factor that influence GT1 and GT3 for sure.

Memory Bandwidth influence a LOT GT1 and a bit GT2.....
Now tell me this..... Isn't Phenom supposed to have INTERNAL MEMORY CONTROLLER that wipes out EVERY Intel rig?......Yes or No?..... ;)
This is GT1 of the 2k5....
Now do you want to compare ALL 3Ds to see how Phenon is NOT doing that well at all?..... ;)


You're right if one would compare what cpu is better for extreme benching purpose ... but then, you would choose Intel just because you can clock it at 5,8G using ln² ;).

To find out what cpu is better clock per clock, you need to use the same hardware/software :) .

SOFTWARE used the SAME now......Windows Vista and it seems that it performs BETTER as for 2k5..... ;)

SO we will NOW "avoid" the "SOFTWARE influence" since BOTH benchies where done on VISTA and we keep on, right?.... :)

Hardware: Look my above answer to BeardyMan.... :)

We are testing at the EXACT SAME FREQUENCY/bench/software and EACH system is MAXED out with whatever methods the manufacturer provide us.... :)


anyways, George CAN run the setup at 4.7Ghz on air, while macci is stuck at 3Ghz on air.. enough said at xs.org? :D

:D


Again that 2750Mhz George compared to Badnizze had 458MHz FSB. That's insane FSB for quads. Try ~333MHz and you have a proper comparison.

"Insane fsb for QUADS" is the 500MHz+ M8....This 428 is peanuts.... ;)

.

[Sampsa]

hipro you cheater.. no wonder your scores are higher because your GPU clock is 863 MHz instead of 860 MHz ;)

[hipro5]

hipro you cheater.. no wonder your scores are higher because your GPU clock is 863 MHz instead of 860 MHz ;)

....and?..... what frequency is that?... :D
Plus Riva Tuner can't take it at 860 sharp....it puts it on 863.... :)

[dinos22]

good job hipro

what's the vMCH on the board for that PL?

[BeardyMan]

Ok i will do it myself next week on my QX9650, and teamgroup DDR2.
Omg a stupid run on DDR2 seems to be mission impossible :rolleyes:

[hipro5]

good job hipro

what's the vMCH on the board for that PL?

I used 1.95VMCH dinos.... :)
Also - a secret - Strap 200:D


Ok i will do it myself next week on my QX9650, and teamgroup DDR2.
Omg a stupid run on DDR2 seems to be mission impossible :rolleyes:

I don't own a DDR2 mobo any more M8......What can I do?.....Bench with ABIT AW9D MAX and X6800?..... :D
I'll borrow an ASUS Blitz Formula for you.... ;)

[zakelwe]

I don't care who wins this is fun to watch. I think Macci can tweak a fair bit more yet but i get the feeling Hipro has more in hand hardware wise.

It's nice to see Macci back benching again :toast: Looking forward to results with LN2.

Regards

Andy

[dinos22]

I used 1.95VMCH dinos.... :)
Also - a secret - Strap 200:D




that explains low PL value besides monsterous voltages on MCH

what cooling do you have on MCH
looks like i'll have to crack open my nautilus 500 with chipset cooler block from corsair to really push the damn RAM and straps

buying one of OPBs boards :p for super fast CDT ownage hahahah j/k
really need a decent board for RAM clocking and ddr3 GB isn't cutting it atm
bios needs a lot of work on them :(

[Timbosan]

[B]"Insane fsb for QUADS" is the 500MHz+ M8....This 428 is peanuts.... ;)

.

yep - P5k does 480 FSB with quad at 1.4v vMCH & 1.7v PLL ;)

[hipro5]

what cooling do you have on MCH

I have a cooper Enzotech heatsink with a DELTA 4x4cm 10000rpms on it..... :)

[massman]

Memory Bandwidth influence a LOT GT1 and a bit GT2.....
Now tell me this..... Isn't Phenom supposed to have INTERNAL MEMORY CONTROLLER that wipes out EVERY Intel rig?......Yes or No?..... ;)
This is GT1 of the 2k5....
Now do you want to compare ALL 3Ds to see how Phenon is NOT doing that well at all?..... ;)

SOFTWARE used the SAME now......Windows Vista and it seems that it performs BETTER as for 2k5..... ;)

SO we will NOW "avoid" the "SOFTWARE influence" since BOTH benchies where done on VISTA and we keep on, right?.... :)

Hardware: Look my above answer to BeardyMan.... :)

We are testing at the EXACT SAME FREQUENCY/bench/software and EACH system is MAXED out with whatever methods the manufacturer provide us.... :)

:D

You sick bastard :D. PL=4 is just TOO fast :D

Anyway, I take your point, going as fast as you can with the hardware that the manuf supports does seem to make more sense :)

Now, I'm hoping that Macci is going to pass this again, just to keep this battle going. You know ... I have a flashback to 2003 :rofl:

[hipro5]

You know ... I have a flashback to 2003 :rofl:

:D :clap:

[BadNizze]

Hipro: 4,5k @ 2750 is FAST, Don´t think i can match that...ever, But I will try ;) My DFi board, does not suffer from the Low NB clock bug. As my older Gigabyte did\does.

Regarding the FSB thing, K8 does not scale with HTT\FSB I have not done testing with K10, but i think it´s safe to say that running high HTT does not give much in terms of preformance on K10 either. But it would be nice to see what a york does @ 3Ghz with 333FSb for reference. Could someone run 2,4ghz on a York to? Would be nice to see how it scales with Mhz aswell! No one have tested york @ low mhz, right?

[massman]

No one have tested york @ low mhz, right?

I'm keeping track of as many 45nm 1M results as possible. This is 1M vs MHz, not useful here, but I have one result with very low clock freq, which shows that it scales pretty good.

[Simps]

Macci will beat it :)

[SF3D]

Here is totally stock result with yorkfield. Over 20K, but not much. (This is my 24/7 machine, so there is a lot of crap in background)

QX9650 9x333
Maximus Formula
2900Pro 863/999
2GB ram/ DDR2 1005 / 4-4-4-6
"strap" 266 (-->200 wasn't stable because NB was too hot.)
Performance level= 4
Vista 32-bit.
Catalyst 7.10 /7.11 doesn't work for some reason.

68721

[Simps]

@ SF3D

Nice!
Can you show us some 3DMARK06 with those stock settings?

[]'s
Simps

[mongoled]

nice one guys, come on Hipro play nice :D

good to see Macci benching ti is fun to watch

good to see ipc is close

[BeardyMan]

Here is totally stock result with yorkfield. Over 20K, but not much. (This is my 24/7 machine, so there is a lot of crap in background)

QX9650 9x333
Maximus Formula
2900Pro 863/999
2GB ram/ DDR2 1005 / 4-4-4-6
"strap" 266 (-->200 wasn't stable because NB was too hot.)
Performance level= 4
Vista 32-bit.
Catalyst 7.10 /7.11 doesn't work for some reason.

68721

as i thought so, the reason why hipro doesnt show DDR2 cause he knows that phenom with DDR beats teh crap out of yorkfield with DDR2 and for sure when macci is roling the dices , thx for the info man ;)

[dpa]

as i thought so, the reason why hipro doesnt show DDR2 cause he knows that phenom with DDR beats teh crap out of yorkfield with DDR2 and for sure when macci is roling the dices , thx for the info man ;)

why do you really need ddr2 for...?? ddr3 is the new thing now, just let DDR2 go, okey?

[Oliver]

Really impressive there George.

Keep pushing it.

[LowRun]

why do you really need ddr2 for...?? ddr3 is the new thing now, just let DDR2 go, okey?

DDR2 is still what 99% will buy today.

[IvanAndreevich]

DDR2 is still what 99% will buy today.

Ok so to satisfy the majority, then they should also be benching on integrated graphics and with low-end dual core chips :ROTF:

[NoX]

as i thought so, the reason why hipro doesnt show DDR2 cause he knows that phenom with DDR beats teh crap out of yorkfield with DDR2 and for sure when macci is roling the dices , thx for the info man ;)

Cut it down man. Your remarks are pathetic.:shrug:

[knightwolf654]

id like to see it run with ddr2 since cheapest ddr3 is $200 :o
then again if u have the $1149 to dish out for a 9650 $200 must be chunk change.

[hipro5]

as i thought so, the reason why hipro doesnt show DDR2 cause he knows that phenom with DDR beats teh crap out of yorkfield with DDR2 and for sure when macci is roling the dices , thx for the info man ;)

You MUST be blind or something..... :D
I wroted to you above that I DON'T HAVE a DDR2 mobo right now to test..... :D
I'll borrow one maybe today or tomorrow...... ;)

[Harshal]

@ Hipro5 & Macci, Nice showdown for us poor guys!!!
Honestly, its good to see AMD performing in the same range as Intel in 3DM05 atleast :P

[Astri]

Good, now go extreme to see the results ;)

[Harshal]

^^ I think AMD chip is @ extreme [wid air] only... :rofl:

[massman]

why do you really need ddr2 for...?? ddr3 is the new thing now, just let DDR2 go, okey?

With good DDR2 sticks, you're close to the performance of average DDR3 AND DDR2 is supercheap ;)

[KTE]

Regarding the FSB thing, K8 does not scale with HTT\FSB I have not done testing with K10, but i think it´s safe to say that running high HTT does not give much in terms of preformance on K10 either.If FSB gave him no perf increase, he wouldn't have increased it by that much but ran it stock, no? Does it hike power consumption and need higher volts for no reason? ;)
As for K10h HT means nothing -> it makes a good difference if you're below 1800MHz anyway. All things constant (stock) Phenom 9500 2600XT 2k6:
HT 200MHz: 4438 (http://img99.imageshack.us/img99/4025/2k6qy8.png)
HT 1000MHz: 4742 (http://img208.imageshack.us/img208/6976/2k6kc9.png)
HT 1800MHz: 4783 (http://img205.imageshack.us/img205/8819/2k6ey1.png)

Forgive the card. :lol:

[massman]

KTE, why doesn't the cpu score go up?

[Periander6]

Honestly, its good to see AMD performing in the same range as Intel in 3DM05 atleast :P

It's good to see a max overclock Phenom going up against a stock clockspeed Yorkie in a bench cherrypicked to favor Phenom and still lose? :confused: Phenom will be "in the same range as Intel" when it has stock clocks over 3GHz and over, overclocks to 4.5GHz on air, performs across a range of benchmarks not a single cherrypicked one, and doesn't come with a performance destroying bugfix.

Until then the only comparisons between Phenom and Yorkfield will be like this thread, a kind of carnival show where a pro fighter is taking on drunken challengers from the crowd with one hand tied behind his back.

[Harshal]

^^ No my my point is clock to clock... and yes all of us know AMD should have aimed a lil higher then they did.
Also, I completly agree that with same percentage of overclock on Yorky will leave Phenom miles behind. :)

[Reckless187]

Keep up the good work guys this is awesome! :clap: :-|

[BadNizze]

If FSB gave him no perf increase, he wouldn't have increased it by that much but ran it stock, no? Does it hike power consumption and need higher volts for no reason? ;)
As for K10h HT means nothing -> it makes a good difference if you're below 1800MHz anyway. All things constant (stock) Phenom 9500 2600XT 2k6:
HT 200MHz: 4438 (http://img99.imageshack.us/img99/4025/2k6qy8.png)
HT 1000MHz: 4742 (http://img208.imageshack.us/img208/6976/2k6kc9.png)
HT 1800MHz: 4783 (http://img205.imageshack.us/img205/8819/2k6ey1.png)

Forgive the card. :lol:


I think I´m missing something in youre post, because of the language barrier. Anyway, my point is that running HTT over stock on an AMD rigg does close to nothing for preformace, Right? I was actualy defending hipro by saying that high HTT does close to nothing for preformace with AMD cpu´s. No the other way around. I think that hipro´s score is 100% fair BTW. Towards me or anyone else that for that matter. I´m doing the best i can @ 2750mhz and so is hipro, if hi runs 200 or 333 or 500 fsb is fine by me! But i would be nice to see what a york pulls in 06 CPU tests... @ 2420mhz, 2750mhz and 3000mhz

[ElMoIsEviL]

nice one guys, come on Hipro play nice :D

good to see Macci benching ti is fun to watch

good to see ipc is close

ipc isn't close.

It's what 3DMark 05 does. It's the way it's coded. It really doesn't showcase the IPC of these processors. If you want that you'll have to run 3D Games or compare 3DMark 06 CPU scores.

Heck, using Cinebench or SuperPi would be better for IPC comparing. Mike over at Xtreme CPU has been doing so and the results are less then favorable for the Phenom CPU. It's a horrible processor right now. In time it might mature into something competitive, but as it stands it's junk.

http://www.erenumerique.fr/images/21/20071123/3dmark2005_cpu.gif
http://img.photobucket.com/albums/v51/ElMoIsEviL/Benchmarks/3dmark05.jpg

[massman]

The 3DM05 cpu test is :banana::banana::banana::banana:. Rerun the test and your score will continuesly go up without changing the clocks.

[the-strokes]

higher ht is good for high res 3d benchs, no?

[KTE]

KTE, why doesn't the cpu score go up?
No idea but HT increases bandwidth between GPU>CPU so GPU gains if bottlenecked.

I think I´m missing something in youre post, because of the language barrier. Anyway, my point is that running HTT over stock on an AMD rigg does close to nothing for preformace, Right? I was actualy defending hipro by saying that high HTT does close to nothing for preformace with AMD cpu´s.I understood you fine the first time. I don't think it's a true generalization TBH. I know 1800 from 1000 gets you good gains and people said the same thing when HT was at 1000. Have you tested 2600MHz HT vs 1800MHz HT across all apps? Try it. I would if I could but mine is hevaing like a dud. I don't see why AMD would increase HT from X2 to K10h to Shanghai if it did nothing on perf in any app. ;)

[Stefan Payne]

KTE, why doesn't the cpu score go up?
because the HT Link has nothing to do with the CPU performance 'cause the Memoryinterface is in the CPU.

The HT Link ist just tu communicate with the rest (ie graphics card and other I/O things), therefore the HT Link can have an impact on graphics performance especially if you use a HT3.0 chipset and a graphics card with small amount of memory (like the 8800GT/256MiB or the HD3850 with 256MiB).

[saaya]

Heck, using Cinebench or SuperPi would be better for IPC comparing.
superpi to compare cpus? imo its mostly a cache benchmark now, not really a cpu bench.

[Brother Esau]

ipc isn't close.

It's what 3DMark 05 does. It's the way it's coded. It really doesn't showcase the IPC of these processors. If you want that you'll have to run 3D Games or compare 3DMark 06 CPU scores.

Heck, using Cinebench or SuperPi would be better for IPC comparing. Mike over at Xtreme CPU has been doing so and the results are less then favorable for the Phenom CPU. It's a horrible processor right now. In time it might mature into something competitive, but as it stands it's junk.

http://www.erenumerique.fr/images/21/20071123/3dmark2005_cpu.gif
http://img.photobucket.com/albums/v51/ElMoIsEviL/Benchmarks/3dmark05.jpg



Are you sure that aint the rebadged Intel Score Charts from Toms Hardware 1 1/2 ago:D Besides man what is so god awful about those scores? It just needs some time to mature just Like Intel did with the Prescots.