AMD FX-8120P benchmark from Coolaler

[OCGuy]

It's not that people "buy" tons individually. It's that many "for the Hertz" OCers will decide to try their luck on a particular BD. Some others will buy->test->sell->buy BDs for OCing. Hence my saying 10s of thousands and not 100 thousand plus. Helps AMD avoid the embarrassment of low demand for a brand new uArch.

If AMD's plan to stay afloat is exotic cooling death-runs by extreme enthusiasts.......

 

[BlueBlazer]

As for people saying that IPC will be lower than Phenom II,.... what are you smokeing?
Why would you make a CPU thats slower than your last gen one? also AMD guys like JD-AMD have come out and said its IPC will be higher than phenom II (magny core, mhz pr mhz).

You should not take every word literally (as truth) from JF-AMD. That's my take on this subject. :hmm:

All the "leaked" bulldozer benchmarks are "fakes" done to get page hits as far as Im concerned.

Like the leaked GigaByte screenshot? Its well known fake that was exposed. That one is for "getting page hits". Another blast from the past is "Barcelona" Benchmarks? Don't Believe 'Em Just Yet, and the prankster boasts his exploits here (you can find this joker at XS forums). Of course, when a benchmark looks bad (pictures gone but can be found here and here) then its often ignored and overlooked. That one IMHO is a legitimate leak. :sneaky:

 

[Vesku]

If AMD's plan to stay afloat is exotic cooling death-runs by extreme enthusiasts.......

A lot of casual OCers follow up on what the extreme enthusiasts are doing. Expect plenty of air and turnkey water OCs to be posted when BD becomes available and for these BD speculation threads to be replaced with voltage, etc. threads for at least a while.

Oh, and of coures I forgot the sure to be entertaining SB OC vs BD OC threads.

 

[Piano Man]

A little history here (although its so old it may not be analogous to today's discussions, its still interesting to see how this played out in the past).

Back in May of '99 (one month prior to actual release), AMD invited Firingsquad to come in and bench their upcoming K7. It was an Enginerring Sample, but it was also under AMD's own controlled environment. Needless to say, the tests came back pretty iffy:

http://firingsquad.com/hardware/k7550preview/default.asp

Fast forward 2 months, and now FiringSquad has the release version of the K7 in there own hands, and its like night and day on the benchmarks. It ate the PIII's lunch. They even called the article "The Real AMD K7", because it was such night and day difference.

http://www.firingsquad.com/hardware/athlon600preview/default.asp


Now I know its 12 years later, but I'm just saying. Its happened once before where Engineering Samples have been lackluster, but when it came for release (and only one month after the ES's were tested), that the final chips really shown through.



Just putting that out there for those of us who still have some hope left

 

[BlueBlazer]

Oops.. yeah. I am confusing that with chiphell.

Chiphell often takes news from elsewhere as well, just like any other sites. Thus, its better to find the original source... :hmm:

Anyway.. my point was that AMD knows how to keep people in the shadows. I am not saying that BD will be performing exceptionally.. rather.. I am just implying that AMD has a history of keeping a tight wrap on their ready to launch products.

No matter how tight they try to keep it under wraps, sometimes leaks do happen because they need to send those CPUs out for testing and validation.

This seems so bad, it has to be fake.

Then there are others that are too good to be true. As always take them with a grain of salt if the information doesn't come from well known/recognizeable sources. :hmm:

Andy Glew (the original inventor of CMT) has some fascinating insights about what it's like to shift horses onto a new microarchitecture.

http://groups.google.com/group/comp.arch/msg/82efd5dc0ef96d38?hl=en

Interesting read there....

A lot of casual OCers follow up on what the extreme enthusiasts are doing. Expect plenty of air and turnkey water OCs to be posted when BD becomes available and for these BD speculation threads to be replaced with voltage, etc. threads for at least a while.

Oh, and of coures I forgot the sure to be entertaining SB OC vs BD OC threads.

Interesting times ahead... :thumbsup:

 

[jhu]

Well, opinions are like assholes: everyone has one. Until someone gets his hands on a production CPU and benchmarks it (perhaps someone who works at Cray could help out?), we'll just have to wait.

 

[classy]

Here is another look by a blogger. Who knows??????? One thing is for sure, this is most memorable cpu launch I in my IT lifetime.

http://obrovsky.blogspot.com/

Additional comment. It appears that clearly there is without question a bios issue.

 

[RussianSensation]

Don't think it's a trick, who fakes respins and delays for the sake of some misinformation? Keep in mind how long BD has been in the pipe, it may actually look ok compared to Nehalem, the architecture that they were originally looking to launch against.

X6 1100T performs really well in a few benchmarks such as Cinebench but is slower for just about everything else vs. Nehalem. So IF it takes 8 BD cores to match a Nehalem in Cinebench, then in 1-4 threaded apps, BD is toast against Nehalem and surely against SB. I am still hoping these benchmarks are using ES CPUs with bugs or are fake. Otherwise, AMD is officially off the map for CPU purchases for another 2-3 years. I mean even at $190 the X6 doesn't make any sense vs. SB CPUs, which means there is no way they would be able to launch an 8-core CPU @ $220-260 that can't even beat a $190 chip they have now. I am leaning towards these benchmarks not being representative of final shipping product or AMD is going to need Apple's marketing team to try to sell them.

 

[busydude]

Here is another look by a blogger.

OBR is a massive troll.. those results are all fake(by his own admission).

 

[Phynaz]

The point is just... I dont think it stops at 8.4ghz, engineering sample's and only testing like 4 to get to that record, isnt enough, theres bound to be better cpu's

What makes you think they only tested 4? Those are the four AMD brought, there is no mention of how many AMD tested to get those 4.

 

[BlueBlazer]

Now I know its 12 years later, but I'm just saying. Its happened once before where Engineering Samples have been lackluster, but when it came for release (and only one month after the ES's were tested), that the final chips really shown through.

And that time Sanders was at the helm, not Hector.

Here is another look by a blogger. Who knows??????? One thing is for sure, this is most memorable cpu launch I in my IT lifetime.

http://obrovsky.blogspot.com/

This joker has send fakes out to websites before (example victim), and then announce "you were punk'd". He's very well known now (for the stunts he pulled). Thus, do take tons of salt on information he posted.

This chart is seriously flawed. Reference here, Core i7 2600K scores around 6.89, Core i5 2500K scores 5.43, Phenom II X4 850 scores are too low (should be around at least 3.6 or higher), Core i7 990X scores should be above 9.

 

[Vesku]

X6 1100T performs really well in a few benchmarks such as Cinebench but is slower for just about everything else vs. Nehalem. So IF it takes 8 BD cores to match a Nehalem in Cinebench, then in 1-4 threaded apps, BD is toast against Nehalem and surely against SB.

That's why I'm really skeptical of that Cinebench score, they were targeting Nehalem and had to re-adjust for SB, if after the delays and respins they are struggling against their own x6 cpus . As Idontcare mentioned, if this is the case, I could see why the AMD board would be so abrupt about letting Dirk go. Although they really should have been CEO shopping and waiting until final silicon results to follow through with the boot.

 

[inf64]

This joker has send fakes out to websites before (example victim), and then announce "you were punk'd". He's very well known now (for the stunts he pulled). Thus, do take tons of salt on information he posted.

Also he has a thing for me personally ,it seems D:

 

[RussianSensation]

EVEN if that is true above, taking data from post #32 in this thread:

Phenom II X6 1100 @ 3.3ghz = 5.87 (so 6 Phenom II cores = 8 core FX 8150 ???)

And in overclocked modes:

Phenom II X6 1100 @ 4.0ghz = 7.22
Core i5 2500k @ 4.4ghz = 7.08
Core i7 2600k @ 4.8ghz = 9.45

So FX-8150 @ 4.85ghz = 7.9 would still be disappointing. That means per core performance is severely lacking. To me an 8 core CPU should beat a 4C/8T SB in an 8 threaded app, no questions asked.

 

[BlueBlazer]

Also he has a thing for me personally ,it seems D:

Just noticed that a few moments ago. You're now famous!

 

[OVerLoRDI]

You can have all the collective brilliance of Nobel laureates in physics, chemistry and mathematics at your command, but if you suck at leadership and vision then they will surely flounder under your misguided directives all the same.

What made MS different from IBM? Or Intel different from AMD in the late 1970's and early 1980's? What made Nvidia different from Matrox?

Leadership.

And in Jan of this year, for some very good reason that's never been made all that clear to its shareholders, AMD's Board of Directors elected to dismiss the leadership that gave rise to Bulldozer.

Well said. You assessment of the situation looks very probable. Hopefully AMD can recover from this debacle.

 

[classy]

EVEN if that is true above, taking data from post #32 in this thread:

Phenom II X6 1100 @ 3.3ghz = 5.87 (so 6 Phenom II cores = 8 core FX 8150 ???)

And in overclocked modes:

Phenom II X6 1100 @ 4.0ghz = 7.22
Core i5 2500k @ 4.4ghz = 7.08
Core i7 2600k @ 4.8ghz = 9.45

So FX-8150 @ 4.85ghz = 7.9 would still be disappointing. That means per core performance is severely lacking. To me an 8 core CPU should beat a 4C/8T SB in an 8 threaded app, no questions asked.

I never said it was true. I stated who knows. But the one thing I will say 4.8 is not easy overclock for 2600K. Many and I do many have topped out at 4.5-4.6. And whether or not a hard 8 should beat a 4/8 virtual is also not cut and dry either.

 

[RussianSensation]

I never said it was true. I stated who knows. But the one thing I will say 4.8 is not easy overclock for 2600K. Many and I do many have topped out at 4.5-4.6. And whether or not a hard 8 should beat a 4/8 virtual is also not cut and dry either.

Oh ok, sorry about that. Ya, I agree that 4.8ghz is on the high-end for 2600k on reasonable voltage. I mainly linked it to show performance at the same frequency.

 

[thilanliyan]

This is (if true) definitely disappointing to us enthusiasts, but if most of the money is in servers, then it doesn't make much of a difference to AMD IMO.

 

[Elixer]

This is (if true) definitely disappointing to us enthusiasts, but if most of the money is in servers, then it doesn't make much of a difference to AMD IMO.

It also seems that Trinity is hot on the heels of BD, Anand wrote:

Trinity is a higher end successor to Llano. It will integrate 2 - 4 Piledriver cores and feature some sort of a Radeon GPU on-die. Performance across the board will be higher than Llano. Piledriver should offer better performance per clock vs. Bulldozer, but Trinity won't be targeted at the same high-end market that AMD's FX series based on Bulldozer will address. Presumably we'll see a Piledriver based FX CPU at some point down the road. AMD indicated that Trinity was a few months away, implying an early 2012 launch.

Which could mean that they have worked out all the issues they had with the first revision of BD, and Piledriver will improve upon BD FX's shortcomings (if ppc is a real issue with BD FX).

 

[inf64]

Hmm,multi clock domains in Bulldozer module. Think about that for a second. And don't forget power draw(high power draw of certain unit(s) ).
There could be the "problem".

 

[Arkaign]

All of this talk about BD brings me back, and reflecting on the first-gen releases of many past chips.

1st Pentium fail. 60/66mhz models only, orphaned socket.
1st P2s kinda sucked honestly (66mhz bus, not very good overclocking)
K5 was fail all the way in every way.
1st P3s also mediocre (overpriced, very little gains IPC over P2 of same bus/clock)
1st K7s I thought kinda sucked as well. Hot, slow L2 cache, crappy mobo chipset
1st P4s sucked hard. Rdram, slower in lots of stuff, expensive, short lived socket
1st K8s were mediocre. Socket 754 lame. Socket 940 not great for home users.
1st Phenom was lolbad

Lets see, on the other hand a lot of winners :

Core 2 Duo, insta-win.
Socket 939 X2s/Opterons were boss for a long time, great mobos too
Socket 1366 / i920 was a great value that is still competitive, and they offered 6-cores there too
Super Socket 7 K6s were initially pretty good vs. Pentium MMX at the time, but of course :
Slot 1 Celeron 128k 300a/333 became the chip of choice for a long time, easy OC to 450/500mhz, and the on-die core synchronous cache was actually better than the 512k of half-speed cache in the much more expensive p2-450 and p3-450s.

 

[WhoBeDaPlaya]

Hey there now, Socket 754 was just fine. The cost savings was justifiable and I used it for the longest time with an A64 3700+ until upgrading to a C2D E6600.

 

[Soulkeeper]

Another thing I think we should all consider with these early benchmarks is the compiler and optimizations.

I suspect (being a new arch, having avx, etc.) that we could see some slow/odd behavior with certain software and potentially large improvements as things get updated and software that wasn't compiled with 1yr+ old compilers gets deployed

 

[Arkaign]

Hey there now, Socket 754 was just fine. The cost savings was justifiable and I used it for the longest time with an A64 3700+ until upgrading to a C2D E6600.

Yeah it wasn't entirely useless, but I thought it was something of a trap. I don't like the idea of heavily segmented sockets (like Intel is doing now to some extent), but S754 was pretty annoying. It would be as if the SB launch went like this :

Socket 1155 = core i3 only, no quad core or IB for y00
Socket SB Xeon = Xeon variants only, need special ram + expensive board
Socket 1157 = core i5, core i7, but not available for awhile

Given that Socket 940 was already available, and there were only the most minor differences between that and the eventual socket 939 release, I think it would have done the community a great service to go ahead and release 939 instead of f'ing around with 754, which left no dual-core upgrade path. It's not like 939 chips were any more expensive to make than 754 chips, and the mainboards weren't crazy expensive either unless you wanted the more exotic SLI-type boards.

I guess the same criticisms can be made of many short-lived sockets, but that one irked me more than most. Socket 423 was a clusterfsck as well heh.