AMD CPU speculation... and expert conjecture

[logainofhades]

I would take PClab's results with a grain of salt. They have been known to be quite the intel fanboy website.

that can be said for any website that shows, correctly or otherwise, intel cpus ahead of amd cpus.

We have had members here from Poland speak out about that particular site, hence my comments.

 

[de5_Roy]

We have had members here from Poland speak out about that particular site, hence my comments.

deja vu ..... i remember someone saying the exact same thing about pclab before. iirc the counter-argument was that american member calling thg u.s. biased, german members calling the thg german site biased, brits calling kitguru biased and so on.
a solid proof of bias would be noticeable, verifiable discrepancy in the benchmarking, regardless of nationality.

 

[gamerk316]

"someone said on the internet"

 

[juggernautxtr]

THE 965 is also clocked @4Ghz, THE FX 6300 out does it, the stock clocked 6350 out shines it,
the way i look at the fx line
8 core = 4 core w/ hyperthread
6core = 3 core w/hyperthread
4core = 2 core w/ hyperthread
even the kaveri 7850 does better when clocked up, and it's a low end desktop part.
the stock clocked 8350 pisses on the phII
goes to show that beating a dead horse(phenomII) won't make it faster.

 

[logainofhades]

750k loses to PhII clock for clock. Kaveri upped the IPC enough that is more or less can match an PhII now. Anyone with a PhII would be foolish to buy Kaveri, period. If you have a PhII X4, and are looking for a new upgrade, you go and get an FX or go and get an i5.

 

[Cazalan]

So benching should ideally be done on Windows 8, to get a fair comparison.

And yet HWBot blocks submissions from Win8 due to a RTC bug.

"Due to severe validity problems with the Windows8 real time clock (“RTC”), not all benchmarks results achieved with Windows8 can be trusted. The main problem lies with the RTC being affected when over- or underclocking under the operating system. The operating system uses the RTC as reference clock, and benchmarks use it to reference (benchmark) time. Refer to our original coverage article for more detailed information. The restrictions for Windows 8 due to the RTC bug can be found below."

http://hwbot.org/article/general_rules

Exit: One of these days the quoting system will be fixed.

 

[jdwii]

You can see on my above where the GPU bottleneck comes into play, at about 80ish FPS or so. Note how the i3 is now competitive with with the OC'd 4770k? Core i3 is the big winner here.

Big loser of Kaveri, which still falls behind the Phenom II X4.

Yeah i see that thanks for posting i would of never thought that was going to be the outcome. I still see know reason for someone to "upgrade" to Kaveri if they have a Phenom II x4 even more so if they have it clocked at 3.8+ghz.

 

[Cazalan]

So it seems Mantle does exactly what it set out to do. Devs requested an API closer to the metal and they got it. Even to the detriment of AMD CPUs in many cases. Quite impressive actually. Job well done to AMD!

 

[jdwii]

THE 965 is also clocked @4Ghz, THE FX 6300 out does it, the stock clocked 6350 out shines it,
the way i look at the fx line
8 core = 4 core w/ hyperthread
6core = 3 core w/hyperthread
4core = 2 core w/ hyperthread
even the kaveri 7850 does better when clocked up, and it's a low end desktop part.
the stock clocked 8350 pisses on the phII
goes to show that beating a dead horse(phenomII) won't make it faster.

25% more performance with twice the cores? Comparing the 965 to the 8350fx at the same clock speed.

 

[Embra]

HardOCP:

http://www.hardocp.com/article/2014/02/03/amd_mantle_performance_preview_in_battlefield_4

 

[gamerk316]

25% more performance with twice the cores? Comparing the 965 to the 8350fx at the same clock speed.

And we're back to why I favor IPC over more cores. Hence why the i3's are still hanging around.

 

[juggernautxtr]

sorry for not quoting but you also have to account that the fx line has half the core size per core.
the phenom integer core was split in 2 to fit the fx line.
one phenom core = 2 fx cores

 

[blackkstar]

http://www.hardocp.com/image.html?image=MTM5MTM5Nzk5NlFVelAwM051VWxfM18yX2wuZ2lm

http://www.hardocp.com/article/2014/02/03/amd_mantle_performance_preview_in_battlefield_4/3

So, when did frame times not become an issue? That's massive. It's showing like a 5% fps increase but the frame times are ridiculously better.

And apparently no one read my long post, so here's the tl;dr. AMD removed the CPU bottleneck to encourage more people to buy AMD GPUs when they don't have solid CPUs.

With Mantle, 280x with i3 or FX 6300 or an APU doesn't seem so stupid anymore.

AMD is trying to sell graphics cards with Mantle, not CPUs. If you have a balanced system with a quad core Intel, you're (probably) not looking to upgrade your graphics card because your rig is going to play games fine.

If you've got an older i3 or Ph2 or lower end FX or APU, you now have some sort of incentive to buy a new AMD graphics card. Incentive that you would have never had with DirectX games on a lower end CPU.

It was a really freaking awesome move by AMD. Give a lot of people free performance, offer the standard up to be open, and give people the opportunity for better graphics without having to update their entire rig. If I were stuck on 2m/4c AMD CPU, PH2, or i3, I'd be ecstatic about this because it'd mean I could just upgrade my graphics card and be content with Mantle performance, and then later on update the CPU + motherboard and keep the graphics card, making a slow and steady upgrade path that lets me spend a few hundred dollars a year for a solid PC.

But again, the purpose isn't "wow free performance!" That's just a side effect. The true purpose is "wow my low end CPU I spent $100 on can now be paired with a higher end graphics card."

If AMD can get people with mid-range CPUs to upgrade their GPUs, Mantle has won. And it's put AMD as the sole company in that entire market, as buying higher end Nvidia with an i3 or 2m/4c AMD or Ph2 is going to bottleneck a lot.

People with 3570k or better with GTX 670/7970 or better aren't going to get much out of Mantle (except for intel hex with crossfire), but those people (probably) aren't going to upgrade anytime soon regardless.

Another side effect is that it has completely stopped the bloodshed happening with AMD HEDT platform. FX 8000 series has come a long way in benchmarks from the launch skyrim/starcraft/shogun AMD deathcombo of benchmarks, and Mantle only makes them look better.

I can guarantee you that if you took the average redditor level enthusiast and swapped the names of FX 8350 with launch benchmarks and recent ones so that the new benches had FX 8350 and the old one had some other name (like FX 8150) that people would have no problem believing they were completely different CPUs.

 

[jdwii]

sorry for not quoting but you also have to account that the fx line has half the core size per core.
the phenom integer core was split in 2 to fit the fx line.
one phenom core = 2 fx cores

Not sure if that is 100% true, 1 module in steamroller if built on 45nm would probably be around 1.8 the size of 2 phenom cores. http://www.anandtech.com/show/4955/the-bulldozer-review-amd-fx8150-tested

Actually if we take that number and divide it by the amount of cores we would have 150M transistors for both, the reason for the smaller die size is because of the Manufacturing Process. Also keep in mind steamroller is probably a little bigger now.

 

[Cazalan]

sorry for not quoting but you also have to account that the fx line has half the core size per core.
the phenom integer core was split in 2 to fit the fx line.
one phenom core = 2 fx cores

Not half, but 2/3. FX cores have 2 ALU, Phenom cores have 3 ALU.

 

[juanrga]

The only way to get the big gains with mantle is to build an unbalances system with a weak cpu and an overpowered gpu. Mantle with a balanced system is single digit gains.

No. It was shown immediately above that one can obtain big gains (58%) with an eXtreme i7 chip more a pair of 290X. An eXtreme i7 is not a "weak cpu" but one of the more stronger x86 cpus. Also what you consider "unbalances system" is what Nvidia call a battlebox machine: a top-end gaming PC where the pair of 290X is replaced by two/three Titan/780.

Here more benchmarks of MANTLE vs DX

This is for an i7-3770k plus OC 290X. Again the 3770k is not a "weak cpu". This graphic shows clearly how MANTLE eliminates CPU bottlenecks.

The very convenient fact that you left out is that this is on 1080p, and I can't justify getting a R9-290 for 1080p, let alone two, and we all know a lower resolution makes a game more CPU bound.

What part of MANTLE eliminates the CPU bottleneck generated by DX the overhead is not still understood? If you are in a GPU bottleneck situation, then MANTLE will brings you from small to zero performance gain. "CPU bottleneck" is not synonym for low-end CPU, because one can have a CPU bottleneck with a high-end $1000 CPU as well, it just depends of the situation.

I know people who play games at 1080p, reviews that test the 290x at 1080p, and reviews of crossfire of pair of 290x at 1080p.

At 1080p you can see big gains of up to 60% even with high-end CPUs (3770k, 4960x,...). At 2160p you can see small single digit gains. But not everyone is playing at 4K resolution, true? If you play at resolutions between 1080p and 2160p you will find intermediate results between 7% and 60%.

Another important point that is being ignored here is that MANTLE will provide bigger benefits with future games that make more draw calls. Current games make a limited number of draw calls due to DX overhead. The Oxide demo uses a larger number of draw calls to show what future games could achieve with MANTLE: 2--3x more performance.

 

[noob2222]

The oxide demo is only designed to force a cpu bottleneck. Aka its designed to show what mantle can do, aka synthetic banchmark.

Other than that, your arguement is only valid in that one map. Even at that 1080p is 27% not 60.

 

[Master-flaw]

Another important point that is being ignored here is that MANTLE will provide bigger benefits with future games that make more draw calls. Current games make a limited number of draw calls due to DX overhead. The Oxide demo uses a larger number of draw calls to show what future games could achieve with MANTLE: 2--3x more performance.

Basically my point earlier...kinda strange that little to no discussion is gear towards this with a group of people that should be able to put 2 and 2 together.
People are failing to mention that this doesn't optimize low CPU's only. It shows it in the Star Swarm demo quite obviously. I received almost a 200% gain in performance with my setup.
People are using it's impact on BF 4 a handful of benches to define it's uses.

 

[juanrga]

Now that I've given more thoughts to MANTLE and since one of you pointed out, but it seems no one noticed...

MIN FRAME RATES!

What happened with those? We can all agree that if we see an avg increase of 7% (for example), it will be amazing if the min frame rates increase in a significant amount. In other words, we can forget (a little) about averages and start asking about minimum ones, right?

So, given that, any numbers on those?

Cheers!

Beta testing (although reviewers write "Compared to other "beta" driver releases from AMD, this seems much more like an alpha") and show some inconsistencies, but at least some data with minimum FPS for 7850k + 290x

http://www.pcper.com/reviews/Graphics-Cards/Battlefield-4-Mantle-Early-Performance-Testing/More-Mantle-Results

The best result they got was

 

[juanrga]

pclab has some bf4 mantle cpu benches. look at the haswell pentium and core i3 go after 7850k and fx8350!
http://pclab.pl/art55953.html
i wish i could translate this... i'll read it after i fix my browser.

^ 7% @ 4000
"In the graphics limit the 4K resolution (3,840 x 2,160 pixels) overloaded the Mantle API the Radeon R9 290X little better, the average frame rate increases by 7 percent."

http://www.golem.de/news/amds-mantle-api-im-test-der-prozessor-katalysator-1402-104261-3.html

might mean that mantle won't help in a typical cpu-bound situation like online mp shooters if the graphics subsystem becomes bottleneck, no matter how high end the gfx card is. you must lessen the load on the gfx hardware for mantle to deliver.

I would take PClab's results with a grain of salt. They have been known to be quite the intel fanboy website.

Sounds like Mantle will be good for the budget crowd, at the very least. That I see is a good thing. The more PC gamers there are the better.

yes, we know from inside they invent scores, but pclab benchmarks are always quoted by the same people in this thread.

So benching should ideally be done on Windows 8, to get a fair comparison.

And yet HWBot blocks submissions from Win8 due to a RTC bug.

"Due to severe validity problems with the Windows8 real time clock (“RTC”), not all benchmarks results achieved with Windows8 can be trusted. The main problem lies with the RTC being affected when over- or underclocking under the operating system. The operating system uses the RTC as reference clock, and benchmarks use it to reference (benchmark) time. Refer to our original coverage article for more detailed information. The restrictions for Windows 8 due to the RTC bug can be found below."

http://hwbot.org/article/general_rules

The above quote is from gamerk, not from mine!!!

 

[blackkstar]

The oxide demo is only designed to force a cpu bottleneck. Aka its designed to show what mantle can do, aka synthetic banchmark.

Other than that, your arguement is only valid in that one map. Even at that 1080p is 27% not 60.

Oh no 27% free performance gain for installing different drivers and changing game settings in a supported game. AMD IS PURE EVIL!!!

The lengths some of you go. This reminds me of when Titan fell to GTX and 290x and every Titan owner was out defending their overpriced, inferior card to the death.

I realize that Mantle is going to make everyone who went Intel quad + mid-range gaming card (like 7870) instead of something like FX 8320 + 7950 cry like babies but just stop already.

Mantle is causing an earthquake with every curmudgeon that wanked to Skyrim and Starcraft benchmarks and went "LOL AMD SUX XDDDD", because they can no longer walk around going "LOL AMD CPU SUX FOR GAMING XDDDD NICE SKYRIM BENCHMARK SCORES LOLOLOLO"

What's next from some of you? Are you going to attack free-sync and defend g-sync?

If it really bothers you and you're all so dead set on cherry picking situations where Mantle isn't that good, then just go buy an Intel quad with an Nvidia card and play Skyrim and Starcraft 2 for the rest of your lives.

People said pages and pages ago that AMD would capitalize on their console wins, and some of you went "LOL NO ONE CARES LOL 8 CORE JAGUAR IS IMPOSSIBLE FOR MULTITHREAED GAEMZZZ" and now we have a huge bottleneck being removed and the same sort of people are here going "yeah that's cool you found an example where Mantle is twice as fast but here is an example where it's only 27% faster so it's complete ***** useless vaporware dead technology ***** you AMD"

Some of you really need to learn how the hell to use logic. People making the claim that Mantle can give huge performance gains and then going "yeah there's some huge performance gains but it's not always a huge gain" isn't disproving anything.

You might as well go up to someone who claims to work 40 hours a week, see they have a week where they don't work, and then go "LOL U DONT WORK YOU LAZY BUM", because that's basically what you're doing.

 

[etayorius]

Ill even take 5% Free performance thanks to mantle.

27% is just impressive and up to 45% is pure win.

 

[Lessthannil]

There isnt much illogic (if any) in pointing out that most of the big precentage gains comes from a woefully imbalanced system and that pointing out that x setup is more reasonable and common with it showing lower games.

Nobody is saying that a 45% gain or such is bad. We are saying that those situations are rare (and for good reason because what happens when a game doesnt use Mantle?)

 

[palladin9479]

There isnt much illogic (if any) in pointing out that most of the big precentage gains comes from a woefully imbalanced system and that pointing out that x setup is more reasonable and common with it showing lower games.

Nobody is saying that a 45% gain or such is bad. We are saying that those situations are rare (and for good reason because what happens when a game doesnt use Mantle?)

When hardware assisted rasterizers were first introduced to the consumer landscape the exact same comments were made. The programs that took advantage of them were too few, and that software rendering would work on far more programs. A few decades later hardware assisted rasterization is in everything, everything is using them and software rendering is unheard of outside of ray tracing for CGI environments. And even that is being accelerated via large vector coprocessors.

I expect the current iteration of Mantle to be like MMX / 3DNOW / Glide. Useful in some situations, especially the low cost ones, but not widely adopted. In the next two years we'll see several updates to the API and further refinements as developers start working with it and providing more feedback to what they want and what they need. Eventually it'll be replaced by another version of itself or by another API that does the same thing but better.

In either case I'm really interested in what they can get from a A8-7600 with 2133 memory. Everyone keeps comparing it with the monstrously overpriced 7850K and R290x which isn't remotely realistic. You don't really want to pair APU's up with dGPUs, it push's them out of their segment. Though there not being a mini-itx AM3+ board is a shame, fx63xx + mid range dGPU would be amazing with Mantle.

 

[juggernautxtr]

All the intel optimized software out there, and we are here with mantle, listening to a bunch of whiners complain that their high end cpu+dgpu aren't gaining 45% performance boost. amd does something to help themselves out and people complain.