**Official THIEF-Incorporating AMD™ Mantle™/True Audio™/Bench mark Thread**

[stulid]

Hold on, give me a mo

 

[stulid]

Attached zipped up GPU-Z files.

 

[LtMatt]

Attached zipped up GPU-Z files.

Yeah, as you can see the core clock is fluctuating when it should be locked at whatever clock you set it at. Assuming you used 50% power tune?

 

[stulid]

But its not fluctuating during the actual benchtest, only on either side of the test does it slow down.

So the stock run,

test starts - 2014-03-21 16:10:26 ,
test ends - 2014-03-21 16:11:37 ,


Overclocked,

test starts - 2014-03-21 16:17:54 ,
test ends - 2014-03-21 16:18:37 ,

 

[LtMatt]

But its not fluctuating during the actual benchtest, only on either side of the test does it slow down.

So the stock run,

test starts - 2014-03-21 16:10:26 ,
test ends - 2014-03-21 16:11:37 ,


Overclocked,

test starts - 2014-03-21 16:17:54 ,
test ends - 2014-03-21 16:18:37 ,

I will do some testing myself and report back, thanks.

 

[Kurisu Makise]

Mantle seems to be very hit and miss

Heres my Thief benchmark

Ultra 1080p / 14.3 v1 beta driver / win8.1
i3-3220 + 7950

DirectX
MIN - 16
AVG - 33.4
MAX - 62

MANTLE
MIN - 22
AVG - 31
MAX - 58.5

Quite the dissapointment as i was hoping for a boost from the 33avg up to 45avg or close with mantle enabled, is it that i have intel cpu and not amd or is because it just works on R9 290 and not HD7000 series?
im 16hours in to the game and finished so it doest really matter now i suppose.

i found this

(steam community thread)

Mantle actually hurting performance? [Discussion]
Has anyone diagnosed the common denominator amongst people who are getting worse perf with Mantle on than with it off? I've read a few threads on here regarding Mantle and
they seem to be a mixed bag of people getting solid boosts with no real downsides and those of us who are suffering, lol.

My specs for reference:

Win 8.1 Pro 64-bit
R9 270X
FX-8350
16GB Ram
Traditional HDD

I get stutters, single digit min fps, roughly the same or worse avg fps, and roughly the same or lower max fps.

 

[koldphuzion]

Why oh why is my system sooo slowwwwww.....

 

[Martini1991]

Mantle seems to be very hit and miss

Heres my Thief benchmark

Ultra 1080p / 14.3 v1 beta driver / win8.1
i3-3220 + 7950

DirectX
MIN - 16
AVG - 33.4
MAX - 62

MANTLE
MIN - 22
AVG - 31
MAX - 58.5

Quite the dissapointment as i was hoping for a boost from the 33avg up to 45avg or close with mantle enabled, is it that i have intel cpu and not amd or is because it just works on R9 290 and not HD7000 series?
im 16hours in to the game and finished so it doest really matter now i suppose.

i found this

(steam community thread)

Your CPU is a total mismatch for your GPU, that's the long and short of it.
My brother does similar (I3 2120 with a 7870, he said he'd upgrade the i3 quickly, as we only paid 50 for it, but a year down the line and he hasn't)

 

[melmac]

Your CPU is a total mismatch for your GPU, that's the long and short of it.
My brother does similar (I3 2120 with a 7870, he said he'd upgrade the i3 quickly, as we only paid 50 for it, but a year down the line and he hasn't)

Actually, I think his CPU should be fine with Mantle, I would say the problem is more that the AMD 7xxx series aren't optimised for Mantle yet.

 

[Silent_Scone]

Why aren't the Mantle results on the scoreboard yet, Tommy?

 

[meldarthx]

Mantle seems to be very hit and miss

Heres my Thief benchmark

Ultra 1080p / 14.3 v1 beta driver / win8.1
i3-3220 + 7950

DirectX
MIN - 16
AVG - 33.4
MAX - 62

MANTLE
MIN - 22
AVG - 31
MAX - 58.5

Quite the dissapointment as i was hoping for a boost from the 33avg up to 45avg or close with mantle enabled, is it that i have intel cpu and not amd or is because it just works on R9 290 and not HD7000 series?
im 16hours in to the game and finished so it doest really matter now i suppose.

i found this

(steam community thread)

driver install will cause that; if drivers haven't been completely removed and then reinstalled you will get that. I know I was getting that.

I'm 14.2 - I'm going to move to 14.3s but so damn smooth on 14.2 we'll have to see and the fact when I originally installed them it ate my win 8 drive LOL

 

[Kurisu Makise]

Your CPU is a total mismatch for your GPU, that's the long and short of it.
My brother does similar (I3 2120 with a 7870, he said he'd upgrade the i3 quickly, as we only paid 50 for it, but a year down the line and he hasn't)

i havent had any problems with bottlenecking with this set up but i do know that a better cpu would give me a nice boost.

Actually, I think his CPU should be fine with Mantle, I would say the problem is more that the AMD 7xxx series aren't optimised for Mantle yet.

im thinking this ^

driver install will cause that; if drivers haven't been completely removed and then reinstalled you will get that. I know I was getting that.

I'm 14.2 - I'm going to move to 14.3s but so damn smooth on 14.2 we'll have to see and the fact when I originally installed them it ate my win 8 drive LOL

14.3 first drivers installed on to a new ssd with win8.1

 

[tommybhoy]

Why aren't the Mantle results on the scoreboard yet, Tommy?

Because there is no way to capture them yet-which is the general rule with all the BM's here, just following suit.

I wasn't wanting to get stick for adding any(forum being the forum and all that), I can add them no problem if the majority is in agreement.

 

[Martini1991]

Ignore.

 

[LtMatt]

Where can i get some of these training wheels you speak of?

 

[Kurisu Makise]

essentially your gpu has training wheels on

even if thats so isnt the whole idea of mantle to give me better results when running on a rig like mine?

my usual numbers in game are

cpu 60% gpu 100% usage

the only difference is BF4

cpu 90% gpu 100%

 

[Martini1991]

Yeah, Mantle should definitely benefit you most, and part of your results will be down to needing more optimization.

But even then, I saw Borderlands 2 cause a massive CPU bottleneck for my brother for example, and there's plenty of others.

An i3 is a decent CPU, but it's not meant to push around a 7950.

 

[drunkenmaster]

Your CPU is a total mismatch for your GPU, that's the long and short of it.
My brother does similar (I3 2120 with a 7870, he said he'd upgrade the i3 quickly, as we only paid 50 for it, but a year down the line and he hasn't)

Actually, I think his CPU should be fine with Mantle, I would say the problem is more that the AMD 7xxx series aren't optimised for Mantle yet.

I would say it's neither, there are parts of the benchmark that aren't stressing the cpu much most likely hence the maximum FPS isn't increasing, but look at the minimum's, 40% increase in minimums is a hugely impressive result and that will appear much much smoother in game. I would firstly say the result is genuinely good, but unlike the Star Swarm demo, this is a real world game. DX overhead is a huge problem, but a game is a game, it's variable. Sometimes you are in a corner, hidden, waiting for a guard to pass while in a shadow with most of the screen dark, sometimes you are in the middle of 5 enemies with several light sources, many more dynamic shadows, many more things being done. More things being done = higher overhead.


In real world usage Mantle, DX12, any API, there will be intense and less intense parts of games. Mantle is help bringing the minimums up massively in the most intensive scenes.

Look at other results where the max isn't necessarily improving but minimums are going up anything from 25-40%. I forget who but someone specifically said the average didn't change but their minimum was up and in game it's vastly smoother. Many many people have said the same about BF4.

In a benchmark designed to push the intensity of the scene constantly the difference will be much much bigger. The difference is very clearly there in real game benchmarks, it's just reflecting in the minimums and the playability of the game.


For me, I've not run a 3dmark in 5+ years, I haven't benchmarked BF4, I don't have Starswarm even installed(and likely won't ever bother). I've got the info I need from articles/presentations and other users. Playability is all I've ever given a damn about. 25-40% increase in minimums is the holy grail in "playability" improvements with graphics, increased maximums do nothing, increased average is nice but unimportant.(reductions are bad in general though). Smoothness is important, and for me higher frame rate too, a "better" API will actually prevent extra work being done to improve the minimums, steadying out frames. Ask most dev's and they won't talk to you about FPS, they'll talk to you about latency and most importantly, predictability of it.

DX(as admitted by MS/Nvidia/everyone else in the dx12 talk) has the current issue of being very black box and unpredictable, it will randomly choose when to compile things and can cause an obvious stall that can lead to a frame dropped. Controlling each frame, having a known latency and getting a smoother game is the key here.

25-40% increase in minimums is a phenomenal result for Mantle in a real world game.

 

[Martini1991]

http://forums.overclockers.co.uk/showpost.php?p=26030840&postcount=580


Look at my gains, that's a 4.5GHZ 4770K (Albeit with a 1000MHZ R9 290) but I'm running well over twice Samsara's CPU grunt, but I see decent gains across the board, Samsara's end GPU grunt is what, 45% less? 7950 versus R9 290 (Clocks depending of course)

Some of the problem will definitely be Mantle not being quite ready yet, and especially not for the 79XX, but for me in Thief, Mantle increased EVERYTHING.

 

[LtMatt]

http://forums.overclockers.co.uk/showpost.php?p=26030840&postcount=580


Look at my gains, that's a 4.5GHZ 4770K (Albeit with a 1000MHZ R9 290) but I'm running well over twice Samsara's CPU grunt, but I see decent gains across the board.

Some of the problem will definitely be Mantle not being quite ready yet, and especially not for the 79XX, but for me in Thief, Mantle increased EVERYTHING.

Impressive.