AMD CPU speculation... and expert conjecture

[szatkus]

@juan

You idiot. That cosmology ES is not kaveri. The only way to get the results of your system on the cosmology website is to run it through their program. Anyone on here can verify what I have said. You on the other hand insist on trying to insult me instead of pulling your head out of your back side and face the truth. That retarded "blog" took an 8 core benchmark, divided by 8 (the number of cores in 8xxx) then calculated the clock speed then multiplied by 4 to come up with their fake ass BD numbers.

they cut the BD and PD numbers in half in order to try and justify posting anything.

Don't try to tell me I'm clueless when you cant even offer anything other than your opinion.

http://www.cosmologyathome.org/show_host_detail.php?hostid=199535

48 core intel scoring 8181 integer pts.

4 core "kaveri" scoring 3311.

http://cosmologyathome.org/show_host_detail.php?hostid=187215

that means in your world if cosmology is a multicore benchmark, at 48 cores and 1.8 ghz kaveri will score 39,000+ or 4 times faster than a 2.7 ghz xeon. But according to your favorite blog kaveri is only 32% faster than PD not 4x xeon.

ITS FAKE, WHAT DONT YOU GET ABOUT THE TRUTH.

If I'm a liar. Prove it. Download cosmology and run your own tests.

http://www.cosmologyathome.org/

lets see your results, quit making up bs.

BOINC benchmark is old, single core Whetstone and Dhrystone. It's not interesting at all.

15-Dec-2013 22:49:46 [---] Running CPU benchmarks
15-Dec-2013 22:49:46 [---] Suspending computation - CPU benchmarks in progress
15-Dec-2013 22:50:17 [---] Benchmark results:
15-Dec-2013 22:50:17 [---] Number of CPUs: 4
15-Dec-2013 22:50:17 [---] 805 floating point MIPS (Whetstone) per CPU
15-Dec-2013 22:50:17 [---] 1625 integer MIPS (Dhrystone) per CPU

 

[noob2222]

^^ the point wasn't bionic. It is the fake BD & PD numbers here that have been posted over and over and over.

http://amdfx.blogspot.com/2013/10/amd-kaveri-benchmarks-more-steamrolling.html?m=1

 

[szatkus]

^^ the point wasn't bionic. It is the fake BD & PD numbers here that have been posted over and over and over.

http://amdfx.blogspot.com/2013/10/amd-kaveri-benchmarks-more-steamrolling.html?m=1

Cosmology@home like almost all similar projects is based on BOINC.

http://www.cosmologyathome.org/ (images on the bottom)

It's not fake, it's just a crappy benchmark.

 

[noob2222]

My 8120 bd scored 5000+ at 2.1 ghz. No way its 2300 at 1.8 ghz.

Bionic isnt fake. That blog is.

 

[szatkus]

So better let's talk about more modern benchmarks.

http://postimg.org/image/6aercz557/

You can compare my Llano with Kaveri results (ofc ignore OpenGL score).

 

[noob2222]

^^ Few more to add with those. http://forums.pureoverclock.com/video-card-overclocking/22523-post-your-cinebench-r15-scores.html

I5 2500k @4.5 ghz scores 585
8350@4.5 ghz scores 725
6300@4.6 ghz scores 558. here
6800k @5.0 ghz scores 400.

Kaveri @ 4.9 ghz scores 547.

Cinebench doesnt care about l3 cache so its good for promoting APUs as it doesnt show the weak point of going without.

Even so, kaveri doesnt look like a 2500k or a 63xx on the cpu side.

 

[palladin9479]

Seems like this happens several times before a new product is released you have over hype and then disappointment. We had someone from the BD saying 50% more performance compared to the X6, Now we have someone stating as fast as a 2500K even though a 2500K is faster by 35% on average and even using upto statements almost twice as fast as a 5800K(which is a good 7% slower compared to a 6800K)

It's people's intentions to double down on whatever strategy they were adopting when that strategy doesn't seem like it's winning. That's how fanatics are made and why I always take a balanced approach and caveat everything with "based on current information available and historical record".

Something people need to remember is that Kaveri was geared towards low power consumption more then raw power. I actually expect them to have the same general purpose performance as current APU's with a slightly better graphics engine. They are clocked lower so most of their gains are in efficiency which is a good thing should AMD port those changes to a process that can be scaled up.

 

[-Fran-]

^^ Few more to add with those. http://forums.pureoverclock.com/video-card-overclocking/22523-post-your-cinebench-r15-scores.html

I5 2500k @4.5 ghz scores 585
8350@4.5 ghz scores 725
6300@4.6 ghz scores 558. here
6800k @5.0 ghz scores 400.

Kaveri @ 4.9 ghz scores 547.

Cinebench doesnt care about l3 cache so its good for promoting APUs as it doesnt show the weak point of going without.

Even so, kaveri doesnt look like a 2500k or a 63xx on the cpu side.

I don't really like/trust Cinebench, but here are my own results:

Cheers!

 

[FALC0N]

I disagree, Palladin9475. I'm expecting 10% CPU improvement over the same model with moderate GPU improvement.

It would be shocking developement if the new APU can match the 2500k on the CPU side. High performance parts are not even the target at this point. I can't believe so many seem to be arguing over that point.

 

[jdwii]

^ well if you magically add numbers to a benchmark anything is possible just ask juanrga(sorry mods i had to insult him once after he did it several times with no one caring)

 

[FALC0N]

Friendly jabs are not insults. I think your safe.

 

[szatkus]

Can someone benchmark some 2M Piledriver CPU (Trinity, Vishera, Richland, whatever) with single core test?

 

[juanrga]

Ranth I agree with the idea of keeping expectations low. This is part of correspondence that I had when preparing my article about Kaveri/Steamroller:

your drafted article is very interesting and well written. It has a good level of detail and has a neutral POV regarding rumours and speculations based on them.

There will be other changes in SR besides the doubled decoders. E.g. the 3 FPU pipes (might allow for higher clocks or lower power -> less scheduler logic, less R/W ports in the register file parts), fetch buffer, int scheduler improvements, mem controller etc.

there are improvements throughout the module, from fetch, to decode buffer acting as a loop buffer, branch prediction, schedulers, EX/AGU units, L/S, and so forth.

The FPU might be more efficient, as do the decoders if the loop buffer gets used. Even branch prediction saves power by wasting less for wrong code paths.

There are many possibilities, I prefer to assume a 20% improvement and the final thing was faster than assuming that will be ~40% faster (35% from module + 5% from memory subsystem) and get disappointed. Bulldozer lessons ;-)

Your way to be more conservative is better I think. We both know the internet forums and the way of many posters to create and tout opinions.

tourist thanks by the comment, but please note that I considered 20% from improvements to the module architecture, but I already assumed Kaveri would have lower clocks than Richland. The abstract of my article, quoted above, clearly says "about 17% faster than top Richland APU"; it doesn't say 20%.

 

[juanrga]

jdwii, You continue making me laugh by quoting biased and fake benchmarks and pretending that they are "reliable". This is part of the FTC accusation:

59. To the public, OEMs, ISVs, and benchmarking organizations, the slower performance of non-Intel CPUs on Intel-compiled software applications appeared to be caused by the non-Intel CPUs rather than the Intel software. Intel failed to disclose the effects of the changes it made to its software in or about 2003 and later to its customers or the public. Intel also disseminated false or misleading documentation about its compiler and libraries. Intel represented to ISVs, OEMs, benchmarking organizations, and the public that programs inherently performed better on Intel CPUs than on competing CPUs. In truth and in fact, many differences were due largely or entirely to the Intel software. Intel’s misleading or false statements and omissions about the performance of its software were material to ISVs, OEMs, benchmarking organizations, and the public in their purchase or use of CPUs. Therefore, Intel’s representations that programs inherently performed better on Intel CPUs than on competing CPUs were, and are, false or misleading. Intel’s failure to disclose that the differences were due largely to the Intel software, in light of the representations made, was, and is, a deceptive practice. Moreover, those misrepresentations and omissions were likely to harm the reputation of other x86 CPUs companies, and harmed competition.

60. Some ISVs requested information from Intel concerning the apparent variation in performance of identical software run on Intel and non-Intel CPUs. In response to such requests, on numerous occasions, Intel misrepresented, expressly or by implication, the source of the problem and whether it could be solved.

61. Intel’s software design changes slowed the performance of non-Intel x86 CPUs and had no
sufficiently justifiable technological benefit. Intel’s deceptive conduct deprived consumers of an
informed choice between Intel chips and rival chips, and between Intel software and rival software, and raised rivals’ costs of competing in the relevant CPU markets. The loss of performance caused by the Intel compiler and libraries also directly harmed consumers that used non-Intel x86 CPUs.

64. Several benchmarking organizations adopted benchmarks that measured performance of CPUs running software programs compiled using the Intel compiler or libraries. Intel’s deception affected among others, the Business Applications Performance Corporation (“BAPCo”), Cinebench, and TPC benchmarks.

67. Intel publicized the results of the benchmarking to promote sales of products containing its
x86 CPUs even though it knew the benchmarks were misleading.

In some cases the Intel scores are inflated by a 10-15%, in some special cases they are inflated by up to a 30%.

I also note that you continue lying about what I say. You are been lying about how the numbers in my Kaveri article were obtained. I corrected you: "jdwii, I didn't add 30% to each benchmark". You continued lying. I gave you the links to my article and a copy of the abstract. You continue lying about my work after my correction with your recent "Also again i'm being realistic(not magically adding 30% to everything)".

noob222, excellent work of cherry picking benchmarks, besides you taking different benchmarks, different versions and comparing numbers directly :lol:

One of your links gives two JTR benchmarks. In one of them the A10 perform very well, in the other performs bad. Which did you select? Of course, the poor one. What is more interesting is that if we look to the sigma of the poor benchmark we found abnormally large values (about 4x larger than usual). This is clearly indicating that that piece of silicon had some problem in that particular test. The chip could be defective, the hardware could be incorrectly configured, some compiler regression in the flags selected, who knows.... But we clearly see abnormally low value and abnormally large sigma.

Not only you selected the poor case of two available in your own link, but the one that you selected contains abnormal values indicating some problem in that test.

This is your same tactic, when you linked a game review and selected the poor possible benchmark for AMD of five available in the same link. Or you recent attempt to try to discredit the same source (AMD blog) that you initially used to attack AMD, when was noted that the source said the contrary that yo pretended.

It is pretty clever than both of you have a hidden agenda.

 

[gamerk316]

Palladin, Yuka, Gamerk, 8350rocks and others on the thread who tried out my benchmark app, I have built a new version with an EXE launcher for windows. It lets me control the command line and ensure the app runs at high priority by default. Also, some of your CPUs were not identified correctly - this version (on windows of course) takes a second shot at the CPU name by reading the registry.

http://www.headline-benchmark.com/download.jsp

I'll give it a try some time tonight.

Wonder how my work PC would do, but I'd rather not feel the wrath of the IT/Security department...

 

[etayorius]

I only seen a 30-35% max in PC Games that heavily rely on Single Threaded Performance between comparable CPUs from AMD and Intel:

http://techreport.com/review/23750/amd-fx-8350-processor-reviewed/5


http://www.pcper.com/reviews/Processors/AMD-FX-8350-and-FX-6300-Processor-Review-Vishera-Breaks-Cover/Results-Skyrim-and-

and the higher the graphics and settings, the less difference there will be.

 

[szatkus]

I'm sorry, ~33% worse = ~50% better.

 

[gamerk316]

ICC again? Seriously? NO ONE USES IT.

Seriously, last time we had this debate, I showed every game in my game library that I had was compiled via MSVC (except one, which was GCC). The game industry does not use ICC AT ALL.

Though if you want, I'll run every game in my library against my compiler detection suite, and I'll show, again, no one uses ICC.

 

[noob2222]

@juan

Nice tactic of trying to defend your calculations by explaining nothing.

So from what I get, you took the intel numbers from the "poor" benchmark and made up your own kaveri numbers based on the other chart?

If you got your information based on certain testing, try posting it instead of trying again and again to simply say "im never wrong, my name juanrga translates to I am never wrong."

Show us all where richland is just as fast as the i5 3470. Show us where your best case benchmarks are.

This is a perfect opportunity to prove me wrong. Let see if you can do it.

 

[jdwii]

I only seen a 30-35% max in PC Games that heavily rely on Single Threaded Performance between comparable CPUs from AMD and Intel:

http://techreport.com/review/23750/amd-fx-8350-processor-reviewed/5


http://www.pcper.com/reviews/Processors/AMD-FX-8350-and-FX-6300-Processor-Review-Vishera-Breaks-Cover/Results-Skyrim-and-

and the higher the graphics and settings, the less difference there will be.

@juan

Nice tactic of trying to defend your calculations by explaining nothing.

So from what I get, you took the intel numbers from the "poor" benchmark and made up your own kaveri numbers based on the other chart?

If you got your information based on certain testing, try posting it instead of trying again and again to simply say "im never wrong, my name juanrga translates to I am never wrong."

Show us all where richland is just as fast as the i5 3470. Show us where your best case benchmarks are.

Agree, I'm a Amd fan like everyone here or at least a fan at improving performance but its like he is either delusional or he has a hidden agenda.

If you compare a A10 6800K to a I5 2500K their is a big difference in CPU performance which is not Amd's fault since the A10 was 55-65% cheaper and included a Much better IGPU but were only talking about CPU performance here.
He continues to say the next APU outperforms the I5 2500K and for this to happen you would need a 35-45% boost in CPU performance on average at times it could probably even beat the chip but the majority of the time no way(then again the product is not even out yet i thought we where supposed to have a paper launch)

 

[gamerk316]

...Am I the only one here getting a strong sense of deja-vu?

The [strike]Bulldozer[/strike] Kaveri rumor thread, with Gamerk2 playing himself, noob2222 playing himself, and juanrga playing the star role of JF-AMD.

 

[FALC0N]

Maybe gamerk360, But at least there was good reason with bulldozer. A new, promising architecture, the return of the FX brand, the announced pricing, all pointed to a quality part that never materialized.

In this case, all signs point to a quality part, but there is no indicate that these parts will generate the 30% speed boost needed to catch the 2500k.

We might see a big IPC improvement, but the clocks are lower and that will offset the IPC gains somewhat.

 

[gamerk316]

And I stood here and made the same, exact, arguments, that I am still making about why FX wouldn't be as good as everyone was expecting. Software always lags hardware, and if AMD's master plan is to wait for software to catch up in order to make up CPU performance, I am sorry to say you are going to be very disappointed. And that's before I get into the more architectural arguments.

 

[de5_Roy]

...Am I the only one here getting a strong sense of deja-vu?

The [strike]Bulldozer[/strike] Kaveri rumor thread, with Gamerk2 playing himself, noob2222 playing himself, and juanrga playing the star role of JF-AMD.

SERIOUSLY?!?!? jf-amd's knowledge seemed far more than juanrga's. you're actually highly complimenting juanrga by comparing him to a guy who knows infinite times more than juanrga will ever learn. ever. times infinity. times infinity.
i actually liked reading jf's posts even though i wasn't here at that time. despite what became of him, with people channeling their frustration at his direction. to me the guy looks like he was forced to take a fall. there were plenty of other insanity in the bd thread. i'd still not compare him to someone like juanrga.
this as far as i'll talk about this.

since i am posting already, might as well speculate on the 4.9gig kaveri cb performance. it's score was close to fx6300, a 6 core cpu with l3 cache. the fx will likely be cheaper when a10 7850k comes out, but it'll have the added advantage of a gcn igpu. now it remains to be seen how "typical" that 4.9ghz o.c. turns out to be, and at what voltage and temperature. the bench must have been done with a discreet gfx card.
the benchmarks i wanna see are sandra memory bw, 7zip, winrar open cl, handbreak (cpu, igpu, open cl cpu, open cl igpu), tomb raider @medium @1080p (and with tress fx).

p.s. gamerk - at first i read the post as "gamerk2 playing with himself, ... j...<snipped>" the deja vu i got was that the kind of insanity, delusion and attempts to dispelling such delusion and overhype before real data emerges and subsequent crawling back of people who engage in spreading misinformation.

 

[juanrga]

Szatkus, what games? Those that appear in the well-known "optimized for Intel" list that Intel publicities in its own site? Those that always appear benchmarked in sites known by their bias? Maybe you mean those on purpose tested at low 720p at low settings?

Or are you trying to say me that at 1080p all games will run 50% better in a 2500k than in a A10-6800k? Because I know average difference can be lower than 10%.

Maybe you didn't notice, or maybe you jare just ignoring it, but it was already explained here which will be the expected gaming performance of Kaveri under (i) old games, (ii) modern games, and (iii) MANTLE enabled games.

Noob222, there are nothing can be explained to you or to jdwii. A month after the work got published and lots of explanations and corrections posted here and both you are still pretending that I added 30% to all the benchmarks. LOL