AMD Mantle Performance: Thief & Battlefield 4
- Thread starter Steve
- Start date
hahahanoobs
Posts: 5,221 +3,075
Nice review. The colours used in the charts were hideous, but I'm sure they will get better.
So if you don't know how to build a computer (weak CPU + $400 GPU + 16x10 resolution), and you only play Thief, then Mantle is for you.
PS, Mantle is a joke.
So if you don't know how to build a computer (weak CPU + $400 GPU + 16x10 resolution), and you only play Thief, then Mantle is for you.
PS, Mantle is a joke.
misor
Posts: 1,427 +326
I think AMD repeatedly lied to its loyal followers by claiming huge gains from the mantle technology.
I guess Nvidia engineers knew all along how mantle would perform and not be bothered by some game studios joining the mantle bandwagon.
real price cuts now? or just hype the e-coin business again?
good move for Microsoft's dx12 which has the support of both AMD and Nvidia.
We've got some nicer designs in the works for future stuff. Getting anything to look nice in Excel is... well... difficult.
wastedkill
Posts: 1,423 +348
Right now this just shows mantle isn't ready or atleast someone skimped out on either doing it properly on the game bf/thief or mantle just isn't ready...
G
Guest
Mantle reduces CPU overhead, because you ran the games at such high settings you were mostly GPU bottlenecked and mantle would have little effect. You should benchmark with slower CPU's or lower the games graphical settings so you are not GPU bottlenecked.
It’s as ready as its going to get, it’s not a magical frames per second API. It simply removes the CPU overhead and we saw that. This is all you are likely ever going to see from the current version of Mantle.
Slower CPU’s? Really slower than the A8-7600? Not sure there is any point to that or that it is even necessary.
Furthermore those investing in a Radeon R9 290X or R9 280X are obviously going to want to play these games using maximum quality settings. As for the R7 260X that was tested with lower quality settings and provided playable performance at 1680x1050 and here DX11 was faster. Explain that.
wastedkill
Posts: 1,423 +348
From what we have seen with drivers things will improve just like how CoD ghosts made 2 titans run at 20-30FPS it is as simple as patching it to improve it so your wrong things will improve its like a driver it just needs fine tuning like all things or if it didnt then why do drivers need improving?
Granted Mantle is new I don’t believe you are correct. Just look at the numbers. We are seeing 100% from Mantle right now, the evidence is all there.
Perfect example that Mantle is working as good as it’s ever going to is Thief.
Take the Radeon R9 290X results in Thief.
Core i7 = 42fps @ 2560x1600 with Mantle and DX11.
FX-8350 = 33fps with DX11 and the full GPU limited 42fps with Mantle, it can’t get any better than that.
Finally the example you gave in Ghosts is about Nvidia drivers and more importantly it is in regards to multi-GPU technology. It certainly isn't common for drivers to improve SLI or Crossfire performance in new games, that is an entirely different issue all together.
wastedkill
Posts: 1,423 +348
So OpenGL and Direct X can improve but mantle cant... thats a bit bias dont you think...
Point missed I guess. We will see what future drivers bring then.
Burty117
Posts: 5,108 +3,689
From what I can gather, all Mantle does is reduce the load on the CPU to which in turn frees up the GPU to work harder and therefore get a higher framerate, Since the i7 doesn't struggle anyway it just brings AMD processors in-line with Intel's.
Don't get me wrong, I'm sure there are other optimizations, but this one seems to be key and it's probably quite hard to improve on over time as it already allows developers to use all the threads an 8 core offers etc...
They may improve it over time, but it won't be anything significant unless they come up with some new optimizations they can implement into Mantle. Or developers haven't been using it very effectively.
Skidmarksdeluxe
Posts: 8,645 +3,290
I like the new Techspot 3.0 and I'm quickly becoming used to it but this is the 1st review I've read in this format and I found out that I can no longer quickly navigate to the 'final thoughts' without having to scroll through all the other pages. If it's a subject or product that I find interesting it's not a problem as I will read the entire review but not every review or subject tickles my fancy and I just want to read the introduction and closing, this is one of them.
I'd appreciate it if the devs would include a 'quickly navigate to page' option.
P.S. If this sounds like a slur on the reviewer or article I do apologise (no offense Steve, it is well written and researched) and in no way intended to be that.
I'd appreciate it if the devs would include a 'quickly navigate to page' option.
P.S. If this sounds like a slur on the reviewer or article I do apologise (no offense Steve, it is well written and researched) and in no way intended to be that.
wastedkill
Posts: 1,423 +348
Remember everything still isn't in yet so give it time they probably are working on that as we speak.
Polaco
Posts: 46 +7
"Core i3 through i7 owners can hope to see a 5% boost (a few fps) with Mantle but that's not enough to recommend it over the more mature DirectX 11 API for now."
Sorry, but what are you "not recommending" here?
Are you recommending developers to no use an API ? ( doing that based on this review would be hilarious )
Users to not enable Mantle on their system?
Do not considering Mantle as a feature when choosing their next video card?
Sorry, but what are you "not recommending" here?
Are you recommending developers to no use an API ? ( doing that based on this review would be hilarious )
Users to not enable Mantle on their system?
Do not considering Mantle as a feature when choosing their next video card?
Spykezxp
Posts: 296 +73
Mantle is still in its infantile state. Its still new to the game and needs time to improve. Yes they should have proceeded with releasing Mantle a better way, but it was still released. It does improve fps, even if its a small bump. This will hopefully get better with AMD making modifications and improvements to the coding. I have both an AMD CPU and AMD Graphics Card. I tried BF4 with Mantle and it does help some. We just have to give it time to improve.
On a side question: Why is there no testing down in 1920x1080? I know a lot of gamers that run at 1080 resolution, but I never see benchmarks for it on Techspot. Most budget gamers run at this resolution so shouldn't this be used? Will this ever be an option for testing?
On a side question: Why is there no testing down in 1920x1080? I know a lot of gamers that run at 1080 resolution, but I never see benchmarks for it on Techspot. Most budget gamers run at this resolution so shouldn't this be used? Will this ever be an option for testing?
I would probably gather the same if I hadn't witnessed a different story.
I was running BF4 with crossfire 290's on windows 7, DX11 and an i7 2600K at 4.7ghz. Mostly on medium to high settings as some stuff was turned off purely to give me a visual advantage playing multiplayer. Frames were hitting 200 fps and with some rare dips to approx 100fps and hovered anywhere inbetween these two extremes when playing. Up and down quite a lot.
Cue the 2nd mantle beta drivers (14.2 beta 1.3) and my 1st time trying mantle. I Fired up BF4 and my jaw dropped as I ran around. I thought it was super smooth before but it is way smoother using mantle. Also I noticed the fps were consistently hitting the 200fps limit of the game and dipping now and then to a lowest of approx 140+. So straight off the bat that's a 40% low frame drop improvement. The framerate is much more stable now with mantle and usually hovers around 160-170.
Now that would be great but then I noticed that I had forgotten that the last time I was messing with the game I had maxed everything out to ultra settings, so I was in effect comparing mantle with ultra game settings with DX11 medium to high settings and mantle was still killing it.
theBest11778
Posts: 298 +127
I think everyone's really missing the point of Mantle... it's not really meant for R9 290 owners... in fact it's not really designed for discrete graphics owners. AMD knows (Oh you better believe they KNOW,) their CPUs will never catch up to Intel. Plain and simple, Intel has far too many resources. It'd be like Sears trying to overtake Walmart.
Since the new generation of consoles have released games will become more CPU intensive, and graphically demanding (happens every generation.) For AMD to compete they need to think outside of the box. AMD discrete GPUs pair well with Intel chips... so they make some money there, but the market is focusing on low power, low heat, low noise, cheap APUs that do it all. Here comes Mantle to the rescue. Kaveri wasn't the knockout I was hoping for, but I bet AMD's next APUs keep the same Steamroller cores, but beef up the GPU. Mantle will allow games to run well with the weak CPU and strong GPU the same as an equivalent GPU paired with an Intel chip (I'd bet a 1024 Steam Processor APU is in the works.)
The moral of the story is... if you have a highend discrete GPU, pair it with an Intel chip (Core i5 is the best bang for your buck,) but if you want to buy a cheap APU that'll handle your gaming in your living room (Steambox,) AMD, like with this generation of consoles, will be your best option. The HW isn't there yet, but it will be sooner or later.
Since the new generation of consoles have released games will become more CPU intensive, and graphically demanding (happens every generation.) For AMD to compete they need to think outside of the box. AMD discrete GPUs pair well with Intel chips... so they make some money there, but the market is focusing on low power, low heat, low noise, cheap APUs that do it all. Here comes Mantle to the rescue. Kaveri wasn't the knockout I was hoping for, but I bet AMD's next APUs keep the same Steamroller cores, but beef up the GPU. Mantle will allow games to run well with the weak CPU and strong GPU the same as an equivalent GPU paired with an Intel chip (I'd bet a 1024 Steam Processor APU is in the works.)
The moral of the story is... if you have a highend discrete GPU, pair it with an Intel chip (Core i5 is the best bang for your buck,) but if you want to buy a cheap APU that'll handle your gaming in your living room (Steambox,) AMD, like with this generation of consoles, will be your best option. The HW isn't there yet, but it will be sooner or later.
GhostRyder
Posts: 2,151 +588
Well I have to say your number seemed alot lower than I was expecting on BF4. As for Thief, im actually shocked it gives that much of a performance boost in that game but ive noticed that because my friend is running an old Phenom II X6 1100t overclocked to 4.0ghz and an HD 7870 and it brought up the FPS at max to around 50-60 on a 1920x1080p screen.
Now for me on BF4, im running an FX 9590 and 3 R9 290X GPU's in 5760x1080p which for me is giving me a pretty constant 60 FPS in multiplayer that ive noticed (Albeit I don't stare at the FPS counter in the top of the screen while playing but it does not bother my eye skipping around constantly). On DX11 I noticed lots of skipping around in performance.
Im more curious why the results are so low here for BF4 because other places are reporting higher Mantle results.
Now for me on BF4, im running an FX 9590 and 3 R9 290X GPU's in 5760x1080p which for me is giving me a pretty constant 60 FPS in multiplayer that ive noticed (Albeit I don't stare at the FPS counter in the top of the screen while playing but it does not bother my eye skipping around constantly). On DX11 I noticed lots of skipping around in performance.
Im more curious why the results are so low here for BF4 because other places are reporting higher Mantle results.
wastedkill
Posts: 1,423 +348
Maybe techspot didnt do any crossfire cards like hardocp did?
GhostRyder
Posts: 2,151 +588
Well they didn't I was mostly referring to the Single tests as well in HardOCP and the frame time variance that was very smooth in BF4 with Mantle online. But of course everyones tests are going to be different because all hardware acts differently and scenarios will of course change just be the smallest of details.
cliffordcooley
Posts: 13,141 +6,441
Wouldn't 100 to 140 out of 200 be 20% not 40%?
