AMD Bulldozer Eng. Sample leaked, benched

[Dent1]

Really, i am probably older than you. I am basing my statements on the fact the launch has been postponed 3 times, and that all the leaked ES BD chips are just a FAIL really.

You have got to be kidding, right?

No I'm not kidding. Answer the question please


Earlier you was being negative about its hypothetical power consumption. Now you are being negative about it being postponed. Get your story straight, is your hostility due to the power consumption or delays?


ES chips are supposed to be fail. What’s your point?

 

[heky]

No I'm not kidding. Answer the question please


Earlier you was being negative about its hypothetical power consumption. Now you are being negative about it being postponed. Get your story straight, is your hostility due to the power consumption or delays?


ES chips are supposed to be fail. What’s your point?

1.)both

2.)Since i can remember(Intel 286 33mhz), ES were specially binned chips, shipped all over the worldf for reviews and testing. ES were also always the best overclockers. Just take a look at reviews over the years. Almost all reputable sites used ES. And you state ES are supposed to be fail.

 

[Tatty_Two]

On a serious (and un-biased note), I always thought the point of releasing some ES chips was to display at the very least some of the potential of the forthcoming product, not to highlight it's weaknesses which surely would actually only serve a negative purpose, hence in the past... often ES chips actually performed better than their first revision stepping releases....... I cannot beleive anyone would actually release any form of ES or "prototype" for anything that could even remotely be found to have any major shortcomings, it's a bit like the Britiish Airways, Air France and the American Airlines (not sure if I have the right US carrier there) partnership releasing a Concorde prototype in 1970 that couldn't fly surely?

Just my 2 pennies worth.

 

[cadaveca]

ES samples are used by OEMs to validate thier supporting products. As such, they only need to have very basic functionality, and to simullate TDP/ACP.

Performance is not a consideration, as the testing period is like 6 months....6months after samples go out to OEMs, it's definitely posible a silicon revision can be done, which will vastly change performance characteristics.

AMD, via JF-AMD, has already stated that ES smaples are NOT indicitive of final performance of retail chips, due to tweaks that were made so that the highest number of samples were functional, rather that a retail scenario, where chips are binned for speed/performance.

With this information in hand, I reserve judgement until I see results with a non-ES sample.

 

[repman244]

But heky you still haven't answered my question

What are you basing your assumptions on, that the power consumption of BD will be out of this world.



About ES's: OBR was saying that BIOS was the culprit for low scores (on those first benches that he showed), but then again I really can't take that guy seriously.


But did we have any other benchmarks (like cinebench, wprime, something like that) apart from useless SuperPI and supposed gaming benchmarks? (sorry I may have missed some)

 

[cadaveca]

SuperPI is good to judge ram performance, but not CPU performance, really.

the others, while valid benchamrks, still aren't indictive of "real-world" performance, as they were all written with different tech in mind.

 

[Dent1]

1.)both

2.)Since i can remember(Intel 286 33mhz), ES were specially binned chips, shipped all over the worldf for reviews and testing. ES were also always the best overclockers. Just take a look at reviews over the years. Almost all reputable sites used ES. And you state ES are supposed to be fail.

From what I can remember the high performing ES which are a near final revision are given out to journalists of respected review websites and magazines to create publicity. In turn a full review is conducted on an array of benchmarks and a detailed analysis of the findings is writen up next to some pretty graphs and charts and a conclusion is made.

In this situation, I see no credible review from a reputable website or magazine, I see no detailed analysis, no pretty graphs and charts, just low quality images which you guys call leaks, a credible reviewer don't have leaks or act shady because they have permission to review the product.

Regardless of previous engineer samples, AMD are not obliged to make ES perform better or similar than the final retail product. It's an ES, there is no rules on how it *should* perform. It's a broken, unflished peice of kit, treat is as such.

 

[repman244]

SuperPI is good to judge ram performance, but not CPU performance, really.

the others, while valid benchamrks, still aren't indictive of "real-world" performance, as they were all written with different tech in mind.

Just one question and a bit off topic, regarding SuperPI (maybe you don't know, but it's worth asking if you do), I always noticed a huge difference in performance between Intel and AMD, no other benchmark AFAIK doesn't have such huge delta in performance.
Is it the way it was coded, the way it uses instructions?

 

[cadaveca]

No, AMD ram performance is severely lacking in comparison to Intel(I've mentioned many times that ram performance is key for me with Bulldozer). Even in comparison from Intel platforms, say 1156, and 1155, 1155 is vastly superior, and this shows not just in SuperPI, but in ALL benchmarks that are ram-intensive.


So, no, it's not just the way apps are coded. You'll see the same performance deficits in many of the benches in my reviews, like this one:

Note the 20FPS difference, but look at this bench:

Not so ram intensive, and things are pretty close...even though Intel can use 8 threads vs. 6 on AMD...

WinRAR, however, is more ram intensive, and shows a big performance difference:

So, in the end, your assumption that no other benchmark shows these differences, is not correct, at all.

 

[TheMailMan78]

No, AMD ram performance is severely lacking in comparison to Intel(I've mentioned many times that ram performance is key for me with Bulldozer). Even in comparison from Intel platforms, say 1156, and 1155, 1155 is vastly superior, and this shows not just in SuperPI, but in ALL benchmarks that are ram-intensive.


So, no, it's not just the way apps are coded. You'll see the same performance deficits in many of the benches in my reviews, like this one:

http://tpucdn.com/reviews/Biostar/TZ68A_PLUS/images/f12010.gif


Note the 20FPS difference, but look at this bench:

http://tpucdn.com/reviews/Biostar/TZ68A_PLUS/images/wprime.gif


Not so ram intensive, and things are pretty close...even though Intel can use 8 threads vs. 6 on AMD...

G-d I hope AMD reworked the damn MC on BD. If not well........next rig will be Intel.

 

[heky]

What are you basing your assumptions on, that the power consumption of BD will be out of this world.

I am just speculating, based on the relatively high voltage in the leaked slides(1.4v+)

 

[repman244]

Thank you for that cadaveca (and thank you for warning me that SuperPi isn't a lonely case).

Well Llano is capable of doing 2400MHz RAM 8-11-8-28, would that suggest to an improved IMC or does it make only a small effect on RAM performance?

 

[TheMailMan78]

Thank you for that cadaveca (and thank you for warning me that SuperPi isn't a lonely case).

Well Llano is capable of doing 2400MHz RAM 8-11-8-28, would that suggest to an improved IMC or does it make only a small effect on RAM performance?

Its depends on if BD uses the same IMC. Which NO ONE knows but AMD.

 

[sneekypeet]

This round of cleaning is free, and on BP's request! IF it was up to me points would have been doled out, and this thread would remain closed. You can thank cadaveca's talking me into opening it via MSN for your ability to continue with all this FUD!

 

[cadaveca]

Well Llano is capable of doing 2400MHz RAM 8-11-8-28, would that suggest to an improved IMC or does it make only a small effect on RAM performance?

Just because frequency scales, doesn't mean that that frequency is utilized. You see that lots on AMD chips...little bandwidth gains from frequency, but latency improves. You need to up NB speed to see bandwidth gains, and a near 50% boost from stock NB does not give a 50% boost in performance.

Good indicator something is up. What that problem is, or if it is even really a true problem, must be answered from AMD. I just work here.

 

[Damn_Smooth]

Seronx do you think my motherboard will run Bulldozer ok?

I'm sorry Mailman, I'll take this one for Seronx. I really don't think that one of the best 990fx boards (Probably the best.) on the planet is going to have a shot at running Bulldozer.

You are going to need MOAH POWAH!!!

And from what I've learned in this thread, you better watch those vrm's too.

 

[TheMailMan78]

Just because frequency scales, doesn't mean that that frequency is utilized. You see that lots on AMD chips...little bandwidth gains from frequency, but latency improves. You need to up NB speed to see bandwidth gains, and a near 50% boost from stock NB does not give a 50% boost in performance.

Good indicator something is up. What that problem is, or if it is even really a true problem, must be answered from AMD. I just work here.

Constantly fighting my OC has taught me that timings are everything for AMD RAM.....that and Erocker calling me an idiot. Frequency makes a difference but not like timings. For example I am running cas9 now at a higher frequency then when I was running cas7 in DDR2........The DDR2 was faster due to the latency in everything.

So unless BD has changed the IMC its going to suck IMO.

I'm sorry Mailman, I'll take this one for Seronx. I really don't think that one of the best 990fx boards (Probably the best.) on the planet is going to have a shot at running Bulldozer.

You are going to need MOAH POWAH!!!

And from what I've learned in this thread, you better watch those vrm's too.

OH NOES!

 

[techtard]

I hope they fix the underferforming USB too. Not a dealbreaker, but it would be nice.
And maybe a more robust onboard RAID controller.

 

[TheoneandonlyMrK]

im putting the delays down to production capacity verses strategic mass manu of the RIGHT chips at this time E and A series plus the upcoming 7800+ gfx cards, if i were them i wouldnt be as concerned with the few as i would be with the many, simples, <and also as said before.


like this thread, lookin forward to BD and im willing to wait for deffinate benches before upgradeing again, so i appreciate the clean up

and imho AMD will have done as Cadeveca says, given out ES chips to oems that were mearly rock solid examples not fastest just to do ATE testing with etc, and its one or more of them getting benched, and AMD will as any company does pass the best chips to reviewers who will then give valid reviews and conclusions hence why no sites reviewed it yet, that and the NDA no doubt in place.

+1 on the better RAID and memory stabillity/compatibility

 

[yogurt_21]

On a serious (and un-biased note), I always thought the point of releasing some ES chips was to display at the very least some of the potential of the forthcoming product, not to highlight it's weaknesses which surely would actually only serve a negative purpose, hence in the past... often ES chips actually performed better than their first revision stepping releases....... I cannot beleive anyone would actually release any form of ES or "prototype" for anything that could even remotely be found to have any major shortcomings, it's a bit like the Britiish Airways, Air France and the American Airlines (not sure if I have the right US carrier there) partnership releasing a Concorde prototype in 1970 that couldn't fly surely?

Just my 2 pennies worth.

the only caveat to that I would say is that yours assumes the ES delivers upon consumer expectations. So while engineering expectations may have been met, it may not be up to what consumers demand. Also it could have taken them longer than anticipated to meet the engineering expectation thus putting them behind consumer demands.

in that case the ES would be put out to get a feel for the market to determine if the chip simply needs to be more competitively priced, needs to be scrapped because it sucks ass, or needs to have several things re-worked before final release.

most on here are assuming the latter condition is true.

 

[repman244]

Found some interesting info: http://vr-zone.com/articles/sandy-bridge-e-delayed-until-january-2012/12816.html#ixzz1QpvOUZ9J

I know, I know, not very reliable, and you might also ask what am I trying to do with the link in an AMD thread.
I'm not saying this to hate on Intel or anything but, what if (as a joke) Intel knows how final BD will perform and realized they need more time to tweak SB-E?
This is just my random BS-ing around

@TheMailMan78: So is it better to have the NB at let's say 2600MHz, RAM at 1333MHz with tight timings than having NB at stock (2000MHz) and RAM at 1600 or 1800 (but with loose timings)?

I really need to take a day or two to tweak a bit around RAM and NB, as I have seen that NB can give quite a boost.

 

[renq]

I'm not saying this to hate on Intel or anything but, what if (as a joke) Intel knows how final BD will perform and realized they need more time to tweak SB-E?

Or perhaps just the opposite- Intel realised there's no rush

 

[Velvet Wafer]

On a serious (and un-biased note), I always thought the point of releasing some ES chips was to display at the very least some of the potential of the forthcoming product, not to highlight it's weaknesses which surely would actually only serve a negative purpose, hence in the past... often ES chips actually performed better than their first revision stepping releases....... I cannot beleive anyone would actually release any form of ES or "prototype" for anything that could even remotely be found to have any major shortcomings, it's a bit like the Britiish Airways, Air France and the American Airlines (not sure if I have the right US carrier there) partnership releasing a Concorde prototype in 1970 that couldn't fly surely?

Just my 2 pennies worth.

Did you ever heard of the "Starfighter Affair"?
Many German airmen died, due to a severely flawed design, which even was accepted by military standards... it can happen

 

[seronx]

New leak

[yt]VHI675P-Xc8[/yt]

Idle Clock #0 - 1.4GHz
Stock Clock #0 - 3.2GHz
Turbo Core #0 - 4.2GHz

This is another leak by OBR so be wary

Nothing new for me

 

[Pestilence]

New leak

[yt]VHI675P-Xc8[/yt]

Idle Clock #0 - 1.4GHz
Stock Clock #0 - 3.2GHz
Turbo Core #0 - 4.2GHz

BENCHMARKS!!!!!!!!!!!!!!! God damn leaks