AMD delays Phenom 2.4 GHz due to TLB errata

[red]

http://www.theinquirer.net/gb/inquirer/news/2007/11/18/amd-delays-phenom-ghz-due-tlb

[informal]

This is bad news indeed.Weird the 2.2 and 2.3Ghz version are not affected :shrug:
Could it be because the 2.4Ghz version's Northbridge(L3 included) runs at higher frequency?

[terrace215]

Typical Theo-quality article, but he is one of AMD's biggest fans, so I'll assume the recall part is correct.

The AMD statement he quotes says the TLB problem is *BIOS* correctable, which implies that is NOT the reason for the recall.

But if they really need to wait for B3 to get anything over 2.3GHz... ouch!

Seriously starting to wonder if the "3GHz Phenom" demoed in July was actually running at that speed.

Another thing that is clear: the K10 was no where CLOSE to being production ready when they launched it in September. No wonder tier 1 OEMs aren't touching it until some time next year.

[savantu]

Is there anything else that can go wrong at AMD ? I mean , are they really at the bottom and digging on top of that ?

[Kasparz]

Prove it...
Come on, he needs to increase his post count somehow.
Well, Phenom 9700 is still there, available for preorder.
http://www.pixmania.co.uk/uk/uk/652644/art/amd/phenom-x4-9700-2-4-ghz-l2.html

[R101]

Is there anything else that can go wrong at AMD ? I mean , are they really at the bottom and digging on top of that ?

I just hope Radeon 3xxx series don't fail, that's the only thing that could get them some cash right now.

[Thesavage]

Yes a strange article bios correctable but however a recall? May some of the resident gurus can give us some insight.

[nemrod]

While benchmark software vendors had some problems detecting how new memory controller work - both SiSoft Sandra XII and Everest 4.20 were returning single-channel results, regardless of memory controller being used in SC/DC, dual-72-bit or 144-bit-...
we have seen that on previews on XS? Is it related to tlb error?

[Periander6]

So now it's B3 that's the dancing in the aisles stepping? :rolleyes:

[freeloader]

Probably why a few users have been getting DC/SC memory problems.

You can see a 3ghz Phenom on Youtube for anyone who's interested.

http://www.youtube.com/watch?v=i_q_9a4zDps

[duploxxx]

not good, but if its only bios update then it won't take long.

[xenolith]

So now it's B3 that's the dancing in the aisles stepping? :rolleyes:

Thanks for the good laugh...

Just when DAMMIT was starting to pick themselves up...

[terrace215]

not good, but if its only bios update then it won't take long.

Yet in the same article, he tells us that B3 is needed for 2.4GHz parts.

I guess we'll know more at launch... tomorrow?

Of course he also says: "B3 is considered a speed *bin* of B1 and B2" which makes no sense at all. Did he mean "speed-focused stepping based on..."? Something else?

[gallag]

what will happed to the stocks of 2.4 that have already been sent out? If they have to be destroyed that will seriously hurt there next financial report, No?

[JumpingJack]

This is bad news indeed.Weird the 2.2 and 2.3Ghz version are not affected :shrug:
Could it be because the 2.4Ghz version's Northbridge(L3 included) runs at higher frequency?

It's weird, because Theo is probably blowin' smoke. Not that the errata is not real, probably is... but the core architecture is the same, unless AMD was working a new stepping for the 2.4 GHz specifically... not improbable.

He is wrong though that K8's could not get microcode updated. Here is a procedure for updating K8 microcode.
http://packetstormsecurity.nl/0407-exploits/OpteronMicrocode.txt

What is more odd though is Theo goes on to claim overclocking headroom, but if 2.4 GHz is not stable underload, what will the behavior of an overclocked 2.3 GHz processor be?

[Shintai]

what will happed to the stocks of 2.4 that have already been sent out? If they have to be destroyed that will seriously hurt there next financial report, No?

No im sure there is a usual BIOS workaround. However those 2.4Ghz might have to sit in the warehouse until that is ready.

[JumpingJack]

No im sure there is a usual BIOS workaround. However those 2.4Ghz might have to sit in the warehouse until that is ready.

Well according to Theo, it is not correctable -- as, according to him, AMD does not use microcode updation as a means to correct errors. This is bunk of course, but if they are really flawed with an incorrectable flaw, then AMD would need to take the write-off I suspect.

[terrace215]

It's weird, because Theo is probably blowin' smoke.

Theo is often very very wrong, but his errors are always to AMD's benefit.

BTW, here's a link from Nov 5 stating that the 9700 (2.4GHz) part will not be available until Dec 14:

http://www.nordichardware.com/news,7016.html

Fudzilla (another pro-AMD, Phenom hype guy) just recently said the Phenom FX will launch in Q1 08 at 2.6GHz, and that 3GHz parts have now vanished from AMD's roadmap:

http://www.fudzilla.com/index.php?option=com_content&task=view&id=4256&Itemid=1

Further, Tri-core is backed off to Q208, from AMD's PR of Q108 in September:

http://www.overclockers.com/tips01248/

(Ed is definitely NOT a pro-AMD guy)

All in all, terrible execution that has probably cost AMD their future.

[terrace215]

No im sure there is a usual BIOS workaround. However those 2.4Ghz might have to sit in the warehouse until that is ready.

The other downside is that BIOS workarounds for things like this can often result in a performance hit.

One also wonders about the overclocking potential given these problems... will O/C-ing to 2.4GHz and above cause crashing under load?

[cky2k6]

yep, i have a hunch that there's no actual recall, just a delay to the launch of higher binned parts. as for overclocking, as theo said, some chipse will crash and some will not, so technically you will be able to overclock, just with even more risk of instability than usual... this situation with k10 is just murphy's law at its best.

[GoThr3k]

only happens when you have all cores under 100% load

[JumpingJack]

The other downside is that BIOS workarounds for things like this can often result in a performance hit.

One also wonders about the overclocking potential given these problems... will O/C-ing to 2.4GHz and above cause crashing under load?

Someone suggested in another forum that it may be caused by a hot spot, pure speculation but it would make sense (similar to the 'thermal hot spot, 3K Opty recall a year or two ago).

So say AMD has a hot spot at 100% load that is causing a few functional blocks to latch up.... this would make sense based on what Theo is saying. Higher clocked CPUs would suffer from it and there would be some variation CPU to CPU if AMD is on a cliff process wise with heat/voltage/clock:


Some 9500/9600 parts may even be overclocked to 2.6, 2.8, 2.9, 3.0 GHz and they will have no problems whatsoever, while some will have this error.

Pure speculation of course, however if we assume this is the explanation then no BIOS update will fix this.

[Zytek_Fan]

Unfortunate for AMD...I've already decided to get Yorkfield, but I do hope AMD can do well with Phenom...

[JumpingJack]

Unfortunate for AMD...I've already decided to get Yorkfield, but I do hope AMD can do well with Phenom...

It will do well overall, it is priced well and it is well above AMD's current ASPs so it will help a little on the bottom line. AMD has enough dedicated customers in it's base that it will sell well.

On top of that, despite the somewhat disappointing performance (i.e. not the core killer AMD had originally touted), the IPC is closer overall which for a 3-issue design is quite an achievement. All things being equal, in the absence of a strong Intel performer, a 15% improvement in IPC over K8 is quite healthy and certainly an accomplishment.

If there is a focus for AMD it is to work deligently to debug the architecture and work on the process side to get clock speeds up, for now it will be a good competitive mid-range solution. Especially for those with AM2 boards capable of accepting the chip.

Jack

[largon]

AMD delays (...)

67448

[justapost]

According to sources close to the company, 2.4, 2.6, 2.8, 3.0 and faster parts will appear on the market when the B3 revision comes to life. B3 revision is key CPU stepping for AMD's future, since it is considered a speed-bin of B1 and B2. With this errata fixed, Phenoms will have no problems competing against Intel in clock-per-clock action, but it will bring serious disappointment to AMD fanboys and market itself. They are probably blubbing already.
Blub
:bounce:
I assume B3's will be around when Intel starts shipping 45nm parts. Sounds like B2's are not the overclockers first choice.

Hope in two hours the first reviews will shed a little light on the whole platforms powerconsumtion.

:yawn2:

[Zytek_Fan]

If AMD can get problems sorted out, then MAYBE they can get the speed ramped up fast, and get the OCing they need to be able to achieve.

/optimism.

[JumpingJack]

If AMD can get problems sorted out, then MAYBE they can get the speed ramped up fast, and get the OCing they need to be able to achieve.

/optimism.

I, myself, plan on getting a Phenom -- not right out of the gate, it will take another stepping or two before I will bite, but I do plan on having a Phenom rig up and running.

[terrace215]

It will do well overall, it is priced well and it is well above AMD's current ASPs so it will help a little on the bottom line.

Not clear.

Even for the top line, you have to factor in the much lower yield of parts per wafer. In other words, at a given capacity, they have to give up 2-3x as many Brisbanes for each Phenom they want to make. So, higher ASPs, but lower units.

And for the bottom line, the increased cost per part has an impact.

[JumpingJack]

Not clear.

Even for the top line, you have to factor in the much lower yield of parts per wafer. In other words, at a given capacity, they have to give up 2-3x as many Brisbanes for each Phenom they want to make. So, higher ASPs, but lower units.

And for the bottom line, the increased cost per part has an impact.

That is why I said 'a little' -- large die are always more costly, so in terms of margin ... meh .. not a huge jolt I suspect. Clock for clock though, it is clear, the core uArch is not quite as dominating as it was over K8 (even the Penryn rev).

On top of that, in server, Intel has BW issues in certain workloads, so there are strong drivers/applications for Barcelona in that market.

While not as spectacular as we were led to believe, it is by no means 'another Prescott'.

Jack

[terrace215]

Blub
:bounce:
I assume B3's will be around when Intel starts shipping 45nm parts.

Why?

Intel started shipping 45nm parts already, and the rest of the desktop and mobile lineup launches in early Jan.

B3 looks like some time in Q1 (based on claims of a 2.6 FX part in Q1), assuming it turns out okay, but do you have any information about exactly when in Q1 it might be ready?

[terrace215]

That is why I said 'a little' -- l

Jack

I was suggesting it might not even be 'a little'. It might be a negative to the bottom line, due to increased costs per unit together with fewer units, relative to an all-Brisbane model.

By way of example, suppose AMD can produce either 200 Brisbanes or 60 Phenoms from a wafer. Switching wafers to Phenom might make their finances worse, rather than better. (Note: this assumes they are switching, not adding additional wafers, either due to capacity constraints or market share constraints, i.e. they can sell X units, and can choose to have Y of those be Phenom instead of Brisbane.)

Of course, long term they need Phenom, because it is more competitive. A tough spot to be in.

[Zytek_Fan]

Truly I do hope Phenom will perform well, because I would like to get a Phenom + 790FX + 3870 CF rig, but it's looking like that's not the case... :(

As I said before, I'm not a fanboy, I just like the underdog, but I desire, like any enthusiast, the best product there is available, regardless of company.

[JumpingJack]

I was suggesting it might not even be 'a little'. It might be a negative to the bottom line, due to increased costs per unit together with fewer units, relative to an all-Brisbane model.

By way of example, suppose AMD can produce either 200 Brisbanes or 60 Phenoms from a wafer. Switching wafers to Phenom might make their finances worse, rather than better. (Note: this assumes they are switching, not adding additional wafers, either due to capacity constraints or market share constraints, i.e. they can sell X units, and can choose to have Y of those be Phenom instead of Brisbane.)

Of course, long term they need Phenom, because it is more competitive. A tough spot to be in.

I can't argue with that... I, nor anyone, knows what the yields are like there is no doubt they get much less per wafer pass through the fab... regardless of what AMD tells the investors, the emperical evidence is that they are having a tough time yielding... Barcey launched over 2 months ago, and tier 1 vendors still won't commit to a ship date, the handful of 2.0 GHz and even the 1.8 Ghz have dried up in the channel. We have all seen the reports.. etc. etc.

I agree it is a tough spot to be in... but in this case, AMD made their bed now they need to lay in it... choosing a monolithic design may be technologically 'sexy', but in practicality probably a poor choice... this is the trade off you take with their architectural approach, the MCM is not easy to do.

Jack

[Dimitriman]

wish you well amd. hopefully ill pick up a fusion in 3 years time. this time im going penryn.
good luck.

[justapost]

Why?
Intel started shipping 45nm parts already, and the rest of the desktop and mobile lineup launches in early Jan.
I meant affordable 45nm parts.


B3 looks like some time in Q1 (based on claims of a 2.6 FX part in Q1), assuming it turns out okay, but do you have any information about exactly when in Q1 it might be ready?
I have no information about the B3 release, but your statement sounds reasonable.
If those B3's are on par with 45nm Intel chips, a two month later release is not bad taking all the B1 troubles they had this year into account.

[Movieman]

not good, but if its only bios update then it won't take long.

Agreed. I see this as AMD has run into a number of issues and like any manufacturing firm they will identify,work out a fix, and release that fix.
It's a shame that they didn't have all this done before release but it just proves that even big companies run into issues.
Add in that they were really under the gun to release and it makes sense seeing these issues now.
I have faith that they will get it all done and have an excellent product long term.:up:

[Lightman]

Truly I do hope Phenom will perform well, because I would like to get a Phenom + 790FX + 3870 CF rig, but it's looking like that's not the case... :(

As I said before, I'm not a fanboy, I just like the underdog, but I desire, like any enthusiast, the best product there is available, regardless of company.

I will get that AMD rig (already got mobo and CF is on it's way) and report results here, so you will now how much slower it is compared to similar Intel offer. I have no doubt it will be slower in pure benchmarking terms, but I'm looking for fun and also like to play with something most of people choose not to...:up:

I only hope that AMD will come with more competitive or better solution than Intel at some point in near future (like before 2010) for good of all enthusiasts here on XS.:up:

[informal]

Theo is smoking crack again since (like JumpingJack posted before),there is a microcode update possibility.The problem could be when and who will be doing the update:end users,mobo makers or the sellers in the channel?

[freeloader]

only happens when you have all cores under 100% load

Unfortunately for me, that's all the time. :D

[metro.cl]

Hmm i smell something fishy about this, why only top speed CPU has this issue? why pull it from market if a bios workaround is posible?.

Then why they talk that B3 stepping should fix it all?

[terrace215]

If those B3's are on par with 45nm Intel chips

They won't even be close.

The K10 has an IPC like Intel's 65nm parts, and it will come nowhere near Intel's 45nm clockspeeds.

[justapost]

They won't even be close.

The K10 has an IPC like Intel's 65nm parts, and it will come nowhere near Intel's 45nm clockspeeds.
Unfortunately no phenom review leaked till now. But from what we have seen k10 is very close with core 2 architecture clock for clock.
But for intel, the expected 5-15% speed improvement with the transition vom 65 to 45nm seems to be more like 1-5%. Don't ask for a proof because i do not follow all the intel news atm. It does not sound unrealistic to me that amd can gain a few more percentages with new steppings.
I'm sure intel's 45nm chips will be the first choice for overclockers, but for the mainstream pricing and energy efficieny is what counts.
I just started to play with overclocking and hardware tweaking , but i will not go for higest clock speed but for best energy efficiency settings.
I wait for a cpu with a really low powerusage for daily work. But if required it should have enough power to handle computationally intensive tasks in a short timeframe. The later do occure more seldom here, so til now amd x2's are the better choice for me. So atm i'm really curious about phenoms power usage.

[Omastar]

only happens when you have all cores under 100% load

That might present a tiny problem for WCG people...

[Quintero]

why only top speed CPU has this issue?I wonder about that too. Assuming all models are physically identical, only binned differently, their claim doesn't make sense to me. And I'm further troubled by the following citation:


9500 (2.2 GHz) and 9600 (2.3 GHz) parts are unaffected by the errata. Some 9500/9600 parts may even be overclocked to 2.6, 2.8, 2.9, 3.0 GHz and they will have no problems whatsoever, while some will have this error.I hope and believe that this is little but INQ spin...

[Zytek_Fan]

I wish INQ Charlie could report on AMD stuff more. He makes more sense...

[terrace215]

http://www.dailytech.com/AMD+Spider+Crawls+Up+the+Water+Spout/article9691.htm


AMD guidance originally stated that a 2.4GHz Phenom, dubbed the Phenom 9700, would also launch on November 19, however last minute roadmap updates indicate this chip will come in December instead.

Last minute change suggests a late-breaking bug, as Theo claimed, although his details appear to be jumbled at best, nonsense at worst...

[cadaveca]

B3 being a speed-bin makes sense...not for raw cpu speed, but for memorycontroller speed. Actually makes alot of sense, given current guidance from AMD, and lots of benches we see with ram @ 667mhz, 5-5-5-15...

[terrace215]

B3 being a speed-bin makes sense...

B3 is a *stepping* not a speed-bin. A speed-bin is a frequency, like 2.4GHz.

[cadaveca]

really? never knew...:stick: But do they sell cpu's based on memorycontroller functionality? NOPE.:lol2:

Better inform INtel too...X48, anyone?

[alucasa]

Well, duh, I don't see this as a big deal. It will be fixed eventually, and you guys will have Phenom FX eventually.

I won't get Phenom however. 2P system is my taste.

[Zytek_Fan]

Well, duh, I don't see this as a big deal. It will be fixed eventually, and you guys will have Phenom FX eventually.

I won't get Phenom however. 2P system is my taste.

How about 4P? :yepp:

[informal]

http://www.dailytech.com/AMD+Spider+Crawls+Up+the+Water+Spout/article9691.htm


AMD guidance originally stated that a 2.4GHz Phenom, dubbed the Phenom 9700, would also launch on November 19, however last minute roadmap updates indicate this chip will come in December instead.

Last minute change suggests a late-breaking bug, as Theo claimed, although his details appear to be jumbled at best, nonsense at worst...

Theo overblown the story.DT clearly states that Phenom 9700 is coming shortly after the two 95W models.Probably the launch was delayed in order to get the microcode updates in time for a new launch date.
And more importantly,they found one of the problems that limited them to introduce more speed bins.Let's hope this helps them bring us faster models in Jan/Feb time frame.

[alucasa]

How about 4P? :yepp:

Way too expensive for my taste. :(

[justapost]

lots of benches we see with ram @ 667mhz, 5-5-5-15...
A little off topic:
After I updated the bios on my m2a-vm board to the latest (phenom supporting) version my 800mhz sticks did no longer run orthos-stable at 800mhz (using an X2-3800 at stock), it ran stable at 667mhz. Going back to an older bios version not supporting phenoms fixed that issue. :shrug:

[JumpingJack]

Unfortunately no phenom review leaked till now. But from what we have seen k10 is very close with core 2 architecture clock for clock.
But for intel, the expected 5-15% speed improvement with the transition vom 65 to 45nm seems to be more like 1-5%. Don't ask for a proof because i do not follow all the intel news atm.


From Anand comparing clock for clock Yorkfield to Kentsfield:

http://www.anandtech.com/cpuchipsets/intel/showdoc.aspx?i=3137&p=4

4 to 7 %

Digging deeper we see performance increases on the order of 4 - 7% if we look at the individual 3D rendering or media encoding tests.


2 to 10%

The clock-for-clock improvement under DivX 6.6 is a nice and round 10%, although curiously enough we get a meager 3% from our Windows Media Encoder test. Our QuickTime H.264 test shows a more average 4% performance increase, similar to our WME test.

5-10%

As expected, we're seeing reasonable gains here. 3dsmax shows a 5% increase, Cinebench 8%, Lightwave 6% and Photoshop CS3 has a healthy 10% performance boost in store for us.

about 4 to 9%

while Crysis gives us a mild 3.7% performance boost. Half-Life 2: Episode Two and Unreal Tournament 3 enjoy a 5.1% and 6.4% increase respectively, while World in Conflict and Quake Wars are both at around 9%.

SSE4 is giving more impressive gains.

On our QX9650, the full search with SSE4 enabled runs about 45% faster than with SSE2 only - impressive! Note also that the Penryn QX9650 offers better SSE2 performance in this test as well, coming in about 61% faster than the QX6850. The total performance increase from QX6850 SSE2 to QX9650 SSE4 in this test is an incredible 133%.

Hexus is getting the same results http://www.hexus.net/content/item.php?item=10240

number of evolutionary architectural improvements gives the 45nm processor an up-to 10 per cent performance lead over an equivalently-clocked 65nm (QX6850) part.

Hardwarezone is showing up to 10% as well: http://www.hardwarezone.com/articles/view.php?id=2404&cid=2&pg=13

. If all the QX9650 was able to do was to perform 10% faster than the QX6850, we call it an upgrade. But if the QX9650 can perform 10% faster than the QX6850 while using 20% less power, that's value right there with a capital V.

I simply did a random sampling of the HW sites in my favorites folder and chose ones who made the same sweaping type statements on clock for clock improvements as you did.

1-5%, as you claim, is meager... 4-10% (clock for clock, IPC) with a simple shrink is healthy and well done (there was minimal architectural differences) maybe not what you had hoped for but it is healthy -- typically in shrinks, you don't expect IPC improvements and this is just clock for clock ... clock speeds and thermals are much different.

With a full on architectural overhaul AMD improved IPC 15%, with a larger cache and 2 or 3 tweaks Intel improved up to 10%.... while we compare Phenom to Q6600, because this is where the price point currently lies... the closer to parity Phenom will enjoy will be short lived, as a 2.5 GHz low end Yorkfield is due in a few months.

This is not to say the Phenom is not a good chip, but from a competitive stand point, it's potency is muted by Intel's refresh to 45 nm fairly significantly.

Jack

[Start]

I rather have the problem fixed now, then have it become a problem later. Intel FDIV Bug anyone?

[JumpingJack]

Theo overblown the story.DT clearly states that Phenom 9700 is coming shortly after the two 95W models.Probably the launch was delayed in order to get the microcode updates in time for a new launch date.
And more importantly,they found one of the problems that limited them to introduce more speed bins.Let's hope this helps them bring us faster models in Jan/Feb time frame.

Well, I will believe DT on this one, but do you (and anyone else for that matter) get as much humor out of FUDzilla and the Inq as I do:

http://www.fudzilla.com/index.php?option=com_content&task=view&id=4259&Itemid=1


All 2.4GHz CPUs that were originally planned for launch on Monday, are delayed to late Q1 and naturally all faster Phenoms are to come in late Q1.

Fuad is most likely wrong here.

[justapost]

I simply did a random sampling of the HW sites in my favorites folder and chose ones who made the same sweaping type statements on clock for clock improvements as you did.

Those focus on Improvements, which is ok.

SYSMark 2007's overall performance went up by 2.6% but that's far from significant. If you look at the individual tests however, you'll see that the 3D benchmarks went up by 4.5% and the productivity suite went up by a similar 4.6% margin

I just took the mean of improvements for anandtech's SysMark and WorldBench results.

Sysmark:

Max 4,6%
Min: 0,7%
Mean: 2,63%

WorldBench:

Max: 7,36%
Min: -3,41%
Mean: 1,67%

You have to use alot of significant improving benchmarks to get a better mean. Both companies use this simple technik.

[JumpingJack]

Those focus on Improvements, which is ok.

I just took the mean of improvements for anandtech's SysMark and WorldBench results.

Sysmark:

Max 4,6%
Min: 0,7%
Mean: 2,63%

WorldBench:

Max: 7,36%
Min: -3,41%
Mean: 1,67%

You have to use alot of significant improving benchmarks to get a better mean. Both companies use this simple technik.

I see you focus on synthetics...

Anyway, those numbers were not means, but app by app improvements giving a range of improvement between 4 to 10%.... SSE4 higher for code compiled for SSE4.

What you apparently are doing is taking two data points, breaking them down into the consituent parts from the synthetics and making a generalized statement on the improvements, ignoring the realworld results. Hmmmmmm.... I think I will take the real world result, knowing that if I run Photoshop I can expect 10% or if I run Quake 4, I can expect around 9% ... just what the data says.

Jack

[Inteleron]

so when all is said and done, is phenom launching tomm?

[Zytek_Fan]

so when all is said and done, is phenom launching tomm?

Yepp. I don't think there will be any leaks tonight though. NDA ends at 8 AM PST I believe.

[Start]

Yepp. I don't think there will be any leaks tonight though. NDA ends at 8 AM PST I believe.

Boo I want leaks dammit!

[Inteleron]

Boo I want leaks dammit!

thats what im talking about hahaha. Im thinking of picking up one over the cheaper lower clocked phenoms with the new DFI LP UT 790FX boards and oc'ing that sucker till the faster phenoms come out.

[justapost]

I see you focus on synthetics...

Anyway, those numbers were not means, but app by app improvements giving a range of improvement between 4 to 10%.... SSE4 higher for code compiled for SSE4.
Look at that WorldBech results there a two results with an performance improvement around 4% and one with a 7% improvement. The mean of the other 8 tests shows a 0,32% improvement.
So in 28% of the test penry shows a 4-7% improvement
This review shows in what areas penry improves alot, you can not take this 4-7% improvement as the over all architecture improvement.

[JumpingJack]

Look at that WorldBech results there a two results with an performance improvement around 4% and one with a 7% improvement. The mean of the other 8 tests shows a 0,32% improvement.
So in 28% of the test penry shows a 4-7% improvement
This review shows in what areas penry improves alot, you can not take this 4-7% improvement as the over all architecture improvement.

Ok, that is what the synthetic says, I am not arguing that... but actual apps are showing different:

3DMax9 5.5%
Cinebench 8.4%
Lightwave 5.5%
Photoshop CS3 (one of my favorites) 13.8%

You can tout synthetics all you want.... realworld is what you use, and all review sites are getting the same results. You cannot focus in on one data point and use that to generalize a view you wish to be true.... this is just being blind.

[justapost]

Ok, that is what the synthetic says, I am not arguing that... but actual apps are showing different:

3DMax9 5.5%
Cinebench 8.4%
Lightwave 5.5%
Photoshop CS3 (one of my favorites) 13.8%

You can tout synthetics all you want.... realworld is what you use, and all review sites are getting the same results. You cannot focus in on one data point and use that to generalize a view you wish to be true.... this is just being blind.
Sandra is a Synthetic Benchmark. I see this WorldBench thing for the first time, but it looks like it uses real world apps.

Photoshop CS2, Firefox, 3dsmax 8, Office 2003, Nero and Winzip sound like real apps to me.
I still use CS2 so now performance gain for me.

[JumpingJack]

Sandra is a Synthetic Benchmark. I see this WorldBench thing for the first time, but it looks like it uses real world apps.

Photoshop CS2, Firefox, 3dsmax 8, Office 2003, Nero and Winzip sound like real apps to me.
I still use CS2 so now performance gain for me.

No, Worldbench emulates those code bases and uses parts of that code, but are not the applications... also, those apps you list are out dated in this version of worldbench... again, it is a synthetic and does not represent realworld.

Data is data, I am not arguing that the synthetics only show maybe 2 or 3 %, this is what they show... but for actual apps, the results are much higher, and like I said ... do you use your computer by running only Worldbench? Or do you run Photoshop (or a consumer derivative) or other apps?

If you are insistent on using Synthetics as the only measure of a CPU capability, then you must conclude that the P4 was a better CPU than K8 as it won many of the synthetics (which most reasonable people put in to perspective, bordering on ignoring).

Jack

[justapost]

No, Worldbench emulates those code bases and uses parts of that code, but are not the applications... also, those apps you list are out dated in this version of worldbench... again, it is a synthetic and does not represent realworld.
Thanks for clarification in terms of worldbench.
But if you use the latest versions of apps compiler optimisations for latest processors have an impact on performance. In real world you can not afford to buy all those updates.

[JumpingJack]

Thanks for clarification in terms of worldbench.
But if you use the latest versions of apps compiler optimisations for latest processors have an impact on performance. In real world you can not afford to buy all those updates.

True.... it is a case by case, but again... what I am guarding against is a generalized statement of 1-5% based on 2 of the 20 data points...

It is ok to say, 1-5% based on synthetics.... it is just as ok to say 4-10% based on actual applications you can buy.

[justapost]

It is ok to say, 1-5% based on synthetics.... it is just as ok to say 4-10% based on actual applications you can buy.
:yepp:

[JumpingJack]

http://www.amd.com/us-en/assets/content_type/white_papers_and_tech_docs/41322.pdf


Errata 254:

Description
Under a highly specific and detailed set of conditions, an internal resource livelock may occur between a TLB reload and other cached operations.

Potential Effect on System
The system may hang.

Suggested Workaround
BIOS should apply the workaround based on the extended model, base model, and stepping reported by CPUID Fn8000_0001 as described in the following table:

Curious -- is the Errata that has now reached a heightened sense of awareness around the web? If so, they knew of this since the BA/B2 stepping, alas AMD has not updated their errata documentation therefore this may be completely different.

Anyone out there in the know?

[frankR]

Ouch !

http://finance.yahoo.com/q/bc?s=AMD&t=2y&l=on&z=m&q=l&c=

[Shintai]

Amazing. They are about worth what they bought ATI for 1½ year ago.

[zakelwe]

To be honest when it comes to tech stocks the value does not follow logic. For instance in the past AMD was worth $3 which was grossely undervaued and then they were worth $40 which was grossely overvalued.

Regards

Andy

[JumpingJack]

To be honest when it comes to tech stocks the value does not follow logic. For instance in the past AMD was worth $3 which was grossely undervaued and then they were worth $40 which was grossely overvalued.

Regards

Andy

Yeah, but it is interesting to note that AMD's market cap at today's 'undervalued' price is only about 600 million or so more than what they are in debt.