AMD issues STOP SHIP order on Barcelona

[Clairvoyant129]

Am I the only one who thinks this is good news? Yeah AMD will miss another 2-3 months, but if thats true... that means K10 is not weak, but broken.

If indeed the fix results in low clockspeed and 10-20% performance hit, resolvin the bug will turn K10 into a solid competitor, even on the desktop front....

In server, a 3GHz K10 with 10-20% higher IPC will wipe the floor with anything intel can offer as good as K8 did with pentium 4

fudzilla already posted there are bug free samples of K10... I hope thats true and B3 is finally ready and fully working

Were you reading the samething I was reading? I guess desperate people can interpret bad news as good news.

[Dainas]

Were you reading the samething I was reading? I guess desperate people can interpret bad news as good news.

Yep, all Hail the eternal optimist. Making us take a double takes of things were sure of since 1945. B3 is pretty much just what the reviewer showed us. Its no messiah of AMD's performance deficit, for what those of us without Barcelona (and the 10% hit fix) have seen.

[hollo]

Hector found Intel responsible for AMD failures!

hector makes AMD look so bad sometimes

"If you look at the last five years, if you look at what major innovations have occurred in computing technology, every single one of them came from AMD. Not a single innovation came from Intel," he claimed.

oh please :/
this guy needs to see a psychiatrist

[Timbosan]

It is disappointing, i hope AMD can get it working and provide some competition for intel. I am confused though, as to why people are saying these AMD cpu's will be better in server applications than intel cpu's. Is it just that AMD cpu's are better at the kind of work a server does? I would have thought power efficiency would come into it somewhat- if the servers built on AMD are pulling ~35% more power how are they better?

"If you look at the last five years, if you look at what major innovations have occurred in computing technology, every single one of them came from AMD. Not a single innovation came from Intel," he claimed.

Hilarious.

[Donnie27]

"Not a single innovation came from Intel"
I guess he was referring to High-K and metal gate 45nm.

Sure and he's not using PCI-E and AGP before that. He sure as hell ain't using SIMD's either LOL! Oh brother remove and the Alpha tech from AMD and see how much innovation is left?

[Tekdemon]

Am I the only one who thinks this is good news? Yeah AMD will miss another 2-3 months, but if thats true... that means K10 is not weak, but broken.

If indeed the fix results in low clockspeed and 10-20% performance hit, resolvin the bug will turn K10 into a solid competitor, even on the desktop front....

In server, a 3GHz K10 with 10-20% higher IPC will wipe the floor with anything intel can offer as good as K8 did with pentium 4

fudzilla already posted there are bug free samples of K10... I hope thats true and B3 is finally ready and fully working

Problem is that the hardware fix would only make the Phenom match what the benchmarks had been saying it would be....and the fix isn't coming until late Q1 2008.
And I don't see how this will help them with clockspeeds at all, if anything the fix might end up screwing up their yields even worse since now they're busy trying to fix this erratum instead of focusing on tweaks to improve yields.

There's no way to spin this as good news I'm afraid. But then again I remember back when I used to cheer for AMD in the K5 days, so I'm kinda used to them being on the verge of bankruptcy and screwing up chip launches all the time. Still, AMD pissed me off when they introduced the x86-64 stuff, because I had honestly been looking forward to seeing itanium on the desktop.

[Zytek_Fan]

Sure and he's not using PCI-E and AGP before that. He sure as hell ain't using SIMD's either LOL! Oh brother remove and the Alpha tech from AMD and see how much innovation is left?

Hector is just trying to make himself look better before he leaves
It's not working.

[JumpingJack]

So with B3 it should be about same performance as Core 2 clock for clock. Only cuz of this fix that has a huge performance hit on B2 stepping, they have to lower the price on K10 and they can't really higher the price when B3 arrives either. That seems quite embarrassing to me. Guess they will try to price it a bit expensier than it's worth compared to Core 2 CPUs meanwhile.

Unfortunately, this will not be the case I suspect.... the errata is a livelock error when multiple processes are competing for the same resource. Simple solving the logic to correctly share the resource in those rare cases will not necessarily translate to any performance improvement.

If you are expecting 10-20% improvement for B3, then the data we saw one would expect implemented the work around... this, in fact, is not the case:

Saucier confirmed to us that the test systems at the Tahoe press event did not have the workaround enabled.

http://www.techreport.com/discussions.x/13721

Thus the data collected from most all the reviews indeed let the TLB run without any portion diabled ... so what you saw is likely what you should expect.

But, then it could improve for other reasons ... so there is some hope.

[Zytek_Fan]

Lovely mess that we've got right now

[Brother Esau]

Hector found Intel responsible for AMD failures!

I would honestly agree with that and say thats more than likely very true!

[RAMMAN]

I wonder.. why AMD had to BUY ATI, couldn't they just merge, like activision and blizzard do it now... no one is buying anyone... they just pour some cash in the deal and thats it. A merger wouldn't cost both AMD and ATI more than a billion... That deal set amd in a position where they lack any cash to work with... they should have spent that $$$ on RD and a new chip facility. I think there ware some personal interests behind the buyout...

hector was too proud to merge with ati, and too stupid to merge with nvidia.if amd had merged with nvidia their debt troubles would have been gone forever!

[awdrifter]

But Nvidia probably doesn't want to merge with them. Sometimes merger (and buyouts) works out, sometimes it doesn't. AMD had a great idea buying ATI, but ATI's product didn't deliver, and at the same time AMD's having trouble with K10. So it's part bad decision, part bad luck.

[saaya]

Seems to me that AMD should be publicly *recalling* all K10 parts, including Phenom.

Intel is going to take virtually all of their server share over the next few months.

very reasonable comments

[Hornet331]

I own 2 personal PCs based on intel chips, and 1 AMD, based on your comment you are twice the fanboy you think I am

owning x ammount pc pc/cpus, tells nothing about beeing a fanboy or not.

I own currently 3 Amd PC and one Intel pc (+ one intel notebook), and i prefer intel right now, because it delivers the performance im looking for.

[BeardyMan]

This post made by legend macci was pretty promisfull where he tested that a 3ghz phenom beated an 3ghz yorkfield with 300points although the phenom was on the slowish vista and the yorkfield on XP¨, but all this lol pretty sad, i guess this partly due the media pushing things, if you can't get the job done in X time then don't lol, releasing an unfinished product :/

[DoubleZero]

See how techreport, manage to turn a overall positive article about AMD in a negative one.

http://www.xtremesystems.org/forums/...d.php?t=162378
http://www.crn.com/white-box/202402138?pgno=1
http://www.techreport.com/discussions.x/13389

The low availability of barcelona is nothing new, but techreport managed to used it again for yet more AMD negative crap.

When asked for comment, spokesman Phil Hughes said AMD is shipping Barcelona Opterons now, but only for "specific customer deals." Industry sources have suggested to TR that those deals are high-volume situations involving supercomputing clusters. Such customers may run workloads less likely to be affected by any workarounds for the erratum that reduce L3 cache performance, and those customers could potentially consume hundreds of thousands of CPUs.

It's only me that thinks this is incredibly dumb? Supercomputing clusters don't run the workloads, but the desktop everyday guy run? lol it's dumb but it's extremely convenient for techreport article.

The fact that the most rabid Intel fans on this forum are constantly praising and using techreport fud, should tell you something...

[Jacky]

@Beardyman
A heavily overclocked K10 with a heaviy overclocked NB speed beats a stock (?) Yorkfield on an useless and almost obsolete 3D benchmark? and there was different software involved, wow

btw the stock lost 4% yesterday (nasdaq 0.66%). I wonder if that's the reason or if it's just the overall trend and it may get even worse..

Anyway I don't understand the situation. No confirmation by any other news sites. Is it even true? No Barcelona for major OEMs and most of server vendors who want to be sure not to have crashing systems? But desktop gets "faulty" phenoms which may crash when doing e.g. virtualization? (as mentioned in the article..)

doublezero, your post only confirms that it always has been a paper launch and the articles you posted are VERY, VERY old, what meaning does a 2 months old article have in the tech industry?!
And do you think an AMD executive would not lie/spin to make AMD look not as bad?
" Supercomputing clusters don't run the workloads, but the desktop everyday guy run? lol it's dumb but it's extremely convenient for techreport article."
Your making up your own story here.. it's about Barc, that's not a Desktop part and HPC is absolutely different to most server workloads, so it's very possible that some HPC customers - running always the same workload - will never encouter this bug.
However, you're right, the whole story may be overly exaggerated but you should keep your accusations of credible techsites to yourself until we have a definite answer or you may look very stupid in the end.

[alucasa]

I believe Bracelona are already being withdrawn. Try to look for a chip now. You won't find any stock at all.
At least, Newegg used to have few left, but they removed the chips from their list. (Can anyone else confirm this as well?)
Furthermore, I also ordered extra two chips (Besides the ones I have), but their ETA also has been postponed indefinitely.

[savantu]

...
Can you support your claim with anything else than fanpoism? Explain this: QX9770 - Hello Global Warming

...

Anandtech got a bad chip and the mobo was overvolting it.

http://www.legitreviews.com/article/596/11/

Look for any other 9770 review and you'll see the same story : the chip is hot , but not that hot.

[Syn.]

This is the way i believe things happened.

Hector calls a meeting of management to decide when to launch K10. In the meeting he is told that if he wants to launch K10 in Q2 07 he will be left with slow speed ramp which will be no way near the performance needed for the chips to worth buying. They work out that with B2 stepping they might get an improved speed steppings but will have to delay launch by another quarter. After some discussion it is decided to wait for B2 to be ready for mass production instead of launching with slow BA chips.

Few month's later and another meeting is called.

The B2 samples show that the chip can take the speeds up to 3Ghz but the TDP is too high at that speed so they will have to launch at more moderate speeds to keep with in their company strategy of increasing performance but keeping power levels the same. The B2 chips go into QC testing to see if they are ready. Early results are showing all OK and Hector takes a gamble and decides to go forth with the already delayed launch before all test are complete. With AMD being in loss for 2 quarters the pressure was to deliver new product as fast as possible to prevent further losses. As the mass production is already underway some bugs start appear in the faster bin's AMD investigates and shows that L3 cache is dodgy and will need some patching up. To prevent the press from catching wind of any technical faults with the chips Hector and his team decide to have a controlled review event. The patched systems seem too unworthy of having and faced with poor reviews and poor sales AMD makes a move to fully disclose what some industry folks have already found out about the chips and while putting on a fake smile and trying to keep as many customers buying to its products AMD stops the B2 production and moves to fix all bugs and relaunch with B3 stepping. This meant that the chips wont be available till few months later due to their fab constraints. Meanwhile the CPU's that they have made go to special orders which have controlled environments and can deal with the bugs in the B2 CPU's, these orders are probably kept due to lowered price of the chips as a sort of "Closing Down Sale".

As a result of another delay of K10 motherboard manufacturers decided to lower AM2+ motherboard production leaving low supplies of CPU's and motherboards all round.

Folks like TechReport are only thinking about them selfs. Intel had the power to release early samples of its crappy P4's because its PR machine can eat up any negative points AMD does not which is why it was left with very poor choices which they had to make. If i remember correctly Core 2 was also controled review event.

In other words none said life is a smooth ride AMD was lucky with K8 working out nicely however with K10 they where faced with having their PR team put the best spin they can so that as a company they can cushion their looses as much as possible like any other company would do. In the mean time they try to point finger at Intel to divert attention from them self's. Which backfires because none cares about the monopoly discussions anymore they just want AMD to keep up with Intel so that when they buy their PC's this Christmas they would be nice and cheap.

[BeardyMan]

@Beardyman
A heavily overclocked K10 with a heaviy overclocked NB speed beats a stock (?) Yorkfield on an useless and almost obsolete 3D benchmark? and there was different software involved, wow

btw the stock lost 4% yesterday (nasdaq 0.66%). I wonder if that's the reason or if it's just the overall trend and it may get even worse..

Anyway I don't understand the situation. No confirmation by any other news sites. Is it even true? No Barcelona for major OEMs and most of server vendors who want to be sure not to have crashing systems? But desktop gets "faulty" phenoms which may crash when doing e.g. virtualization? (as mentioned in the article..)

doublezero, your post only confirms that it always has been a paper launch and the articles you posted are VERY, VERY old, what meaning does a 2 months old article have in the tech industry?!
And do you think an AMD executive would not lie/spin to make AMD look not as bad?
" Supercomputing clusters don't run the workloads, but the desktop everyday guy run? lol it's dumb but it's extremely convenient for techreport article."
Your making up your own story here.. it's about Barc, that's not a Desktop part and HPC is absolutely different to most server workloads, so it's very possible that some HPC customers - running always the same workload - will never encouter this bug.
However, you're right, the whole story may be overly exaggerated but you should keep your accusations of credible techsites to yourself until we have a definite answer or you may look very stupid in the end.

HTT doesn't do much for K10 and NB should be close to 2.6ghz stock for a 3Ghz part

[grunge100]

Six pages of AMD bashing in 1.5 days-- This is so unlike this forum.

AMD has some major problems with K10, We all know this. Quit beating it into the ground.

When Intel was having problem with prescott, these forums were not nearly this visious. What gives?

Even positive AMD threads turn into sh**.

[Jacky]

When Intel was having problem with prescott, these forums were not nearly this visious. What gives?

Perhaps they should have been. You get what you deserve.

EDIT:
annandtech/dailytech "the rabid Intelfanboys" ( ) comment on the issue here and here.

[BeardyMan]

Perhaps they should have been. You get what you deserve.

EDIT:
annandtech/dailytech "the rabid Intelfanboys" ( ) comment on the issue here and here.

Heres a yorkield with overclocked FSB and 1100mem ves 1000mem on the phenom

and still it can't overhoal K10, K10 ain't bad at all, it just needs maturing and to give yorkfield a run for ist money a magic touch of overclock wonder

And that's where it will fail sadly enough :/

[grunge100]

Jacky,

Thats not the point. I guess people post on forums for various reasons. Within the past year, these forums have become very anti AMD.

K10 sucks, AMD sucks, Hector sucks, AMD process tech. sucks, etc....
Whats the reason for pointing this out in every thread.