Phenom B3 in late Q1. B2 BIOS fix costs 10% performance (Fudzilla)

[terrace215]

He also says the 2.4GHz Phenom 9700 will get pushed out to Q1.

http://www.fudzilla.com/index.php?option=com_content&task=view&id=4261&Itemid=1


Phenom 2.4GHz has an "errata"
Written by Fuad Abazovic
Monday, 19 November 2007 04:11

Serious enough to push it to Q1 08


AMD has a huge problem with its B2 revision of Phenoms. It was supposed to launch 2.2GHz, 2.3GHz and 2.4GHz CPU, but it looks like the company will delay its 2.4GHz CPU to Q1 2008. AMD doesn't have any choice as the 2.4GHz part is crashing in some heavy load scenarios.

According to senior sources from the company, AMD's 2.4 GHz Phenom has an errata that can crash the CPU in some heavy load multitasking but it is “highly unlikely that this will ever happen”. Luckily there is a fix the “errata”, but if you turn the fix on, you automatically lose about 10 percent of the performance. In the real world, it won't really work well enough to allow the company to launch the 2.4GHz CPU.

The new B3 stepping will solve the errata and will allow AMD to go to 2.6GHz or faster and the new stepping comes in Q1 2008, rather late than early.

This is definitely catastrophic news for this fallen hero, but the 2.2 and 2.3GHz parts will sell for less than €200 to try to fight Intel.


(I figured there was enough new info here, re: B3 stepping timing, to warrant a new thread. If not, feel free to move this into the 9700 Phenom recall thread.)

[TSFroggy]

Labeled Fud until more reliable sources can confirm.

[terrace215]

Labeled Fud until more reliable sources can confirm.

Well, Inq, Fud, *and* DailyTech all confirm that the 2.4 part is not launching tomorrow, and is delayed.

It certainly would be expected that any BIOS workaround for this sort of problem would have a performance impact, and considering that it sits in the cache/mem interface, not in some obscure instruction, it would not be surprising for the hit to be substantial.

EDIT: Apparently AMD confirmed, on the record:

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

2.4GHz Phenoms and faster to come in late Q1
Written by Fuad Abazovic
Monday, 19 November 2007 03:24



B3 stepping


Today, on Monday the 18th AMD will finally introduce 2.2GHz Phenom 9500 and 2.3GHz clocked Phenom 9600. The 2.4GHz will be delayed until late Q1 and so will any faster Phenom. There won't be any 2.6, 2.8 or faster parts until late Q1.

The B2 revision of Phenom Agena based CPU, works well at 2.2 and 2.3GHz and anything faster have this infamous errata that can freeze the machine. That is why AMD plans to launch 2.2 and 2.3GHz CPUs only.

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.


BA stepping of Agena K10 core was looking good and the B2 stepping was a production one, but this bug unfortunately found its place on silicon and AMD cannot get around it.

AMD said on the record that Phenom 2.4GHz and faster parts will be the B3 stepping and this one is planned in late Q1. Doesn't look good at all.

[JumpingJack]

Labeled Fud until more reliable sources can confirm.

Well, as stated above -- a credible site has already confirmed it. What, why and how much comes from the FUD sites, so that is debatable.

[hecktic]

ouch.... bad for intel fans too..... intel can revise release dates of new chips now since qx9650 is the current rein champion and bringing in the $$$

hopefully amd gets their foot out of their ass and fixes this before christmas and still releases them:cord:

[JumpingJack]

ouch.... bad for intel fans too..... intel can revise release dates of new chips now since qx9650 is the current rein champion and bringing in the $$$

Yes indeedy, bad overall.

[hecktic]

i was looking forward to the none "X" or unlocked intel chips at a good price but there goes that idea... at least Im speculating what will happen now to Intel plans

[terrace215]

http://biz.yahoo.com/bw/071119/20071118005069.html?.v=1

AMD's Phenom PR does not even mention the 9700.

I would guess that, if it were coming within the next month or so, they would've mentioned it, given a price, and said "available within 30 days".

So it looks like Fudzilla is correct on this one.

[JumpingJack]

i was looking forward to the none "X" or unlocked intel chips at a good price but there goes that idea... at least Im speculating what will happen now to Intel plans

Well, if the FUDzilla and Inq rumors are to be believed (or true), then tomorrow's launch will produce a top bin that will underperform Intel's lowest bin quads.... i.e. Intel has no incentive to push pricing down because they already have the performance advantage.

I am most certain that if AMD had launched at speeds as planned, on time then we would have seen a much wider breadth of 45 nm parts launched from intel at better price points. For now, AMD is providing Intel no real motiviation to push their 65 nm line to obsolesence.

[JumpingJack]

http://biz.yahoo.com/bw/071119/20071118005069.html?.v=1

AMD's Phenom PR does not even mention the 9700.

I would guess that, if it were coming within the next month or so, they would've mentioned it, given a price, and said "available within 30 days".

So it looks like Fudzilla is correct on this one.

Do you find it interest the sudden emphasis on platform and not the CPU??? It wasn't too long ago that AMD derided Intel for the 'platform' approach, claiming it took choice away from the consumer (and there is a bit of truth in that).

[terrace215]

Do you find it interest the sudden emphasis on platform and not the CPU??? It wasn't too long ago that AMD derided Intel for the 'platform' approach, claiming it took choice away from the consumer (and there is a bit of truth in that).

Yes, I was amused. Also, of course, the lack of any performance claims. What can they say, really?

They're left with some babble about "cutting edge", power efficiency, and... well, that's it.

TechReport kinda pans the 790 chipset, as well.

http://www.techreport.com/articles.x/13628/14


Conclusions
Spider is an ambitious attempt at an enthusiast platform, and while Phenom and the Radeon HD 3800 series might grab all the headlines, the 790FX chipset is undoubtedly a key component of the overall equation. With support for HyperTransport 3 and second generation PCI Express, the 790FX has the high-bandwidth connectivity needed to fully exploit Phenom processors on one end and Radeon HD 3800 series graphics cards on the other. In fact, the 790FX is the only chipset currently on the market that fully supports all the features packed into new Phenom processors. Couple that with excellent OverDrive tweaking and overclocking software that we're told will be widely-supported by motherboard makers, and the 790FX looks pretty good.

Except for one or two not so minor problems.

Despite being the first new enthusiast chipset for AM2 processors in more than a year and a half, the 790FX relies on a dated SB600 south bridge chip whose limited connectivity options hardly befit a cutting-edge enthusiast platform. The SB600's comparatively poor PCI and USB performance only make matters worse. With a new SB700 south bridge due to arrive early next year on a second wave of 790FX boards, it's hard to get excited about current implementations.

The first crop of 790FX boards certainly has its share of problems, too. Gigabyte's GA-MA790FX-DQ6 suffers from a number of BIOS-related issues that really need to be fixed before we would even consider recommending the board. And it's not like this is a pre-release product whose kinks are still being worked out; the DQ6 was selling online before the 790FX embargo even lifted. The DQ6 also suffers from a poor graphics slot layout and alarmingly high power consumption given the 790FX's 10W TDP. It's not cheap, either. With street prices currently hovering around $270, the DQ6 costs nearly as much as the fastest Phenom you can buy.

Our problems with 790FX motherboards haven't been limited to Gigabyte, either. We haven't yet had the opportunity to test MSI's riff on the Spider platform, but we've found Asus' M3A32-MVP Deluxe to have numerous BIOS-related issues and questionable stability. The fact that the top two mobo makers can't get their initial 790FX offerings dialed for the chipset's launch doesn't give us a lot of faith in the platform.

I'm not even convinced there will be much demand for a high-end Phenom chipset. The processor's performance isn't compelling enough to inspire current Core 2 users to jump ship, and those running Socket AM2 processors will probably be more inclined to drop a Phenom into their existing motherboards rather than going out and buying new ones.

In the end, the 790FX is a bandwidth-rich new north bridge chip held back by a dated south bridge and flakey motherboard implementations. AMD will have to do better if it expects the Spider platform to take off with enthusiasts. In its current state, we simply can't recommend the 790FX.

[terrace215]

Hexus confirms 9700 pulled, will not be available until some time in Q1 2008:


The fact that 9600 will get an unlocked multiplier means it is effectively the Phenom FX until late Q1. So there you have it: Phenom FX at 2.3GHz, and Phenom at 2.2 GHz.

Pathetic.

http://www.hexus.net/content/item.php?item=10427&page=3


Facts and figures

Three quad-core Phenom processors were scheduled to be launched today - the 2.0GHz-clocked Phenom 9500, 2.2GHz Phenom 9600 and 2.4GHz Phenom 9700. However, at the last minute, AMD pulled the release of the Phenom 9700 for reasons that we have still to fathom. We can only surmise that it had significant issues in speed-binning parts at that speed and chose to run with the two lower-clocked models instead. Phenom 9700 will now launch in Q1 2008, we're informed, putting a significant dampener on the AMD roadmap for 2007.
Processors Clock speed L2 cache (total) L3 cache Voltage TDP Price
AMD Phenom 9700* 2.4GHz 2MiB 2MiB 1.25V 125W ?
AMD Phenom 9600 2.3GHz 2MiB 2MiB 1.2V 95W £159
AMD Phenom 9500 2.2GHz 2MiB 2MiB 1.2V 95W £145

* - now rescheduled for Q1 2008.

The two released quad-core Phenom processors have the same architectural underpinning but the 9600 will be released with an unlocked multiplier, we were informed at last minute.

[Brother Esau]

Too bad they are just getting around to replacing Hector after he has done the damage:mad:

[xlink]

PIII 1.13GHz...

[terrace215]

Too bad they are just getting around to replacing Hector after he has done the damage:mad:

To be fair, Hector was responsible for the boondoggle of spending 5.6 Billion in cash to buy ATI, but you can't really blame him (other than in a 'CEO is ultimately responsible' way) for the process and design problems that plague 65nm and the K10.

[BrowncoatGR]

TechReport kinda pans the 790 chipset, as well.

http://www.techreport.com/articles.x/13628/14


Conclusions
Spider is an ambitious attempt at an enthusiast platform, and while Phenom and the Radeon HD 3800 series might grab all the headlines, the 790FX chipset is undoubtedly a key component of the overall equation. With support for HyperTransport 3 and second generation PCI Express, the 790FX has the high-bandwidth connectivity needed to fully exploit Phenom processors on one end and Radeon HD 3800 series graphics cards on the other. In fact, the 790FX is the only chipset currently on the market that fully supports all the features packed into new Phenom processors. Couple that with excellent OverDrive tweaking and overclocking software that we're told will be widely-supported by motherboard makers, and the 790FX looks pretty good.

Except for one or two not so minor problems.

Despite being the first new enthusiast chipset for AM2 processors in more than a year and a half, the 790FX relies on a dated SB600 south bridge chip whose limited connectivity options hardly befit a cutting-edge enthusiast platform. The SB600's comparatively poor PCI and USB performance only make matters worse. With a new SB700 south bridge due to arrive early next year on a second wave of 790FX boards, it's hard to get excited about current implementations.

The first crop of 790FX boards certainly has its share of problems, too. Gigabyte's GA-MA790FX-DQ6 suffers from a number of BIOS-related issues that really need to be fixed before we would even consider recommending the board. And it's not like this is a pre-release product whose kinks are still being worked out; the DQ6 was selling online before the 790FX embargo even lifted. The DQ6 also suffers from a poor graphics slot layout and alarmingly high power consumption given the 790FX's 10W TDP. It's not cheap, either. With street prices currently hovering around $270, the DQ6 costs nearly as much as the fastest Phenom you can buy.

Our problems with 790FX motherboards haven't been limited to Gigabyte, either. We haven't yet had the opportunity to test MSI's riff on the Spider platform, but we've found Asus' M3A32-MVP Deluxe to have numerous BIOS-related issues and questionable stability. The fact that the top two mobo makers can't get their initial 790FX offerings dialed for the chipset's launch doesn't give us a lot of faith in the platform.

I'm not even convinced there will be much demand for a high-end Phenom chipset. The processor's performance isn't compelling enough to inspire current Core 2 users to jump ship, and those running Socket AM2 processors will probably be more inclined to drop a Phenom into their existing motherboards rather than going out and buying new ones.

In the end, the 790FX is a bandwidth-rich new north bridge chip held back by a dated south bridge and flakey motherboard implementations. AMD will have to do better if it expects the Spider platform to take off with enthusiasts. In its current state, we simply can't recommend the 790FX.


1) There will be SB700 based boards in Q1 next year
2) Bios problems are to be expected when a new chipset launches. I dont see how that is suprising

[xlink]

1) There will be SB700 based boards in Q1 next year
2) Bios problems are to be expected when a new chipset launches. I dont see how that is suprising

it's not necessarily a BIOS problem per-se

it's just that the 'fix' more likely than not, dramatically loosens timings. Again think back the the 1.13Ghz PIII which was slower than the 1Ghz PIII due to "BIOS optimization"

[STEvil]

P3 1.13 was faster than P3 1.0, the problem was with vcore supplied to the cpu not allowing quite a few of them to operate reliably at 1.13ghz.

Early P4's were slower than P3's (mostly 1.2ghz+ Tualatin cores, not the 1.13ghz and under Coppermine core) is what you are probably remembering.

[villa1n]

Think we can still clock the 9600 to 3.0ghz with the unlocked multi? ^^
Well on the bright side, better they found it, than me being stuck with a 9700 that will crash if i orthos test it for stability.

[terrace215]

Anandtech: Phenom is a dud.

http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=3153&p=11

[villa1n]

Anandtech: Phenom is a dud.

http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=3153&p=11


this from the anand article

"Currently, the L3 cache/NB on these chips runs at a fixed frequency that's actually lower than the rest of the CPU frequency: 2.0GHz. We tested Phenoms running from 2.2GHz all the way up to 2.6GHz, and in all cases the L3 cache and North Bridge ran at 2.0GHz. We're not sure if this will ever get fixed, but it's somewhat disappointing as it was supposed to be a major reason for upgrading to Socket-AM2+ (but it's good news for current AM2 owners)."

Do you think this is what they intend to fix with B3?

[[XC]Atomicpineapple]

B3 is all about fixing an errata that causes the system to crash under certain high workload conditions. I doubt the L3 cache and NB frequencies are the cause of that.

[YukonTrooper]

PHENOMinal Phlop.

[Brother Esau]

To be fair, Hector was responsible for the boondoggle of spending 5.6 Billion in cash to buy ATI, but you can't really blame him (other than in a 'CEO is ultimately responsible' way) for the process and design problems that plague 65nm and the K10.

When you are a main contender for Intel and they are ahead and you make a decision to scale back R&D yes I blame him for being a Dumb Ass:mad:

[knightwolf654]

hmmm i just read a few other reviews and they said a few weeks.

[BeardyMan]

LOL screw AMD i have had enough with this :banana::banana::banana::banana::banana: :)

I benched Intel for the past year and it wasn't as fun as overclcoking amd, but this is just the same bs. OMG

Goes order intel stuff :down:

[GoThr3k]

You can lick MR Smartass his ass beardyman :yepp:

[BeardyMan]

You can lick MR Smartass his ass beardyman :yepp:

hes banned so bad luck for him :D

[`danny]

Too bad they are just getting around to replacing Hector after he has done the damage:mad:

Yea how fortunate for us consumers eh?

[Miwo]

wow....really dissapointing news to AMD
hopefully the 3850s and 3870s will help them stay afloat and bring in some much needed cash

[DoubleZero]

Lets hope AMD got all the screwing up out of it system with k10 and that 45nm & Bulldozer will be flawless.

[Epsilon84]

LOL screw AMD i have had enough with this :banana::banana::banana::banana::banana: :)

I benched Intel for the past year and it wasn't as fun as overclcoking amd, but this is just the same bs. OMG

Goes order intel stuff :down:

LOL, didn't think I'll see this coming. Weren't you just saying yesterday that Phenom 'PWNED' C2Q? :ROTF:

[Hornet331]

you can't escape reality. :)

[GoThr3k]

you can't escape reality. :)

perception is reality ;)

[BeardyMan]

LOL, didn't think I'll see this coming. Weren't you just saying yesterday that Phenom 'PWNED' C2Q? :ROTF:
Yes i was and i will keep continueing doing that :D

To much intel fanboys in here, and the amd boys can't stand their ground :D


:flame: :shoot: :CTF: :explode2: p

[Epsilon84]

you can't escape reality. :)

I beg to differ.

http://www.welovetheiraqiinformationminister.com/

[justapost]

According to this german article (http://www.heise.de/newsticker/meldung/99177), the bios fix disables the whole L3 TLB.

[informal]

According to this german article (http://www.heise.de/newsticker/meldung/99177), the bios fix disables the whole L3 TLB.

There goes those 10% :(

[xlink]

P3 1.13 was faster than P3 1.0, the problem was with vcore supplied to the cpu not allowing quite a few of them to operate reliably at 1.13ghz.

Early P4's were slower than P3's (mostly 1.2ghz+ Tualatin cores, not the 1.13ghz and under Coppermine core) is what you are probably remembering.

I'd heard BIOS work arounds to get the 1.13Ghz copermine stable negatively affected performance.

[Hornet331]

According to this german article (http://www.heise.de/newsticker/meldung/99177), the bios fix disables the whole L3 TLB.

lol, what are you reading?

heise said amd can bypass the TLB bug, by disabeling the L3 cache and it would cost 10% or more of perforamnce if they disable it.

Theres nowhere mentioned that the bios disables anything.

[justapost]

lol, what are you reading?

heise said amd can bypass the TLB bug, by disabeling the L3 cache and it would cost 10% or more of perforamnce if they disable it.

Theres nowhere mentioned that the bios disables anything.

Bist du a zuagroasder? :)

AMD kann den Fehler durch Deaktivieren des gesamten TLBs für den L3-Cache umgehen

AMD can fix the bug by deactivating TLB's of the L3-cache.

Other sources say the bug can be fixed with a bios fix, so i assume this fix simply disables TLB's.

I wonder if the actual bioses for phenoms have this fix included.

[Katzenschleuder]

lol, what are you reading?

heise said amd can bypass the TLB bug, by disabeling the L3 cache and it would cost 10% or more of perforamnce if they disable it.

Theres nowhere mentioned that the bios disables anything.You are not correct too.
The article says that AMD can only work around the problem by disabling the TLB of the L3 cache.

[Hornet331]

You are not correct too.
The article says that AMD can only work around the problem by disabling the TLB of the L3 cache.


Alle bisher fertigen Phenoms haben noch einen Bug im Translation Lookaside Buffer (TLB) des L3-Caches. Er soll laut AMD nur unter sehr seltenen Lastfällen auftreten und entspricht den Fehlern im Barcelona. AMD kann den Fehler durch Deaktivieren des gesamten TLBs für den L3-Cache umgehen, allerdings kostet das laut eigenen Aussagen zehn und mehr Prozent Performance.

ok ich möcht jetzt nicht deutsch unterricht geben, aber wo liest du raus, dass es die einzige möglichkeit ist?

Da steht amd kann den l3 cache deaktivieren um den bug zu umgehen und das es 10% oder mehr perforamnce kosten kann.

Es steht weder da, dass es die eizige möglichkeit ist, noch das sie es sicher machen... also bitte.


Sorry for that huge amount german, but i had to point out my viewpoint. :p:

[justapost]

ok ich möcht jetzt nicht deutsch unterricht geben, aber wo liest du raus, dass es die einzige möglichkeit ist?

You are they are not writing that this is what the fix does.
They write the bug can be fixed by disabling TLB's and it costs ~10% performance.
fud wrote the possible bios fix costs ~10% performance.

Anyway they decided not to use the bios fix. ;)


Er soll laut AMD nur unter sehr seltenen Lastfällen auftreten und entspricht den Fehlern im Barcelona.
The bug occures under seldom occuring loads and is equalent to the bug in barcelona.

Now I wonder if current barcelonas have TLB's disabled via bios tweaks.

[villa1n]

So it is the fact that the nb and L3 are stuck at 2.0ghz and cannot change frequency.. if the only fix is to disable the cache... hrm, that being the case, if they fix it, maybe we will see some real clock speeds in the near future :D

[terrace215]

maybe we will see some real clock speeds in the near future :D

If by 'near future' you mean ~3 months from now, maybe.

They need some new masks, and they obviously only just discovered the problem, otherwise review sites would not have been given 9700s to play with, only to have them pulled back at the last second. Then one would hope they would sample and validate this new stepping for a month or so before releasing it, to avoid another fiasco.

p.s. The L3 is not disabled by the patch.

[justapost]

More info about TLB's can be found here (http://www.xbitlabs.com/articles/cpu/display/amd-k10_9.html).

[informal]

If by 'near future' you mean ~3 months from now, maybe.

They need some new masks, and they obviously only just discovered the problem, otherwise review sites would not have been given 9700s to play with, only to have them pulled back at the last second. Then one would hope they would sample and validate this new stepping for a month or so before releasing it, to avoid another fiasco.

p.s. The L3 is not disabled by the patch.

What is really strange is late discovery of TLB errata :shrug: .It is beyond me how they found out the errata 3 days prior the launch...
Let's hope they do it right with the next stepping.Also lets hope they clock the NB higher then the poor 2Ghz.

[JumpingJack]

What is really strange is late discovery of TLB errata :shrug: .It is beyond me how they found out the errata 3 days prior the launch...
Let's hope they do it right with the next stepping.Also lets hope they clock the NB higher then the poor 2Ghz.

I have had a few discussions on this topic, and one person provided a theory that was a reasonable explanation.

Do you remember the Opteron recall a year or two back, 3000 opty's had a problematic hot spot when run under load with a certain instruction over and over... it was a test escape.

If AMD is have thermal problems (this is the assumption, loosely supported) it could be that they have a hot spot causing device failure at a particular point. This would explain why the 2.4 and higher are affected and the lower speeds are not... it would also explain why some people are seeing relatively higher OC's, 2.8 GHz stable while others can't get past 2.5 GHz stable.

These type of problems are hard to identify because they are random and systematic. I.e. they may be on a cliff and it is unfortunate that they just now found it.

[informal]

I have had a few discussions on this topic, and one person provided a theory that was a reasonable explanation.

Do you remember the Opteron recall a year or two back, 3000 opty's had a problematic hot spot when run under load with a certain instruction over and over... it was a test escape.

If AMD is have thermal problems (this is the assumption, loosely supported) it could be that they have a hot spot causing device failure at a particular point. This would explain why the 2.4 and higher are affected and the lower speeds are not... it would also explain why some people are seeing relatively higher OC's, 2.8 GHz stable while others can't get past 2.5 GHz stable.

These type of problems are hard to identify because they are random and systematic. I.e. they may be on a cliff and it is unfortunate that they just now found it.
Yep,i sure do recall.You are on to something here,there is possibility of a hot spot somewhere (in L3?).

[Shintai]

Yep,i sure do recall.You are on to something here,there is possibility of a hot spot somewhere (in L3?).

Not L3 since it wont be that hot. It would be some logic. Maybe the bridge between the L3 and the L2s.

[Jimmer411]

I feel sorry for the early adopters. As much as I want a X4, or even dual X4 system I dont think things are going to smooth out till the spring.


Unless Nehelem is the reincarnation of the prescott, the next few years are going to be just as depressing as it is now when you try to ask AMD advice in any AMD forum on the internet...

[Hornet331]

Now someone should run a K10 with no heatsink on, shoot a heat shot - and get a heat map, so we can speculate over the problem based on what kind of logic it is affecting :)

wouldn't make much sense with the ihs on the cpu. :p:

but maybe someone wants to remove it.

[GAR]

AMD always has an excuse for everything..........its getting old already, either come out with a good product or face the facts.........stop making excuses......future drivers, future bios, future stepping, all false hopes to inspire sales.