Zambezi ES performance weirdness

[DarthShader]

One problem with the gimping theory: JF said somewhere (amdzone?) that they don't cripple ES chips on purpose. So it's either genuine problems, or some other means of crippling, via BIOS for example?

[TESKATLIPOKA]

Think about it, if they did something like that would they admit it before the end of NDA?

P.S. its still possible JF is telling the truth, we will see when BD is released

[DarthShader]

JF didn't have to say anything. But he did.

[liberato87]

nobody can ask these guys with the ES to see if they work on a crosshair IV?
it will be intersting for me

[Daveburt714]

One problem with the gimping theory: JF said somewhere (amdzone?) that they don't cripple ES chips on purpose. So it's either genuine problems, or some other means of crippling, via BIOS for example?

I saw his post on this subject too, but I can't find it now...
The way I read it was "We produce ES chips for the best yeild / not the best performance", of course I'm para-phrasing
but that's what I got from his comment.

[2senile]

JF-AMD said they didnt cripple the ES sample?so it must be a bios code problem right?

this is from AMDzone

Bios, software glitch .... I'd certainly expect a little "update" from M$ when the chip is finally launched to add to the AMD specific stuff already in the OS. AMD ESs have always been crap which is one of the reasons why I've not been toooooo concerned ..... thus far.

[informal]

"Gimping" is just a term. In reality we have at least 2 different levels of performance between B0 and B1 (check my latest blog). This proves that ES are not performing the same and thus that these are not final performance numbers.

[El Gappo]

"Gimping" is just a term. In reality we have at least 2 different levels of performance between B0 and B1 (check my latest blog). This proves that ES are not performing the same and thus that these are not final performance numbers.

Where are these B1 numbers coming from?

[informal]

OBR,but he has been banned here long time ago. He has a blog and he posted it there.
As a matter of fact,we have no idea if it's B1,that's my personal guess.And it makes zero difference if it's still B0,since we would then have 2 B0 samples performing massively different. Note that his B1/B0 could indeed be final performance level of Zambezi,but we don't know it yet.

[El Gappo]

obr....Duly noted and taken with an appropriate amount of salt.

[demonkevy666]

OBR,but he has been banned here long time ago. He has a blog and he posted it there.
As a matter of fact,we have no idea if it's B1,that's my personal guess.And it makes zero difference if it's still B0,since we would then have 2 B0 samples performing massively different. Note that his B1/B0 could indeed be final performance level of Zambezi,but we don't know it yet.

he's also making fun of you on his blog too.

But your not claiming to know How fast it really is, because it's just speculation, until the cpu is out.

[informal]

he's also making fun of you on his blog too.

But your not claiming to know How fast it really is, because it's just speculation, until the cpu is out.

Yeah I know,he is just a jackass,that's why he is banned on many forums. And that's why he violates NDA before anyone else(especially if it is AMD hardware in question).He can't help it,that's his nature I guess.

[dinos22]

Superpi is absolutely useless, it's a relic from nearly 30 years ago when we used to install our own co-processors.
AFAIK AMD has deprecated x87 in BD and it will be emulated in the fp pipe. Don't be surprised if 1m times are out over a minute.

Still think it's a great performance indicator?

that's what Intel guys were saying during AMD K8 era too

[informal]

SuperPI will be handled via 2xFMAC.I guess it should be a good step up from Phenom II .

[Daveburt714]

that's what Intel guys were saying during AMD K8 era too

Believe it or not, SuperPi was one of the few benches where Pentium4 could actually beat K8 (drove me crazy too (hehe))..
So yeah, SPi has never been a good judge of overall performance. Especially when comparing AMD and Intel.

I'm with informal on this one though, I get the feeling BD may perform much better in SPi.
It still won't really mean much though....

[JF-AMD]

Based on this conversation I have a feeling nobody here is in the semiconductor business, so let me try to explain it to you.

Engineering samples are designed to validate the design and for partners to validate their systems. They are not meant for benchmarking.

There are multiple steppings of samples, comparing one stepping to another tells you nothing unless you know WHY the stepping changed.

When we make engineering samples, we target the best yield, not the highest performance. Then we have more chips to work with. Typically you get a couple of the highest clock speed anyway, those typically stay in our labs, so it is unlikley that you see those out in the wild.

We don't even do any real performance analysis until we get the final silicon. Prior to that, we are totally focused on design validation.

Performance is determined based on the following:

1. The processor
2. The processor microcode
3. The BIOS
4. The Operating system
5. The drivers
6. The compiler code

So, unless you are looking at final silicon and you have all of the other pieces in place, making a statement about the performance of ES chips is completely pointless.

Which is why we just don't comment on those numbers. They probably aren't real. And if they are, they are not representative of what people will see when they get the real products in their hands.

Imagine making a cake. You start with a pile ingredients. You put it all in a mixing bowl. You stir it up. You put it in the oven. You let it cook. You take it out. You put frosting on it.

You are all looking at bowlful of batter and arguing about whether or not it is appetizing.

[[XC] Oj101]

Based on this conversation I have a feeling nobody here is in the semiconductor business, so let me try to explain it to you.

An entire thread just got pwnt by JF-AMD

Thanks for the explanation John

[Olivon]

Happy Holiday JF ...

[DarthShader]

When we make engineering samples, we target the best yield, not the highest performance.

And here lies the answer to some questions, like the lower clock than expected rumour.

Happy Holiday JF ...

He is visiting the forums in his free time, so take the trolling elsewhere.

[El Gappo]

Imagine making a cake. You start with a pile ingredients. You put it all in a mixing bowl. You stir it up. You put it in the oven. You let it cook. You take it out. You put frosting on it.

You are all looking at bowlful of batter and arguing about whether or not it is appetizing.

Bad example, the batter is ALWAYS better than the cake nomnomnom

[[XC] Oj101]

^^ I agree /drools

[informal]

OK ,all of the youtube videos of gimped B0 have been removed.A bit late but better ever than never.

[SkullCracka]

well new LIanos are looking real good. So im sure the final product will be awsome sauce omgwtfbbq. <<<< or as AMD calls it Epic and Brutal.


http://forums.tweaktown.com/overcloc...p6160-igp.html

[SEA]

Btw, it is Llano L-L-A-N-O, not "LIano". Also it is not bulldozer, so it is not appropriate for this topic..
But yes, nevertheless it looks good!

[RussC]

Based on this conversation I have a feeling nobody here is in the semiconductor business, so let me try to explain it to you.

We don't even do any real performance analysis until we get the final silicon. Prior to that, we are totally focused on design validation.

Performance is determined based on the following:

1. The processor
2. The processor microcode
3. The BIOS
4. The Operating system
5. The drivers
6. The compiler code

Actually, I do work in the industry, and I believe one of our etchers is in one of your fabs.

With that said, You don't mean that the engineers that designed the processor dont have some idea of performance before final silicon? Um, that makes no sense to me! So AMD has no idea of where the BD cores stack up against competitors before final silicon, I have a hard time believing that.

RussC