Retail Phenoms slower than review samples + more errata info

epsilon84

Distinguished
Oct 24, 2006
1,689
0
19,780
http://www.techreport.com/discussions.x/13724

To recap, the erratum is a chip-level issue involving the TLB logic for the L3 cache that can cause system hangs in specific circumstances. AMD has a fix for the problem in the works, but it degrades performance. AMD has stated publicly that the workaround can lower performance by as much as 10%, although one source characterized the performance hit to TR as 10-20%.

Saucier clarified the exact nature of the workaround for the erratum that AMD has provided to motherboard makers and PC manufacturers. The fix comes in the form of a BIOS update, and this BIOS patch includes an update to the CPU microcode. This update disables the portion of the chip's TLB logic that is problematic. Saucier noted that the L3 cache "still works" with this logic disabled, and he said AMD has no plans to implement the fix for existing chips in a different way.

We don't yet have a BIOS with the workaround to test, but we've already discovered that our Phenom review overstates the performance of the 2.3GHz Phenom. We tested at a 2.3GHz core clock with a 2.0GHz north bridge clock, because AMD told us those speeds were representative of the Phenom 9600. Our production samples of the Phenom 9500 and 9600, however, have north bridge clocks of 1.8GHz. We've already confirmed lower scores in some benchmarks.

Given everything we've learned in the past few days, our review clearly overstates Phenom 9600 performance, as do (more likely than not) other reviews of the product. We can't know entirely by how much, though, until we can test a Phenom system with the TLB erratum workaround applied.

Poor form, AMD. Phenom should NEVER have been released in this state, B2 is clearly a problematic and immature stepping. AMD should do the honourable thing and recall these processors IMO, their customers deserve better than this. :non:

Also, handing out review samples that are faster than retail versions borders on unethical conduct. Say what you will about Intel and they shady business practices, but AMD is no saint either.

I just can't fathom what AMD was thinking here. Surely they realise people would notice a 200MHz discrepancy in the NB speed... was the small performance gain in reviews really worth the bad PR image they are getting now? I think not.
 

Harrisson

Distinguished
Jan 3, 2007
506
0
18,990
I also think they shouldnt have released immature stepping, but AMD was hard pressed for the roadmap timing and probably couldnt allow yet another postponement of launch. Imagine 1000s of threads how "AMD lied again! No cpu's released!!!!" Another reason - its possible AMD found out too late about this, and recalling would be heavier hit on finances than letting it go 10% slower in certain cases.

BTW, C2D also had a bug, but cpu's werent recalled either, same approach of microcode update.

"Also, handing out review samples that are faster than retail versions borders on unethical conduct."
Yes its not the best way, but name ONE company who doesnt do that? You wont find any.
 

the_vorlon

Distinguished
May 3, 2006
365
0
18,780
Harrisson wrote :

BTW, C2D also had a bug, but cpu's werent recalled either, same approach of microcode update.



True... but.....

The C2D microcode update actually fixed the issue and there was no performance hit....

That is a meanngful qualitative difference IMHO.
 

BaronMatrix

Splendid
Dec 14, 2005
6,655
0
25,790
First you should change the title. That's not what Tech Report said. They sid the WORKAROUND in the BIOS lowered performance NOT the retail revision.

Second, wow. I guess with all of the negative AMD posts, I guess my positive ones kind of level the field.
 

lolitha

Distinguished
May 24, 2006
364
0
18,780
""workaround can lower performance by as much as 10%"" that's too much i guess. I want maximum return of value for money. I hope everthing will be alright for AMD otherwise everone knows what will happen
 

Eviltwin17

Distinguished
Feb 21, 2006
520
0
18,990
wow AMD wow

lets see here

AMD: **** processors but great upgradeability in its AM2 platform
Intel: really awesome processors but **** upgradeability since it keeps switching chipsets every new processor iteration it creates: p965 now pretty much useless, p35+ good for penryn but screwed for 1666mhz?, yay new chipset for nehalem when it comes out. WTF intel, take some notes from AMD
 



You are rediculous Baron. People have to apply the fix so that their AMD Crapron system won't hang and crash. So therefore, your system will be slower than all the fancy benchmarks you've seen.

If Intel had an issue like this you would have an electronic orgasm on these forums and would break into song and dance.

Phenom is a disaster of a release with recalls, sub-par performance, it didn't live up to AMD's promises, and the fact that Phenom systems will need a BIOS update that source are quoting as making the systems 10-20% slower...

Phenom = Phailure

Title is accurate, unless you want me to make up one of my creative titles!
 

epsilon84

Distinguished
Oct 24, 2006
1,689
0
19,780


How about you reread the article? Retail Phenoms have a NB speed of 1.8GHz instead of the 2GHz used in reviews.

I even QUOTED the bit where they clearly say a retail Phenom would underperform the reviewer samples. Must have been filtered out by those huge fanboy glasses or something...
 

Mathos

Distinguished
Jun 17, 2007
584
0
18,980
This would be why I never jump right on to new processors or hardware until there has been at least one revision cycle on the process since release. And if I remember correctly it was stated in one of the reviews, Anandtech, if I remember correctly that the phenoms he had were running on 1.8ghz nb memory plane speed. I believe this may likely be due to the chipset used, or the platform. For example they were supposed to run on a 1.8ghz memory controller speed on any regular AM2 board, but could be upped to 2Ghz on the AM2+ boards that used the newer spider platform chipsets.
 

NMDante

Distinguished
Oct 5, 2002
1,588
0
19,780


Anandtech did use the new 790FX based motherboard, which is the Spider platform chipset of choice, if I remember correctly.
As you can see in the test configurations in the link: Anandtech's Test platform.
 
eviltwin17, same can be said about AMD taking notes from Intel. Don't release a product with claims that it is better then underperforms badly.

Intels upgrade right now is pretty bad compared to AMD but as I have said before if you want HT3, CFX and the latest and greatest you will need the newer chipset/mobo.

And the P965 is not useless. Asus shows support for Penryn on their mobos using the P965. Its more a BIOS update than the chipset. The P965 supports up to 130wTDP. It can run a 1333MHz FSB. Its just an official support.

Yes Nehalem will need a newer chipset but hey same thing happened when AMD moved to a IMC so that is expected.

As for the retail versions performing slowr than the tests versions that just goes to show AMD was trying to make their setuo look better than it truly was. And Anandtech can use a 790FX chipset but what about those who use a older AM2 mobo? Does this mean that the chip will perform even slower? Thats just sad to me.
 

That was for the most part a bad idea IMO too

@ NMDante \/ - Yes its more a timing thing then anything else....
 

NMDante

Distinguished
Oct 5, 2002
1,588
0
19,780


Actually, it was a good move for AMD, but the timing was very bad. It put AMD in a large debt, that it never had before, and when Core 2 was introduced, it limited their profits, which made that debt even worse for them.

If they had gotten ATI during the K8 heyday, it might have been a very good thing for AMD, unfortunately, they did the buyout at the worse possible time for them, and now they are feeling the after effects of the deal.
 

pausert20

Distinguished
Jun 28, 2006
577
0
18,980
I agree with NMDante. To compete effectively with Intel, AMD needed to get their own chipset in house and not leave it to third part vendors. I also agree that they could not have bought ATI at a worse time.

I will go on record that the AMD + ATI acquisition will go down in the business annuals as a text case on how not to do a acquisition with issues of timing being the main theme.
 

Evilonigiri

Splendid
Jun 8, 2007
4,381
0
22,780

I think you should. I've been a fan of your creative titles. What will you call it this time? "Another Phenomenally Phantastic Phailure Phrom Phenom" LOL :D

A noob question here but NB speed? I never knew there was one...does it change with the FSB?
 

BaronMatrix

Splendid
Dec 14, 2005
6,655
0
25,790



So you mean I make posts about Intel? NOT!!! If they had a problem with 400M transistors, I'd say that's what happens. NONE OF THE REVIEWERS HAD THE PATCH APPLIED IN TAHOE AND NO ONE REPORTED CRASHES.

But I've got an idea. Stop thinking about it and don't plan to buy one.
 

BaronMatrix

Splendid
Dec 14, 2005
6,655
0
25,790



I've heard enough doom and gloom, I guess. I guess we'll see what B3 runs at. It is possible that the error is exacerbated by the higher link speed. But the Tech Report also said that early samples were B3 so I guess they had to pull the trigger with the BA and B2 revs.

I'm confident that the makers of X2 will sort out the problems. You all act like this is AMDs first chip and it means they can't make CPUs.
 

zenmaster

Splendid
Feb 21, 2006
3,867
0
22,790
 


Tahoe was a freaking joke. AMD wasn't confident enough in their product to ship it out to reviewers and forced them to come to Tahoe and wouldn't let them raise voltages (read: OVERCLOCK).

We'll have to see Baron. I'm not convinced by one TechReport article, but I fear more will follow this week, if so, look to see AMD @ 8.50 a share