Fudzilla: Bulldozer performance figures are in

[Idontcare]

I have never understood why people get so emotional over things.

If the product is right for you, you buy it. Pretty simple.

That rant reminds me of this: http://www.youtube.com/watch?v=kHmvkRoEowc

Oh man you had me laughing out loud on that one

 

[BlueBlazer]

How long have you been a member here? You should know better than to believe faked, doctored up BS benchmarks. I'd expect that from a newbie. Not from you.

Hmmn, let's see, go off the BS links you believe in, or, perhaps this one? From an actual AMD event with B2F\G stepping and REAL benchmarks, not just the world record\gaming FPS one?

http://www.xtremesystems.org/forums...es-Bulldozer&p=4950183&viewfull=1#post4950183

Also note he doesn't just say "multimedia" machine.

He says GAMING machine.

Do note the "real" benches part (including the "" quotes). He could be referring to the (questionable) leaked "slides" that shows the GPU limited graphs? The only real benches IMHO that could be considered reliable are those coming from Coolaler and his company of known leakers. And it could be argued that those are B2 stepping engineering samples but then how much improvement in the final stepping can be achieved? There's also that NordicHardware quick test on an actual demo machine. :hmm:

 

[Abwx]

And it could be argued that those are B2 stepping engineering samples but then how much improvement in the final stepping can be achieved? There's also that NordicHardware quick test on an actual demo machine. :hmm:

Unless the previous revisions were badly bugged or deliberately crippled,
a final revision will bring very few perfs improvements , the most likely
is increased frequency as the main target of such steppings.

Indeed , the true final stepping is rather to make some compilation
to take advantage of the new CPU uarch.

 

[iCyborg]

If Interlagos is supposed to be shipping out, how come there are no benches for it? Still under NDA until end customers get it or what's the deal here?

 

[BlueBlazer]

If Interlagos is supposed to be shipping out, how come there are no benches for it? Still under NDA until end customers get it or what's the deal here?

Indeed, and still waiting to see some SPEC benchmarks but alas none. AMD was shipping them to Cray mostly/exclusively (for supercomputer installations). The only glimpses of Interlagos can be found on Sisoftware website (possibly from motherboard testing and validation, like this MSI one). :hmm:

 

[Black96ws6]

Do note the "real" benches part (including the "" quotes). He could be referring to the (questionable) leaked "slides" that shows the GPU limited graphs? The only real benches IMHO that could be considered reliable are those coming from Coolaler and his company of known leakers. And it could be argued that those are B2 stepping engineering samples but then how much improvement in the final stepping can be achieved? There's also that NordicHardware quick test on an actual demo machine. :hmm:

He's not. If you read the whole thread (which takes awhile), he saw other REAL LIVE benches as well, not slides. He was invited to the event. He just can't talk about them.

And in another thread he said his next build will be AMD.

I think that speaks volumes.

And again, I could care less about x87 SuperPi benches on an ES sample.

JF-AMD has already gone on record as stating single threaded performance INCREASES.

So, there's logically only 2 options here:

1. Those benches are crap\BS AND\OR there's something wrong with the ES they tested,
2. Those leaked benches are correct and JF-AMD and MovieMan are lying.

Seriously, how can you honestly believe it's #2? Which option has the higher probability?

 

[Cogman]

He's not. If you read the whole thread (which takes awhile), he saw other REAL LIVE benches as well, not slides. He was invited to the event. He just can't talk about them.

And in another thread he said his next build will be AMD.

I think that speaks volumes.

And again, I could care less about x87 SuperPi benches on an ES sample.

JF-AMD has already gone on record as stating single threaded performance INCREASES.

So, there's logically only 2 options here:

1. Those benches are crap\BS AND\OR there's something wrong with the ES they tested,
2. Those leaked benches are correct and JF-AMD and MovieMan are lying.

Seriously, how can you honestly believe it's #2? Which option has the higher probability?

Easy, this is a major architecture change. There is a GOOD chance that some things will do worse in single threaded performance as a result of the changes. JFAMD doesn't have to be lying, rather, he could have been referring to specific benchmarks/tests that he used.

 

[sm625]

HD4870 was faster than the GTX260 and cost $100 less. I don't think anyone expects the $205 FX-8120 to be better than the $317 2600k.

I do. Absolutely. In games anyway. I expect a vertically integrated cpu/gpu company to build a cpu with a special instruction set that boosts gaming fps by at least 30%, when used in combination with their own drivers. Anything less is an utter failure. They may not have OpenCL down yet but the radeon driver should take advantage of their ability to create and use whatever instructions they want or need.

 

[Black96ws6]

Easy, this is a major architecture change. There is a GOOD chance that some things will do worse in single threaded performance as a result of the changes. JFAMD doesn't have to be lying, rather, he could have been referring to specific benchmarks/tests that he used.

We won't find out the truth until mid-October when we can see legitimate benchmarks.

Again, if you compile all the facts from LEGITIMATE sources, it paints a picture of BD being a good bang for the buck. Here's a few more examples:

<totally rhetorical question, NOT real numbers>
Which would you rather have:

80&#37; of the performance with 50% of the cost and 50% of the power consumption
100% of the performance with 120% of the cost and 120% of the power consumption

</end rhetorical question>

People keep seeing that 80% number and thinking that it is a compromise.

What they don't understand is that by sharing components we are able to add more cores in the same die space and same power budget.

It is by no means 80% of today's performance.

And from another board:

People do not get one thing.

The comparison of the BD module having 180% performance of 2 cores (which would have 200%) is not done in regards to a 10H core. So... a BD module (with two cores) is not 180% of 2 thuban cores. That is plain dumb to believe.

The 180% is vs 2 proper BD cores. So, instead of having 2 full Bulldozer cores for 200% perf, they chose a module aproach, giving 180% but at 50-60% the size of 2 full cores. Thus, you pack more cores into the same die.

This is exactly the thing people don't get, that BD cores lose 5-10% compared to Thuban. That is just naive to believe. They lose 5-10% in fully threaded applications vs 2 theoretical BD full cores.

How many times do I have to tell you that bulldozer has higher IPC than our current architecture?
Is somebody being paid by intel to continually post these statements?

See, that statement is what gets people in trouble. Someone reads that statement and assumes 10% lower performance.
IPC will be higher than previous generation
Single threaded performance will be higher than previous generation

Well, if it holds it's own against a 980X that sort of speaks volumes as to how it will do against a 2500k or 2600K yes?

There's a ton of rumors going around on the BD chips and I can honestly tell you
that many of them are SO wrong it's not even funny.I was there in Austin and saw with my own eyes a lot of what these can do!

Look at the above quotes I have mined from other boards and judge for yourself. That's all we can really do until Anand benches this thing and releases the results.

 

[trollolo]

the amount of anal frusturation in this thread is astounding.

 

[Cogman]

We won't find out the truth until mid-October when we can see legitimate benchmarks.

Again, if you compile all the facts from LEGITIMATE sources, it paints a picture of BD being a good bang for the buck. Here's a few more examples:

And from another board:

Look at the above quotes I have mined from other boards and judge for yourself. That's all we can really do until Anand benches this thing and releases the results.

And I wasn't trying to say that BD won't perform well. Rather, I was just trying to explain that there is the possibility that some benchmark somewhere could suffer from the changes in architecture (I have no clue what will really happen).

I haven't looked at the leaked benches, I just don't trust them. Rather, I'm waiting for a REAL review from someone like anand.

 

[BlueBlazer]

We won't find out the truth until mid-October when we can see legitimate benchmarks.

Again, if you compile all the facts from LEGITIMATE sources, it paints a picture of BD being a good bang for the buck. Here's a few more examples:

And from another board:

Look at the above quotes I have mined from other boards and judge for yourself. That's all we can really do until Anand benches this thing and releases the results.

Since you like to quote Movieman that much, you may have missed this.....

We have two companies that compete for our and the rest of the worlds business.
They both bring out at times excellent products and once in a while real dogs.
There's a ton of rumors going around on the BD chips and I can honestly tell you
that many of them are SO wrong it's not even funny.
I was there in Austin and saw with my own eyes a lot of what these can do BUT
that is different from setting up your own machine and then KNOWING what it will do.
Same with Intel, they all want to show you their products in the best light possible and that they should be doing.
After all, this is business right guys?

You cut short at the "BUT" statement, while this is the full statement. That's open for interpretation. As for JFAMD, I caught him here when he earlier denied there was an Opteron 6282SE part (from Sisoftware site). :hmm:

He's not. If you read the whole thread (which takes awhile), he saw other REAL LIVE benches as well, not slides. He was invited to the event. He just can't talk about them.

And in another thread he said his next build will be AMD.

I think that speaks volumes.

And again, I could care less about x87 SuperPi benches on an ES sample.

JF-AMD has already gone on record as stating single threaded performance INCREASES.

I've also asked him about the IPC thingy earlier (still awaiting answer).

So, there's logically only 2 options here:

1. Those benches are crap\BS AND\OR there's something wrong with the ES they tested,
2. Those leaked benches are correct and JF-AMD and MovieMan are lying.

Seriously, how can you honestly believe it's #2? Which option has the higher probability?

There's a 3rd option - The leaked benches are correct (meanwhile AMD was fixing performance issues that caused the delays and lack of confirmation dates), JF-AMD was quoting estimated performance (theoretical based on simulations) and Movieman was shown demo of AMD FX in a better light (see Movieman's quote above).

 

[Idontcare]

Easy, this is a major architecture change. There is a GOOD chance that some things will do worse in single threaded performance as a result of the changes. JFAMD doesn't have to be lying, rather, he could have been referring to specific benchmarks/tests that he used.

While you are absolutely correct when taken to a pedantic extreme, with >700 instructions in the supported ISA it IS true that it would be astounding if every one of them experiences an improvement in their execution rate with the PhII -> BD transition.

(^ isn't that graph the bomb?)

But I would think that to people ingrained into the industry these sorts of caveats would go without saying, i.e. JF would never think he need say such self-evident caveats.

However, I would expect the onus to be on JF to clearly delineate in his own comments, when he originally made them, if the case really were that he was specifically referencing some cherry-picked hand-selected niche set of applications which proved his point.

Its not our responsibility to assume this to be the case, that is not the unwritten rule. Its his responsibility to make clear if this was the case when he made those statements. (<- yet another one of those things that just foes without saying)

Since he has not detailed such caveats, I do think it only fair to ourselves, and to him, for us to simply take him at his word and move forward with confidence in accepting as fact the operating condition that bulldozer's IPC does not decrease (either stays the same, or improves) compared to that of PhenomII product class microarchitecture in all but the corner-cases of applications.

There will always be exceptions to the rule, but we are within our rights to assume that it is not the exceptions that are themselves the basis of the assertions that have been made so far. This was true for HT, it will no doubt be true of BD as well.

My concern, THE bigger concern, is not the performance of Bulldozer but rather the profitability. AMD's new CEO is not going to doggedly pursue Intel in the x86 space if he comes to view it as a fool's errand from an investor returns perspective.

As captain of the ship he has a fiduciary responsibility to pilot the ship away from the hurricane, and bulldozer's profitability envelope will determine the funding priorities and funding levels for its successors.

 

[Idontcare]

I've also asked him about the IPC thingy earlier (still awaiting answer).

I don't want to offend you, but if I were a professional in this industry and I saw your post I would have written you off as yet-another-troll and simply would have never wasted my time responding to it.

I can't speak for John but he may have arrived at such an impression on the basis of your post, complete with trollish-type embedded graphic, and opted to just "move on".

Regardless how he took it, if he even saw it, John is not obligated to respond to every post addressed to him, and him not responding to a post is surely not to be taken as him avoiding a question somewhere in the post.

If that were the case then there are prolly dozens, if not hundreds, of dissatisfied members of these forums who are still waiting on a response from me to their post xyz.

 

[Abwx]

While you are absolutely correct when taken to a pedantic extreme, with >700 instructions in the supported ISA it IS true that it would be astounding if every one of them experiences an improvement in their execution rate with the PhII -> BD transition.

I pointed this instructions continual inflation earlier on this thread...

Since this graph was made , there s now AVX as well as at least
20 FMA related instructions in BD plus AMD s ex SSE5 that has
been rebranded XOP...

The 1000 instructions bar is really in sight...

 

[Black96ws6]

Another quote from JF-AMD related to performance that I forgot to include in the previous quote.

If you add all these up, put emotions and unreliable sources (OBR for example) aside, the picture painted does in fact seem to indicate that Bulldozer will be a winner:

Today's processors have 3 execution units that are shared between ALU/AGU. That is essentially 1.5 ALU and 1.5 AGU. With BD we get 2 AGU and 2 ALU. Much better.

So with BD you get 4uops per cycle vs 3uops per cycle with PhenomII.

That, to me, would seem to indicate an INCREASE in IPC, rather than a decrease. I realize that many other factors are involved as the architectures are completely different, maybe IDontCare can chime in with his thoughts...

 

[Cogman]

While you are absolutely correct when taken to a pedantic extreme, with >700 instructions in the supported ISA it IS true that it would be astounding if every one of them experiences an improvement in their execution rate with the PhII -> BD transition.

(^ isn't that graph the bomb?)

But I would think that to people ingrained into the industry these sorts of caveats would go without saying, i.e. JF would never think he need say such self-evident caveats.

However, I would expect the onus to be on JF to clearly delineate in his own comments, when he originally made them, if the case really were that he was specifically referencing some cherry-picked hand-selected niche set of applications which proved his point.

What if the application was cherry picked to be the worst case for the new architecture? I would assume that in general, everything will improve, however, it is not impossible for some benchmarker to find some test that AMD saw decreased performance in.

I'm really just trying to explain a crappy benchmark. I'm not trying to say anything about general single threaded performance. The guy I quoted said that "you would have to believe JFAMD a liar to believe those results" my response is "No, you wouldn't, here is why". I am not saying that the results are true, correct, or anything. I'm just saying that the results could be accurate and JFAMD telling the truth.

Perhaps I am being pedantic.. IDK, I was just born with that tendency.

 

[LOL_Wut_Axel]

He's not. If you read the whole thread (which takes awhile), he saw other REAL LIVE benches as well, not slides. He was invited to the event. He just can't talk about them.

And in another thread he said his next build will be AMD.

I think that speaks volumes.

And again, I could care less about x87 SuperPi benches on an ES sample.

JF-AMD has already gone on record as stating single threaded performance INCREASES.

So, there's logically only 2 options here:

1. Those benches are crap\BS AND\OR there's something wrong with the ES they tested,
2. Those leaked benches are correct and JF-AMD and MovieMan are lying.

Seriously, how can you honestly believe it's #2? Which option has the higher probability?

Which means nothing in this context. They could have a bit higher IPC and just clock it 500MHz higher and it'd have higher single-threaded performance.

I'll let you in on a hint: AMD is putting so many cores and such high clock speeds because they NEED to.

 

[Vesku]

That is why I'm glad to hear it has a lot of headroom on quality air cooling. Should be quite the price-performance ratio if I don't get a bum chip. I'm 1 for 3 with AMDs current CPUs. Have a really nice Athlon II x3 that unlocked but my emergency Sempron testing CPU's 2nd core is very unstable and my Phenom II x2 will unlock to x3 but while stable on my Asrock DDR2 board is finicky on my Gigabyte DDR3 board.

Hopefully all these gloomy rumors have made AMD decide to slip in a lot more 8150 validated chips as 8120s, hehe.

 

[CHADBOGA]

That rant reminds me of this: http://www.youtube.com/watch?v=kHmvkRoEowc

That guy was attacking AMD, not AMD's detractors.

The "Leave Britney Alone" video applies to AMDroids who demand that no criticism of AMD ever be uttered, like those fools on AMDZone.

 

[NostaSeronx]

http://phx.corporate-ir.net/phoenix.zhtml?c=74093&p=irol-newsArticle&ID=1611144&highlight=

AMD Announces Preliminary Third Quarter ResultsSUNNYVALE, CA, Sep 28, 2011 (MARKETWIRE via COMTEX) --
AMD (NYSE: AMD) today announced that revenue for the third quarter ending Oct. 1, 2011 is expected to increase four to six percent as compared to the second quarter of 2011. The company previously forecasted third quarter 2011 revenue to increase 10 percent, plus or minus two percent, from the second quarter of 2011.
In addition, AMD expects third quarter gross margin to be approximately 44 to 45 percent. The company previously forecasted third quarter 2011 gross margin to be approximately 47 percent.
The less-than-forecasted preliminary third quarter 2011 revenue results are primarily due to 32 nanometer (nm) yield, ramp and manufacturing issues at GLOBALFOUNDRIES in its Dresden, Germany factory that limited supply of "Llano". Additionally, 45nm supply was less than expected due to complexities related to the use of common tools across both technology nodes. AMD continues to work closely with its key partner GLOBALFOUNDRIES to improve 32nm yield performance in order to satisfy strong demand for AMD products.
The less-than-forecasted preliminary third quarter 2011 gross margin results are primarily due to less-than-expected supply of "Llano" and associated products with higher average selling price (ASP). Additionally, shipments of AMD's next-generation server processor, codenamed "Interlagos", occurred later in the third quarter than originally anticipated.
AMD will report third quarter 2011 results after market close on Thursday, Oct. 27, 2011. AMD will hold a conference call for the financial community at 2:00 p.m. PT (5:00 p.m. ET) that day to discuss third quarter financial results and to provide information regarding expected fourth quarter results. AMD will provide a real-time audio broadcast of the teleconference on the Investor Relations page at http://www.amd.com. The webcast will be available for 10 days after the conference call.

 

[LOL_Wut_Axel]

That is why I'm glad to hear it has a lot of headroom on quality air cooling. Should be quite the price-performance ratio if I don't get a bum chip. I'm 1 for 3 with AMDs current CPUs. Have a really nice Athlon II x3 that unlocked but my emergency Sempron testing CPU's 2nd core is very unstable and my Phenom II x2 will unlock to x3 but while stable on my Asrock DDR2 board is finicky on my Gigabyte DDR3 board.

Hopefully all these gloomy rumors have made AMD decide to slip in a lot more 8150 validated chips as 8120s, hehe.

The Athlon II X3 isn't all that hot now that the new Celeron Dual-Cores are out. They have 15-20&#37; higher single-threaded performance and only lose by ~10-15% in multi-threaded. That, and they're more efficient and cost less. Sure, they can't be overclocked, but why would you want an Athlon II X3 that will consume more power than a Core i7?

The Phenom II X2 I'll simply say is the dumbest CPU to have come from AMD since the original Phenom. Unlocking is a hit-or-miss, and if it's a miss, you end up with a measly dual-core that gets trashed by the $57 Celeron G530 in both single and multi-threaded. The Phenom II X2 needs to be priced at $55 to make sense.

 

[Black96ws6]

Here's another interesting thought:

If Bulldozer is a good chip, and it has 8 cores and overclocks up to or even past 5Ghz on air...

I wonder how far you could push the chip if you disabled 4 cores, or even 6, for those legacy apps that don't use them?

Interesting days ahead indeed! Mid-October is really only a couple weeks away and all these speculations and rumors will finally be put to rest

 

[Idontcare]

The less-than-forecasted preliminary third quarter 2011 gross margin results are primarily due to less-than-expected supply of "Llano" and associated products with higher average selling price (ASP). Additionally, shipments of AMD's next-generation server processor, codenamed "Interlagos", occurred later in the third quarter than originally anticipated.

Sadly, Motorola experienced similar proofs of the bountiful "synergies" that were to come from unlocking shareholder value by spinning off their fabs too.

Sanders, who definitely looks to be old enough to be in his grave, is surely rolling in his bed tonight.

 

[grimpr]

That guy was attacking AMD, not AMD's detractors.

The "Leave Britney Alone" video applies to AMDroids who demand that no criticism of AMD ever be uttered, like those fools on AMDZone.

Go on Champ, let it all out.