AMD CPU speculation... and expert conjecture

juanrga · Feb 1, 2014

gamerk316 :

Hum! When MANTLE was announced some people in forums said that it was unneeded because would provide negligible performance gain over DX: "the overhead is minimal", they said us. My reply was that devs would not be interested in MANTLE for minimal single digit performance gains and that we would expect 30--50% gains in BF4. Benchmarks are showing that MANTLE provides big gains, as expected. Now you pretend that MANTLE is not needed because Microsoft will eliminate the overhead in DX12...

Nice to know that MANTLE is useful on putting pressure on Microsoft, but the problem is that DX12, even if it fits your expectations, will be only released for Windows when developers are interested in other platforms as well. Your argument about no dev using MANTLE was answered many times before giving to you announcements of main devs already using MANTLE or planing to use it soon.

gamerk316 :

That graph represents the average of a 'GPU' test (1080p and medium settings) and a 'CPU' test (720p and low settings). The names are irrelevant, call them "A" and "B" if you want. Take BF4 benchmarks, in the 'GPU' test the gain from using 2400MHz instead 1600MHz is the double of the gain measured in the 'CPU' test, which is the expected trend because bandwidth is more needed for GPU than for CPU. In more GPU oriented tests we can see nice bumps in gaming performance by using 2133 and 2400 memory

When interpreting those graphs, it must be understood that 2133MHz is clocked 33% above 1600MHz whereas 2400MHz is clocked only 13% above 2133MHz.

Here more memory scaling benchmarks with 2400 compared to 2133 and 1866

And here you can find gaming memory scaling (800/1066/1333/1600/1866/2133) under linux, which can give an idea of how SteamOS will perform. Again 2133MHz provides nice gains and it is cheaper in my country ;-)

P.S.: Thanks to 8350rocks by the post about clocks and voltages.

noob2222 · Feb 1, 2014

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.

as well mantle introduces its own set of problems that need to be fixed.

Mantle is great at what its designed to do. Allow a cheaper system to utilize a higher end video card than it could normally handle. It wont push cpu sales because it works just as well for a low end intel system (aka pentium).

gamerk316 · Feb 1, 2014

^^ And a chart like that is what I'd expect for a balanced system, in all honesty.

I think the net effect of Mantle is going to be making i3's look that much better.

jimmysmitty · Feb 1, 2014

palladin9479 :

I guess we shall see as I am in the Steam beta for it and will test it out over a 802.11n network, which I might upgrade my TVs HTPC to 802.11ac which my router does.

And I wasn't talking about dCPU and dGPU. I was talking about the fact that the majority of HTPCs are not used for high end gaming if any at all. Mine isn't nor will it ever be so a Intel HD graphics will work just fine as I use it to watch movies and shows.

Honestly, I am not sticking up for juan here at all, but that is typical. AMD advertises what they can guarantee, there will be fluctuation, but the stock configuration will be 100% achievable on any CPU, with any hardware compatible with the design. Most PD CPUs could hit 10% overclock without bumping voltage, my FX8350 is one of numerous such examples.

Really it shows how little you know about historical trends with a given architecture.

The truth is that every company cherry picks. Look at the R9 290X review samples. They were the exact same as the retail ones but for some reason, they clocked higher and even with driver fixes, still clocked higher. Just because you got a FX 8350 that OCs well does not mean that all of them do.

I got a Core 2 Q6600 G0 that not only OCed to 3GHz with a simple FBS bump but I was also able to lower the voltage by .10v and keep it stable from its stock voltage. Most G0s would hit 3GHz on stock, some would need a slight voltage bump, very few would undervolt and OC to 3GHz.

It is all based on luck. Look at the HD7970s. Some people hit 1200MHz on stock voltages with aftermarket coolers (Vapor-X, DCUII) others couldn't pass 1100MHz.

AMD and Intel advertise their CPUs as they label them but review samples are normally always cherry picked better than retail.

wh3resmycar · Feb 1, 2014

mantle preview up on anandtech.. thoughts?

jimmysmitty · Feb 1, 2014

wh3resmycar :

So if you have a low end CPU with a extreme high end GPU, you get a decent performance boost. But if you have what the majority of gamers have, say a FX8K series or i5/i7 4K series, its so minimal that who even cares.

I do find it interesting that a A10 7700K, per AMD that is, benefits the most.

If it was 40% with any CPU, I would be pretty amazed but 40% with a APU only leaves me saying meh. I want to see higher res numbers where the GPU matters more before I make a final opinion though.

Master-flaw · Feb 1, 2014

Looks to me as though mantle was set to reduce CPU bottleneck...and it does so by 40%....it seems the more of a bottleneck the CPU is the better it works...reason being why only weaker CPU's work with it is becuase there really is no bottleneck on the CPU's that produced single digit gains(8350...ect).
I'd like to see this work in an RTS or MMO...probably bring up frames by a large margin for higher end systems.

-Fran- · Feb 1, 2014

Well, I understand MANTLEs' benefits with the "glass half full": the let you use a cheaper CPU to achieve what you would with a more expensive one. That's actually a pretty good thing. And at the same time (glass half empty? haha), it would give even less incentive to put more powerful CPUs out there, at least, for the time being.

Now, on the other hand (half full, lol), it seems that XFire get a massive speed bump. That might make dual and triple GPU systems more common.

Also... What could nVidia be thinking as an answer to MANTLE? It seems like they won't back AMD on this one (no surprise there), so will they push their own API now?

Cheers!

palladin9479 · Feb 1, 2014

Seems to be pretty good for APU's which is where I figured they would direct it the most at. I want to see what it's like on the A8-7600 as the A10-7700K is about $20 USD too high.

esrever · Feb 1, 2014

Im assuming the APUs are still bandwidth limited which is why we don't see more improvements.

palladin9479 · Feb 2, 2014

esrever :

The GPU components of APU's are bandwidth limited not the CPU component. Most of the reviews have pointed out that the iGPU components on APU's are quite strong but the CPU can't feed it after once you get to DDR3-2133 memory. They were using the 7850K paired with a 290X card and showed a massive gain in MP performance, damn near 90%. People have to realize that 99% of benchmarks are single player timed demo type runs which aren't indicative of how that setup will play in a MP environment.

So I'm expecting a 20~30% gain for the 7600/7700 line, depending on the settings.

Master-flaw · Feb 2, 2014

If it can increase the CPU overhead in the 7700 in BF 4 then I imagine it can do the same for higher end CPU's under more strenuous conditions.

DLing 14.1 now...put it through a test run on BF 4 then check to see how star swarm preforms.

de5_Roy · Feb 2, 2014

^^ you mean decrease the cpu overhead.

Master-flaw · Feb 2, 2014

Yea, spaced it...
Doesn't look like I gained much from BF 4(MP and SP)...though there are people all over the chat channel saying min frames went up from 30 to 60...
Though the Star Swarm stress test more than doubles my frames with a single GPU and almost triples them while using two on both tests...
wondering how well it would do with more CPU intensive games.

de5_Roy · Feb 2, 2014

from the initial previews and tests, entry level cpus and apus are getting good boost out of mantle. in that price segment, any perf gain is appreciated. this means that apu, fx4xxx-6xxx owners will be getting more out of their pcs without replacing it with a core i5(-which will gain even more).

mantle first impressions (compilation)
http://www.techpowerup.com/197455/amd-passes-on-catalyst-14-1-beta-to-the-press-public-release-shortly.html
http://www.guru3d.com/articles_pages/amd_mantle_preview,1.html
http://www.legitreviews.com/amd-mantle-api-real-world-bf4-benchmark-performance-catalyst-141_134959

kaveri case mods
http://community.amd.com/community/amd-blogs/amd/blog/2014/01/23/kustom-kaveri-pushing-the-limits-with-small-form-factors

noob2222 · Feb 2, 2014

^^ Until/if mantle is implemented across the board, your still building an unbalanced system 99% of the time.

AMD doesn't have the revenue to push software and its not going to be widely adopted. Sure it works, but that doesn't mean its going to sell all by itself.

Basing a decision on one game alone would be pretty ignorant, but those people do exist.

de5_Roy · Feb 2, 2014

it is here, finally
http://www.techpowerup.com/197469/amd-catalyst-14-1-beta-1-6-posted.html
none of the sweet spot cards are supported at launch. users/owners groan.

noob2222 :

that's the problem with mantle hype. if you have a pc with weak cpu, mantle will deliver in loads. the more balanced your gaming pc is, the less will mantle deliver (in one case in at's preview, it'll cause regression). worse, even if you Do build a pc with weak cpu with mantle in mind, you'll be stuck with:
-the few games it supports,
-cpu underperformance in everything else, including cpu-intensive apps (pentium, core i3 and fx4xxx come to mind),
-underused gpu resource in non-mantle games, resulting in less perf/price. this diminishes mantle's worth.
-constant software upgrades as amd continues to fix bugs and issues. the proof is in it's delayed launch and the last moment withdraw to fix another bugs. on the plus side, Never A Dull Moment.

-trolls and clueless fanboys use vague percentage figures to misrepresent mantle's usefulness. although this one isn't amd's fault at all.

noob2222 :

they don't have the revenue, but they're not pushing mantle to games, but game engines. i like that policy. with engine optimizations, any game built using the engine will get mantle support/choice by default. the real software issue is whether amd can continue releasing 80-20 driver/api updates. they failed to do that at launch, but i'll chalk it up to a rough start.

noob2222 :

i agree. amusingly, the statement reminded me of the sbm build with pentium and radeon hd 7850 aiming for 1080p gaming. in pcs like that, mantle will deliver. only in optimized games though, in everything else, there won't be any improvement (may be regressions).

juanrga · Feb 2, 2014

noob2222 :

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.

con635 · Feb 2, 2014

gamerk316 :

A chart like that is in single player, check out the 64mp benches in juans post above. Dont have bf4 yet myself but hearing real world reports of 'it just plain smoother' as well...

Lessthannil · Feb 2, 2014

juanrga :

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.

de5_Roy · Feb 2, 2014

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.

tourist :

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.

Master-flaw · Feb 2, 2014

de5_Roy :

tourist :

Slap another 290X in there and you may reduce min frames and performance halts due to CPU bottleneck. Resulting in more fluid 4K gaming that wasn't possible on BF 4 before mantle.

de5_Roy :

Yea its really as simple as CPU performance...Balanced high end systems are not going to need the performance boost. Though there are games that could've benefited from mantle due to an unbalanced engine. MMO's, RTS's and the like would probably benefit heavily from Mantle. With high end CPU's some of these games still seem to drop frames in intensive scenario's. I think people are defining it on it's impact on one game a bit to much. This has potential to be much more useful than the benefit of unbalanced systems.

juggernautxtr · Feb 2, 2014

mantle is designed to reduce cpu overhead, your going to see less as you get into higher end cpu's.
mantle still helps even up there most likely with latency.

blackkstar · Feb 2, 2014

tourist :

The purpose of Mantle is to provide people with older CPUs a reason to upgrade their GPUs by giving them better performance.

I was playing with Star Swarm Demo and I was seeing close to double performance when using Mantle over DirectX. For reference, I have a 1440p Catleap with a 5ghz FX 8350 and a 7970 at 1250/1725.

Mantle was designed to remove CPU bottlenecks. I think some of you are missing the point. Think of all the people with older CPUs who want to upgrade and everyone on the internet goes "lol ur cheapo CPU will bottleneck a 280x!" Mantle completely ends that. Which means more GPU sales for AMD.

3 R9 290xs in Crossfire are going to be bottlenecked by 4960x in DX. FX 8350 will bottleneck in games that don't scale well to many cores with a single 7970. And then you have the whole i3/Phenom/FX6350 and lower class of gaming PCs.

If Mantle takes off and is widely adopted, it's going to completely change the way you build PCs. It's actually a massive blow to Intel, because right now the chips to get are the unlocked Intel quads. With mantle, those chips are complete overkill and you can get something cheaper and put that money towards a GPU.

This is what Mantle is supposed to do. It's supposed to be a direct blow at Intel Quad Unlocked parts and it's meant to encourage users to spend that money on a cheaper CPU and a more expensive GPU.

So, of course it's not going to offer a lot to people with balanced systems with quad core Intels. That's not who it is for. It's for making PC gaming more accessible to those who can't or don't want to spend $200+ on a CPU. So APUs, older CPUs like Phenoms, lower core FXs, Core i3, etc.

This is a long range plan and it should give you some hints as to what AMD is going to do with future CPU lines. I can tell you that they didn't bother writing Mantle and making it scale to lots of cores just so they could sell 2m/4c APUs.

This probably opens up the door for AMD to push small cat cores (like Jaguar) to 3ghz or so and 16 or more cores post-excavator, and to keep it around for gaming purposes. It's the one thing ARM isn't going to be able to take over, and I don't see ARM ever taking off for PC gaming. It's difficult enough to escape WIndow's grasp, let alone x86's when it comes to gaming.

Of course, I'm assuming that Mantle is going to be used solely to increase performance. There's no reason why Mantle can't be used to offer the same performance while offering significantly better graphical quality. But at the same time there's no reason why Mantle can't do both.

I don't expect BF4 to be the gold standard of what Mantle can do. DICE can't even make a game that has basic functionality in it like respawning that works without bugs. Star Swarm is more along the lines of what Mantle should be used for, and it's more of an example of something designed around Mantle. BF4 is more of just throwing Mantle onto an existing product.

And it should be pretty obvious why Mantle is a boon to game developers. The money Oxide spent on going Mantle is basically free advertising for them. It's mentioned all over the internet. I don't know about eveyrone else, but I've never heard of Oxide or Star Swarm before this.

So really, with Mantle, you should be focusing on where it works great and the situations where it works best. No one in their right mind goes around finding single thread benchmarks where i3 is close to 4960x and then goes "lol 4960x IS TOTAL GARBAGE IT'S BARELY FASTER THAN i3 IN THIS BENCHMARK!"

Because when you go hunting for worst case scenarios with benchmarks, that's exactly the kind of thing you're doing.

truegenius · Feb 2, 2014

tourist :

Me too want to see performance boost provided by mantle in case of phenom 2, is there any review in which they used any phenom chip :??:

AMD CPU speculation... and expert conjecture

Distinguished

Distinguished

Glorious

Champion

Distinguished

Champion

Honorable

Illustrious

Splendid

Splendid

Splendid

Honorable

Splendid

Honorable

Splendid

Distinguished

Splendid

Distinguished

Honorable

Honorable

Splendid

Honorable

Honorable

Honorable

Distinguished

Share this page