AMD Bulldozer Thread

[FlanK3r]

Olivion: but this is multi-GPU, not single GPU. This was "clear"....

[kaktus1907]

AMD FX-8150 Multi-GPU Gameplay Performance Review

November 03

http://www.hardocp.com/article/2011/...mance_review/1

hmm slower but not smoother either


i wonder what causes this the frame rate is all over the place in some games.. (Arma 2, DA2)

[Sgt.McRuff]

TDP limit is still on and thus its down clocking?

[xdan]

BD is a crap cpu, broken arhitecture, that's the only motiv.
Plain and simple!

[AliG]

looks like AMD is cutting 10% of its workforce,according to the Associated Press.
AMD to cut 10 pct of workforce as PC weak

maybe they can put that extra $10 million to good use?

Steve Jobs proved that if you push people to their limits, you'll see greatness. I don't think AMD's problem is money, it's leadership. They have so much instability in their board and the CEO seems to be a rental position these days. I think all they need is a guy that's not afraid to get into their employees faces from time to time and tell them they aren't cutting it.

Granted I'm way over simplifying things, as Jobs did have a tremendous ability to read people's emotions and be charismatic as well. With that being said, before he went off on the deep end of price gauging etc, he basically changed the entire computing industry with a team 6 of people. If AMD can find a guy like that they'll be competing with Intel again in less than a decade (yeah that seems like a while, but let's be realistic, this architecture sucks, it'll take at least 1 or 2 more to get back on the ipc playing field).

[gallag]

I see a lot of talk about windows 8 helping bd due to better thread managment, seems to be the prefered excuse atm, am just wondering if that will also benifit intel cpus? Probaly not the 2500 with no ht but will better thread managment not help ht enabled cpus like the 2600?

[informal]

Unlike the bulldozer's case,Windows 7 already handles i7's SMT cores quite well.

[gallag]

How do we know that? Mabey win8 will give both amd and intel a boost?

[Leeghoofd]

clock speeds are crap in that review, 4.6ghz on an FX-8100 would be a complete waste, and someone would rather disable some cores or go with a 6100 and overclock it higher. it wouldnt be enough to beat a 2500k, but it would bring the gap up way closer is some titles.

I tested 4/4 and it isn't faster clock per clock, at least not in the games I tested it with... and to be honest the end user shouldn't be bothered with this fine tuning at all... Only thing that it allows you to do is to clock up higher and drop power consumption... But it will need to be at least at 5Ghz and beyond to give the raw power of the Intels a run for their money... and for that to be usable 24/7, Global Foundries has to make up for their crappy production process.... as they are responsible for the final blow to this new architecture

The weak multi GPU game performance can be easily spotted in 3DMark06 and Vantage as the GPU tests are way too low. Even lower as a Thuban at 4Ghz ( FX8150at 4.6) Anyone ever wondered why we saw so many Heaven benchmarks ? Coz its 99% if not more GPU limited, if you even loose 100 points there on the score, something is terribly wrong...

Solution to this all : harsh price cuts and a better GF production to allow us to run these things over 5Ghz... All the other software recompiling and stuff is encouraging, but most of the image damage has been done... and later this month another thing is coming...

Also most reviewers are testing with ES CPUs that don't seem to clock as high as the retails, mine would run the full test suite at 4.9, though prime stability was at max 4.6-4.7Ghz...

[freeloader]

Leeghoofd......"and later this month another thing is coming...", care to elaborate?

[Olivon]

I guess Leeghoofd talked about SB-E launch coming in 10 days, the 14th

[tom1]

other two reviews

http://www.xtremehardware.it/eng-rev...-201111146032/

http://www.xtremehardware.it/eng-rev...-201110126039/

maybe if you want to add

[tom1]

other two reviews

http://www.xtremehardware.it/eng-rev...-201111146032/

http://www.xtremehardware.it/eng-rev...-201110126039/

maybe if you want to add

[Gambit_2K]

other two reviews

http://www.xtremehardware.it/eng-rev...-201111146032/

http://www.xtremehardware.it/eng-rev...-201110126039/

maybe if you want to add

How is that a review? It's an analysis of the architecture, they dont even have any form of performance test in the article...

[fellix_bg]

AMD confirms: Bulldozer FX only 1.2 instead of 2 billion transistors

How we test the bulldozers FX-8150 indicate the underlying Orochi chip has, according to AMD's over 2.0 billion transistors details - we can not count, so we must rely on the values ​​communicated by the manufacturer. Der ist sich über die Anzahl der Transistoren offenbar nicht sicher, denn im Zuge des Launches der Opteron 6200 und 4200 sprachen die Kollegen von The Register auf einmal von nur 1,2 Milliarden Transistoren. Which is about the number of transistors does not seem safe, because in the wake of the launch of the Opteron 6200 and 4200 , the colleagues of The Register spoke at once of only 1.2 billion transistors.

Wir haben bei AMD explizit nachgefragt und erhielten als Antwort, dass der zuvor kommunizierte Wert von 2,0 Milliarden Transistoren falsch sei und es "nur" 1,2 Milliarden Transistoren seien - wohlgemerkt wie gehabt verteilt auf ein Die mit 315 Quadratmillimetern. We asked specifically for AMD and received in reply that the previously reported value of 2.0 billion transistors is wrong and it is "only" 1.2 billion transistors are - mind you had as distributed on the one with 315 square millimeters. Zum Vergleich: Intels Sandy Bridge DT, also etwa der Core i7-2700K, kommt auf 216 Quadratmillimeter sowie 1,16 Milliarden Transistoren (auf dem IDF von ursprünglich 995 Millionen nach oben korrigiert), ist sparsamer und meist schneller. For comparison, Intel's Sandy Bridge DT, or about the Core i7-2700K comes to 216 square millimeters and 1.16 billion transistors...

[STaRGaZeR]

????

Huge ass die on GF's 32nm process is only 1,2 billion transistors? Llano has 1,45 billion transistors, 228 mm^2.

BS.

[-Boris-]

Why does so many people blame GF for bulldozers failures? Even if Bulldozer was at 5GHz+ it wouldn't touch 2600K territories. No company can produce bulldozer at those frequencies. And I doubt any company can make a huge chip with 2 (or 1.2?) billion transistors energy efficient at frequencies around 4GHz, compare with SB-E.

[informal]

????

Huge ass die on GF's 32nm process is only 1,2 billion transistors? Llano has 1,45 billion transistors, 228 mm^2.

BS.

GPU part in Llano supposedly has very high transistor density.

[Hornet331]

Lol wat?

marketing doesn't even know how much transistors there own product has?

WTH is going on in AMD?

[saint19]

Lol wat?

marketing doesn't even know how much transistors there own product has?

WTH is going on in AMD?

Bulldozer

[STaRGaZeR]

GPU part in Llano supposedly has very high transistor density.

I know you like to suppose a lot, but the official figures are ~2B transistors for the die and this is pure BS. That or AMD's PR department just hit another level of mediocrity. 1,2B on a process that is known to be more dense than the competition?

[Hornet331]

I know you like to suppose a lot, but the official figures are ~2B transistors for the die and this is pure BS. That or AMD's PR department just hit another level of mediocrity. 1,2B on a process that is known to be more dense than the competition?

And you to... density is also dedicated by the logic used and not only by the process itself. Amd always had worse cache density on there cpus and suddenly they beat intel by a large marigin...

[fellix_bg]

The SOI process technology is rather counter-productive for dense logic structures - certainly not as good as node-matching bulk process.

[informal]

I know you like to suppose a lot, but the official figures are ~2B transistors for the die and this is pure BS. That or AMD's PR department just hit another level of mediocrity. 1,2B on a process that is known to be more dense than the competition?

Official number has been corrected now,it's 1.2B and die size is 315mm^2. Bulldozer is server oriented chip and a lot of stuff was made with that target in mind. That's why it's somewhat larger die. Also as Hornet said, intel always had higher cache density versus Kx generation.

[STaRGaZeR]

Where is that official correction?