Wow--New Black Edition Phenom Next Week-- Comments?

[ryman554]

I'm not praising AMD, I'm saying that CRAP happens. I say, it will improve, you say For all i know the chip tested by THG....

It's just sickening.

Phenom will get better, there's no question about it. It just won't reach parity with Core2 on a per-core basis -- excepting those memory hungry tasks. That's not FUD, that's a fact. All of the single threaded benchmarks point in that conclusion.

If I've missed something, please provide me with a tangible link that would make me change my mind.

I just wish I shared your optimism that AMD will make a miraculous improvement. What tangible evidence (not rumours from Fudzilla, not editorials from Dirk Meyer, just honest to goodness data) do you have that Phenom is fundamentally more capable than it has shown?

 

[epsilon84]

No, this very site had an ES (unlocked) that ran stable at 3GHz and several other places easily got to 3GHz. Only Anand got lower but I believe he had a retail chip (locked).

Can you please list the sites that 'easily got to 3GHz'? I don't recall seeing many, and I've read quite a few Phenom reviews.

 

[homerdog]

I pity the fool who responds to a post made by BaronMatrix.

 

[someguy7]

No way a black edition is coming out any time soon.

 

[gallag]

I to would like links to all the easy 3ghz k10 threads please

 

[3Ball]

I call bullshit.

+1

 

[3Ball]

But I'm sure that no matter what they do, your opinion won't change.

The sad thing is...is that neither will yours!

Best,

3Ball

 

[croc]

The sad thing is...is that neither will yours!

Best,

3Ball

The saddest part of it all is it looks very like AMD is in a bad way. If you take all of the other bad news articles, (WSJ, EDN, EETimes) things look bad for AMD. Hector, on the other hand, can take his golden parachute and jump off of the plane with megabucks already vested, so what does he care?

As to who would buy AMD, I do believe that would be limited to an American company. And then there is the issue of Intel cross-licensing agreements...

All in all, a very messy outlook.

Not a good look, all around.

 

[Reynod]

Thanks tamerelo for the link to the northbridge issues at extremesystems ... they seem to be doing some good work there with the bios.

 

[nruo]

if it is correct then yes, it is almost a fraud from AMD
AMD has alrady publicly annonced that the B2 stepping is defected, and in the analyst day said that new "fixed" chips (probably B3) will only in Q1 next year (and this is probably not January, otherwise they wouldn't hide the exact date )

so what can AMD sell next week?
the defected B2 parts, which work correctly only in 1.9-2.2 GHz.
but no one wants it as is, and AMD had to find a way to sell it (other wise these "hundreds of thousands" will have to be write off)
so they found a great solutution - lets sell it to the enthusiastics (i.e. suckers) - they don't gurantee that it works in higher frequencies (because thy know it doesn't) but hypig it as an good for overclocking - no one can sue them over the hype, or on the fact that OCed units are causing the system to hang

 

[elbert]

Not trying to shoot you down, but Fudzilla isn't the most reliable news source around. I would believe what TechReport said, which B3 is due in Q1 of 08.

True but the tech report is old by near 2 weeks so its just as unreliable at this point. As I stated before a limited B3 supply this could be. Being that Q1 08 is only 3 weeks away I would see little reason that some B3 could not be released now or shortly.

 

[nruo]

As I stated before a limited B3 supply this could be. Being that Q1 08 is only 3 weeks away I would see little reason that some B3 could not be released now or shortly.

http://www.x86watch.com/

But when pressed by an analyst about when exactly faster Barcelona parts (i.e., 2.5 GHz) would be available, there was a pause, then Rivas offered early Q2 in the "April-May timeframe."

 

[elbert]

http://www.x86watch.com/

But when pressed by an analyst about when exactly faster Barcelona parts (i.e., 2.5 GHz) would be available, there was a pause, then Rivas offered early Q2 in the "April-May timeframe."

And do you know the differance in Barcelona and phenom? We can not imply this to cover phenom as the question was barcelona and not phenom. This is a black edition phenom we are speaking of and barcelona being 2.5 is no big deal as this could be about the 2GHz server part. No distinction was made by rivas.

 

[harmattan]

Not trying to shoot you down, but Fudzilla isn't the most reliable news source around. I would believe what TechReport said, which B3 is due in Q1 of 08.

B3 may very well be "released" next week. However, for any volume of the chips to get to reseller, and then for those resellers to clear out their B2 stock first will take longer. I'd say mid to late Q1 is a good estimate.

 

[AnGeLuS69]

It should be great since nearly every sample I saw was able to get to 3GHz by turning up the multiplier. And with AMD OverDrive, great may be an understatement.

I'll try this more diplomatically...

I believe you are not being truthful/factual. Now I am not insulting you in any way shape or form, I just believe you should perhaps look towards Macci as an example here.

Taking a Phenom riddled with a TLB bug to 3GHz is impossible unless you patch that said Phenom (~20% performance hit). The amount of heat generated would require more extravagant then air cooling (as proven by Macci).

Macci had an Engineering Sample provided to him directly by AMD (he works for AMD's Chipset division). This ES variant does not suffer from the TLB bug. As such I truly believe you're misleading potential buyers. I am not accusing you of doing so on purpose just stating a fact.

 

[homerdog]

I'll try this more diplomatically...

You can try...

Also, how is it that his ES didn't have the TLB bug? If AMD has bug free chips, why aren't they releasing them?

 

[TechnologyCoordinator]

If this stuff would clock to 3.0ghz you'd see high clocked Phenoms on the market being sold for more money.

This is a good direction for AMD, but enthusiast overclockers may favor faster and more OCable Intel processors.

 

[AnGeLuS69]

You can try...

Also, how is it that his ES didn't have the TLB bug? If AMD has bug free chips, why aren't they releasing them?

Engineering Samples often have issues with them. Probably other issues and errata's. Most of the Phenom ES chips don't suffer at all from a TLB bug.

As for being more specific then that, I cannot be as I do not work for AMD. Perhaps asking someone who does wuld give you the answers you seek.

 

[ryman554]

Engineering Samples often have issues with them. Probably other issues and errata's. Most of the Phenom ES chips don't suffer at all from a TLB bug.

As for being more specific then that, I cannot be as I do not work for AMD. Perhaps asking someone who does wuld give you the answers you seek.

Unlikely.

Engineering samples have identical designs (at this stage, at least) as production chips. They come from the same physical photomasks. So I can guarantee you that the Barcy ES in production today have exactly the same erratas and issues as what is going to be sold -- B2 OR B3.

Now, that being said, there may be *process* reasons why ES will differ from the final chip (poly gate CDs, implant times, and the like) which *can* give rise to different performance metrics (and potentially stability issues), but you're not going to see any more or less design bugs.

In early production/design work you'll see ES with a ton of bugs and stuff in them -- (ie Barcy A0), but in principle, each stepping is a "release candidate" (probablility however small). If they don't plan on selling it, it's called an ES. If they do, it's called production. But they are made (more or less) the same as the process is maturing.

 

[AnGeLuS69]

Unlikely.

Engineering samples have identical designs (at this stage, at least) as production chips. They come from the same physical photomasks. So I can guarantee you that the Barcy ES in production today have exactly the same erratas and issues as what is going to be sold -- B2 OR B3.

Now, that being said, there may be *process* reasons why ES will differ from the final chip (poly gate CDs, implant times, and the like) which *can* give rise to different performance metrics (and potentially stability issues), but you're not going to see any more or less design bugs.

There's a reason why they're called B2 and B3.

Much like comparing an E4 to an E6 Athlon64 X2, they've got differing errata's and bugs. Some are present throughout the chips life, others are fixed as the newer steppings are released.

The TLB issue is not present with most of the ES variants out there. Macci's chip does not suffer from the TLB problem.

 

[ryman554]

There's a reason why they're called B2 and B3.

Much like comparing an E4 to an E6 Athlon64 X2, they've got differing errata's and bugs. Some are present throughout the chips life, others are fixed as the newer steppings are released.

The TLB issue is not present with most of the ES variants out there. Macci's chip does not suffer from the TLB problem.

Then the ES samples out there are B3 steppings -- as I was trying to say, but probably self-obfuscated. =) What they are going to (or are planning to) sell. If *these* have additional errata (as you have insinuated in your previous post), then AMD is in a world of hurt.

Most likely, these ES are the pilot line (or close to it) wafers -- and they are working through potential process issues to enable the ramp. That's good news for AMD. And the reason they can't sell them -- they still need to convert the line over. They won't commit significant amounts of production until they are assured they have a working part. And that will take ~10-12 weeks from the time the lead lot gets through validation, depending on the layers touched for this stepping.

 

[AnGeLuS69]

Then the ES samples out there are B3 steppings -- as I was trying to say, but probably self-obfuscated. =) What they are going to (or are planning to) sell. If *these* have additional errata (as you have insinuated in your previous post), then AMD is in a world of hurt.

Most likely, these ES are the pilot line (or close to it) wafers -- and they are working through potential process issues to enable the ramp. That's good news for AMD. And the reason they can't sell them -- they still need to convert the line over. They won't commit significant amounts of production until they are assured they have a working part. And that will take ~10-12 weeks from the time the lead lot gets through validation, depending on the layers touched for this stepping.

Some new Errata's are usually created though with a move to a newer stepping. The TLB issue seems to be just that. At least it's what we're assuming seeing as the ES variants handle Virtualization without issue.

 

[justinmcg67]

You know I read all of this AMD stuff, and I can't help but ask myself this question:

BaronMatrix, do you even own a Phenom processor?

 

[dragonsprayer]

wow paint the box black and that gets you excited?

if the cpu was black you could not see it under the hsf - u know?

maybe a glass hsf and black steel package - then it would be a black edition

 

[yomamafor1]

You know I read all of this AMD stuff, and I can't help but ask myself this question:

BaronMatrix, do you even own a Phenom processor?

No... but he has a Skulltrail