Sign in with
Sign up | Sign in
Your question

AMD Bulldozer details on August 24th?

Last response: in CPUs
Share
a b à CPUs
August 5, 2010 8:35:21 AM

One can only hope...
m
0
l
a c 85 à CPUs
a b À AMD
August 5, 2010 2:12:53 PM

The thing I want to know is whether its 8 full/real cores, or 4 cores with 8 interger cores and 4 FP cores. Exciting times coming up folks, no matter which camp you sit at.
m
0
l
Related resources
a b à CPUs
August 5, 2010 3:42:20 PM

4745454b said:
The thing I want to know is whether its 8 full/real cores, or 4 cores with 8 interger cores and 4 FP cores. Exciting times coming up folks, no matter which camp you sit at.


what it sounds like is that there are 4 modules, and each module has 2 integer cores and 2 128 bit FP cores that can combine to become a 256 bit FP core

from wikipedia (don't know it is right or not):
Quote:

* Two tightly coupled "classic" x86 out-of-order processing engines which AMD internally named module
(Single-Module ==> Dual-Core, Dual-Module ==> Quad-Core, Quad-Module ==> Octa-Core etc...)
* 8MB-16MB L3 cache shared between all modules inside same processor (on same silicon die)
* Dual channel DDR3 integrated memory controler (support for pc3-12800 (1600MHz))
* Higher Memory Level Parallelism
* Continuous Multithreading (CMT) Technology
Bulldozer module consists of
** 1024kB L2 cache inside each module (shared between module cores)
** Two independent integer cores
each consist of 2ALU and 2AGU and capable of 4 independent arithmetic or memory operations per clock
** One 256-bit wide floating point unit
with fused multiply-add (FMA) capability and flexible configuration to work with up to four 64-bit wide FP data or one 128-bit wide SSE FP data per clock
m
0
l
a c 85 à CPUs
a b À AMD
August 5, 2010 4:39:36 PM

Not sure. Last rumor I saw said each module should have 2 int cores and one FP core. For an "8 core" CPU it will have 4 modules, 8 int cores and only 4FP cores. I'm wondering about this because this should mean (depending on the FP improvments) that the FP power might not be that much better then the 955/965 because they would both have 4 FP cores. I'm assuming that this means there will be some times where BD isn't much faster then what AMD has on the market right now.
m
0
l
a b à CPUs
August 5, 2010 6:46:58 PM

Hmm, this blog is by John Fruehe, a Director of AMD Marketing (and no, that's not a non-sequitur :p ). He posts a lot over on AMDZone, which should tell you something - along the lines of "it's pretty safe preaching to the choir" esp. when you have a mad dog of a mod banning anybody who questions anything concerning AMD.

Previously he was touting Magny Cours as putting paid to Intel's server business. Unfortunately, both AMD's and Intel's last quarterly earnings reports show Intel actually increased its server marketshare, despite (or maybe because of) MC being available the entire 3 months and competing against both Nehalem EP and Westmere Xeons.

So (1) JF is a paid 'shill' by definition, and (2) his track record is rather abysmal.

Hopefully Global Floundries will get their gate-first HKMG sorted out and yields up on 32nm, since Bulldozer will be released on that node. However GF's problems have already caused AMD to push Llano back, in favor of the netbook fusion CPUs which are being made on TSMC's bulk 40nm process.
m
0
l
a b à CPUs
August 5, 2010 6:53:41 PM

Oh great, the day that bulldozer gets unveiled, is the same day I go back to school. woohoo?!?
m
0
l
a b à CPUs
August 5, 2010 7:20:45 PM

^ You make no sense dude...
m
0
l
a c 85 à CPUs
a b À AMD
August 5, 2010 10:18:48 PM

Quote:
AMD called it DX11 product line the Visual Proccessing Unit, kind of different from the Nvidia Geometry Proccessing Unit! It seems they have a patent that puts gpu on the cpu as a co proccessor to enhance graphics and cad functions to speed up the gaming and workstation market. Its better to have embedded graphics functions on the mainboard to speed up functions, an upgradeable socket for a mainboard wouldn't be that bad either maybe in the future as a next step! Faster than using the PCI-E Slot, but the peripheral component interconnect slot gives you more options for upgradability for future releases of products, limiting the need for throwing out your mainboard every 1-3 years depending on when you upgrade!


Interesting, but none of that applies to what I wrote.
m
0
l
August 5, 2010 10:36:21 PM

I'll wait for the benchmark results before anything else. We need to see the actual performance.
m
0
l
August 6, 2010 12:56:08 AM

fazers_on_stun said:
Hmm, this blog is by John Fruehe, a Director of AMD Marketing (and no, that's not a non-sequitur :p ). He posts a lot over on AMDZone, which should tell you something - along the lines of "it's pretty safe preaching to the choir" esp. when you have a mad dog of a mod banning anybody who questions anything concerning AMD.

Previously he was touting Magny Cours as putting paid to Intel's server business. Unfortunately, both AMD's and Intel's last quarterly earnings reports show Intel actually increased its server marketshare, despite (or maybe because of) MC being available the entire 3 months and competing against both Nehalem EP and Westmere Xeons.

So (1) JF is a paid 'shill' by definition, and (2) his track record is rather abysmal.

Hopefully Global Floundries will get their gate-first HKMG sorted out and yields up on 32nm, since Bulldozer will be released on that node. However GF's problems have already caused AMD to push Llano back, in favor of the netbook fusion CPUs which are being made on TSMC's bulk 40nm process.

Lol do you just look for threads with AMD mentioned in them so you can vent your anger towards them? :lol: 
m
0
l
a c 85 à CPUs
a b À AMD
August 6, 2010 2:44:37 AM

Wanted to point out real quick that just because there is this forum on the 24th and just because AMD is giving a speech doesn't 100% mean we get new details. It might be very possible that they will talk in the general sense and we don't get any real new info. Just thought I'd mention this before some fanboi gets his hopes up to much.
m
0
l
a b à CPUs
August 6, 2010 5:12:54 AM

They did say that they'll give some more details about the architecture.
m
0
l
a c 85 à CPUs
a b À AMD
August 6, 2010 5:19:07 AM

Everyone has a different definition of what "some" means. I'll bet we'll get a firmer release date, along with another high level view of the chip. I doubt we'll really learn to much more about it, but heres to hoping I'm wrong.
m
0
l
a c 110 à CPUs
a b À AMD
August 6, 2010 12:04:34 PM

yannifb said:
Lol do you just look for threads with AMD mentioned in them so you can vent your anger towards them? :lol: 


Sad, ain't it?

John Fruehe posts in the forums at Anandtech and Xtreme Systems, and as a credit to his intellect, avoids the CrapHole which is Toms.
m
0
l
a b à CPUs
August 6, 2010 5:02:18 PM

yannifb said:
Lol do you just look for threads with AMD mentioned in them so you can vent your anger towards them? :lol: 


Exactly which part of "John Fruehe" did you not understand? :sarcastic: 

It seems the OP and others here may not be aware of JF's role at AMD, and thus likely to take his blog opinions without the proper amount of skepticism required by his paid job of promoting AMD products. Have you ever seen him publicly acknowledging any AMD product shortcomings? I haven't, and my guess is that his job would be at risk if he completely candid instead of just flogging AMD stuff.

Now if he limited his postings to just the facts, I wouldn't have a problem with that. However his opinions should rightly be suspect, particularly given his track record. IIRC he has admitted that he is not an engineer as well, so any technical opinions should be even further seasoned with NaCl.

m
0
l
a b à CPUs
August 6, 2010 5:14:07 PM

Wisecracker said:
Sad, ain't it?

John Fruehe posts in the forums at Anandtech and Xtreme Systems, and as a credit to his intellect, avoids the CrapHole which is Toms.


No, what's sad is AMD's perennial 'on-the-cheap' advertising in the form of blogs by employees, posters on the back of the bus, Youtube, etc.

And JF doesn't have much of an aversion to the "Craphole" that is AMDZone, probably because as I mentioned previously, he can hide under Ghost's ban-hammer. Besides, I have seen him post here on THG although not recently. His post got ignored so perhaps he gave it up here as a bad job..

Also, from what I've seen on XS, he tends to disappear when catching flak from other posters. I guess that as director of marketing, he can't be seen alienating too many potential customers.
m
0
l
August 6, 2010 7:28:36 PM

fazers_on_stun said:
No, what's sad is AMD's perennial 'on-the-cheap' advertising in the form of blogs by employees, posters on the back of the bus, Youtube, etc.

And JF doesn't have much of an aversion to the "Craphole" that is AMDZone, probably because as I mentioned previously, he can hide under Ghost's ban-hammer. Besides, I have seen him post here on THG although not recently. His post got ignored so perhaps he gave it up here as a bad job..

Also, from what I've seen on XS, he tends to disappear when catching flak from other posters. I guess that as director of marketing, he can't be seen alienating too many potential customers.


I've yet to see John making any far-fetched comments as, say, Dr. Who - from Intel. The later even threatened XS members over some leaked engineering samples. Just put their posts side by side and the Doctor will look much more of a marketing puppet than the marketing professional John is.

To the credit of AMDZone, I usually read some very good discussions at the place, especially when dresdenboy, Hans De Vries and John are involved - and even the greener members recognize many of Intel's merits where credit is really due. The shenanigans come mainly from its corporate culture and behavior.
m
0
l
a b à CPUs
August 6, 2010 9:34:18 PM

^ TBH I haven't paid a lot of attention recently to the XS forums, so I missed the Dr. Who threats. I'm guessing it was the not-too-impressive SB ES chip reviews at the root of the problem, as Intel is usually not too concerned over NDA leaks when they show the ES chips as impressive.

However I did recently take a look at AMDZone, particularly the http://www.amdzone.com/phpbb3/view [...] 2&t=137815 thread, and must admit I was very surprised to see Abinstein (a mod) hold JF's feet to the fire over the GF spin-off:

Quote:
"JF-AMD wrote: I generally don't comment on these threads because they start out on a false premise and just keep getting worse.

How many "end of the world for AMD" threads over the past 4 years have been right? "

I'd say some of them were partially right.

Yes, you are still in business, but you lost 80% (perhaps more) of your market value, and 70% ownership of your fabs. The ownership of the fabs will keep dwindling to keep the company profitable.

I have to say that I myself underestimated AMD's problem when I said 2 years ago that I believe AMD would not sell off the fabs. It turns out AMD had to sell off the fabs.

My point here is clear: (1) AMD needs access to state-of-the-art fabs in order to stay competitive. Don't even try to tell me otherwise, or why don't you manufacture Bulldozer at 45nm? You had 5 years to do so. (2) The only company who can provide this with capacity is GlobalFoundries. (3) Unfortunately, 32nm SOI is not the highest priority of GF; 28nm bulk is.

If GF takes off, and it starts to make high performance processors for other companies, then AMD can leverage on the cost sharing which would benefit everyone in the industry. If, on the other hand, GF does not attract enough customers, then either AMD can shoulder the R&D cost at GF, or it can say goodbuy to GF (can it?) and send chips off to TSMC or Samsung to manufacture. Then it will be at a greater process disadvantage to Intel.


Quite surprising, to say the least, to see that from Abinstein. However, true to form, JF denies any problems whatsoever with AMD or GF.

Anyway, after the treatment given Johan de Gelas following just about any of his server articles, my opinion of AMDZone remains that it is mostly full of AMD fanbois zombies...
m
0
l
!