• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

AMD shows off BD architecture at Hotchips

nyker96

Diamond Member
bulldozer-die-diagram.jpg


Article Link: http://www.legitreviews.com/news/11315/
 
So does this tell us anything valuable or will it only serve to fuel more speculation? Serious question, as I don't know how big a deal the die map is.



Supposedly we have just under a month to wait.

315mm^2, it's on S/A and in another thread here.

Wait the L3 cahces aren't shareabled???

Don't let the physical layout confuse you, the electrical topology is what matters. The L3$ is shared. Its just not all located in one physically contiguous sram array.

Here's the original uncropped image of the slide:
thumbnail.php
 
Last edited:
I'm still not sure what they are getting on about with the whole module bit and how it is materially different to a core.
 
I'm still not sure what they are getting on about with the whole module bit and how it is materially different to a core.
That by sharing between what would have been two independent cores in the right ways, total complexity/xtors can be reduced, and overall performance improved, since so much of performance as of late has been due to power limitations. The integer part of each module follows research done in this area, which has not yet been proven in a real performance CPU, and is thus rather ambitious; yet makes sense given that AMD has so few R&D and fab resources compared to Intel. Due to the sharing of the front-end, and SMT for the FPU, cores cannot be separated within a module.

From a consumer POV, it's just another 2-, 4-, 6-, or 8-core CPU, and we can only cross our fingers and hope that it won't be Phenom all over again.
 
Zambezi should be quite good desktop CPU. It will address some shortcomings of previous AMD cores and bring some serious clocking potential to OCing community(especially subzero). All this for 300$ .What is not to like? 🙂
 
Zambezi should be quite good desktop CPU. It will address some shortcomings of previous AMD cores and bring some serious clocking potential to OCing community(especially subzero). All this for 300$ .What is not to like? 🙂

That it may never be released? 😛
 
Zambezi should be quite good desktop CPU. It will address some shortcomings of previous AMD cores and bring some serious clocking potential to OCing community(especially subzero). All this for 300$ .What is not to like? 🙂
Will it break current CPU world records (in clock speed and performance when overclocked, which is the motivation for putting these chips under LN2) 😛

That it may never be released? 😛
Or will it end up like DNF? 😀
 
Boring! I want actual benchmarks and CPUs AMD!
There are "leaked" benchmarks of ES (Engineering Samples) which shows awful performance (some rather believeable, still take them with a grain of salt) possibly due to bugs, and there are also "leaked" benchmarks of ES which shows fantastic performance (too good to be true, take them with lots of salt) which later were mostly exposed as frauds (such as my post here, besides the Obrovsky ones). Either way, you decide yourself but I suggest wait for the official release before judgement. 😛
 
Seriously, all of this has bored me to tears. I predict another Phenom launch. Going to wait for Ivy Bridge.

I agree there are parallels that can be drawn between the 90nm Athlon X2, 65nm Athlon X2, 65nm Phenom and the 45nm Phenom II X4, 32nm Phenom II X4 (llano), 32nm Bulldozer.

We didn't know to rationalize it as such at the time, but when those initial 65nm Athlon X2's came out and they could not even meet the clockspeeds of their 90nm predecessors it turned out to be a harbinger of the health of 65nm and what that meant for 65nm phenom clocks.

Now look at Llano clockspeeds versus power...what should we be expecting from bulldozer? Miracles that defy device physics? (maybe...prescott sucked at 90nm but Dothan did not)
 
If someone can give me a good reason why AMD shouldn't release a "leaked but official looking" benchmark on thier flagship CPU in the next 7 days, I will send them 500 internets.

The difference between their GPU side, which has been successful, and the CPU side, which hasnt, is that they will allow NDA-busting leaks, because they know it will stop system-builders in their tracks.
 
If someone can give me a good reason why AMD shouldn't release a "leaked but official looking" benchmark on thier flagship CPU in the next 7 days, I will send them 500 internets.

The difference between their GPU side, which has been successful, and the CPU side, which hasnt, is that they will allow NDA-busting leaks, because they know it will stop system-builders in their tracks.
The only reason I could think of is that they would be tipping their hand to Intel. Intel has a very specific plan for AMD to keep them in the bottom 20% of the market; should they ever do too well, Intel will release new chips at new prices to push AMD back. It's in AMD's advantage to not let Intel see their performance until the last moment to maximize their own benefit of their launch window.

But all the same I can't imagine Intel doesn't have a production Bulldozer. In the long run it would be cheap to pay off an AMD partner to cough up their chip at least for a period of time.

So yeah, you're more or less right. AMD's "tell" is that they don't like to talk about their performance when it's poor. The last time they clammed up like this it was for the Phenom (I), and we all know how that went.
 
The only reason I could think of is that they would be tipping their hand to Intel. Intel has a very specific plan for AMD to keep them in the bottom 20% of the market; should they ever do too well, Intel will release new chips at new prices to push AMD back. It's in AMD's advantage to not let Intel see their performance until the last moment to maximize their own benefit of their launch window.

But all the same I can't imagine Intel doesn't have a production Bulldozer. In the long run it would be cheap to pay off an AMD partner to cough up their chip at least for a period of time.

So yeah, you're more or less right. AMD's "tell" is that they don't like to talk about their performance when it's poor. The last time they clammed up like this it was for the Phenom (I), and we all know how that went.

as did they with K8, same with K7, same with radeon 68xx and 69xx... so no you have absolutely no ground to say they only did this with phenom1... to be precize it isn't even close to it! Phenom was boasted over the competition in the server space. BD is only boasted over their own current cpu line!
 
If someone can give me a good reason why AMD shouldn't release a "leaked but official looking" benchmark on thier flagship CPU in the next 7 days, I will send them 500 internets.

The difference between their GPU side, which has been successful, and the CPU side, which hasnt, is that they will allow NDA-busting leaks, because they know it will stop system-builders in their tracks.

Osbourne effect

To keep up relationships with 3rd parties, because Osbourne would hit them too.

To maximize hype when the product is actually available. People have a short memory, if they leaked a benchmark that shows, lets say FX6130 beating a 2500k, that would make a much bigger impact on sales when the product is already available.

That will be 1500 internets please 🙂
 
as did they with K8, same with K7, same with radeon 68xx and 69xx... so no you have absolutely no ground to say they only did this with phenom1... to be precize it isn't even close to it! Phenom was boasted over the competition in the server space. BD is only boasted over their own current cpu line!
K8 and K7 numbers were available quite early. Even Phenom II and Llano ended up in Anand's hands for a preview. So I stand by my statement: AMD is being uncharacteristically mum about Bulldozer.
 
The only reason I could think of is that they would be tipping their hand to Intel. Intel has a very specific plan for AMD to keep them in the bottom 20% of the market; should they ever do too well, Intel will release new chips at new prices to push AMD back. It's in AMD's advantage to not let Intel see their performance until the last moment to maximize their own benefit of their launch window.

But all the same I can't imagine Intel doesn't have a production Bulldozer. In the long run it would be cheap to pay off an AMD partner to cough up their chip at least for a period of time.

Yep i reckon Intel would've had their hands on just about every relevant stepping of BD, im sure they know exactly how it's going to perform. Hell they probably have it performing better than AMD are able to..
 
K8 and K7 numbers were available quite early. Even Phenom II and Llano ended up in Anand's hands for a preview. So I stand by my statement: AMD is being uncharacteristically mum about Bulldozer.

K7 were crippled and K8 were to.. The other numbers we got from K8 were also from opteron which was released before the desktop. (and was a little preview for what was to come)

Llano preview was very close to the actualy release... same can happen with BD. There is no launch day for BD at this point in time.
Bobcat preview was the same.
And for both of those previews almost nothing was known about eventual specs or performance!
 
Last edited:
I think that people forget how secretive AMD is, compared to Intel. With Intel, they're usually like "check out our next-gen CPU you'll be able to buy in 6 ~ 12 months". AMD you can barely get them to admit they're working on another CPU...


What really grinds MY gears is not the secrecy of performance, but the secrecy of the launch date. Sometimes I get the feeling it isn't secrecy, they just don't know it... Scary.
 
AMD's been "awfully quiet" on Bulldozer, even though its one month or less away from "September launch". Heck they haven't even officially announce an actual launch date (only from rumors, "leaks" and "slip ups"). Look at Phenom II launch, when AMD had something they organize (LnHe) overclock parties early on. I'm not seeing the same scenario here being this close to "launch". I hope there isn't another delay again (to October?). 😛
 
Back
Top