Mantle - Bottleneck?

[Juliogol]

Hi guys!

 

I've just received my brand new R9 290 Sapphire Tri x today (happy face ) and I got a question for you guys. I'm kinda noob when it comes to CPU.

 

I play Battlefield 4 the most, and when I plugged the card and fired up BF to test it, I get ok performance on DirectX (70-100 FPS on medium settings) but when I switch to Mantle I get AMAZING performance (100-120 FPS, Ultra with no AA).

 

From what I've read, Mantle aliviates the CPU load, and utilizes more GPU power right? So, if I'm getting much better performance on Mantle, does this mean that my proccesor is bottlenecking the card when on directX?

 

Testes Tomb raider too, and I get 50-70 FPS, max settings.

 

Here are my specs

 

Core I5 2500k @4.3 GHZ (with Hyper 212 cooler)

Asrock Extreme 4 Z77 motherboard

120 GB SSD

12 GB RAM

Cooler Master Silent Pro 700w PSU.

 

Should I start thinking about getting a better CPU? Witohut changing the motherboard it could be Ivy bridge only, because of the Socket 1155.

 

Would love to here your thoughts on this!

 

Thx in adavance!!!

[flibberdipper]

I think you should be doing better on DirectX with a 290. But... Maybe I'm thinking of a 290x.

[Scratchy1215]

Mantle is better for gaming, though ive had issues of crashing my card sometimes and just, crappy issues that make me turn it off. im waiting awhile before i actually use it when i play BF4 but im pretty sure its just removing some software/cpu usage to go straight to GPU processing power.

 

Theres always a small bottleneck in every system even by 1-2% but if your worried about your cpu. wait for broadwell.

[squirrl]

It's not so much that the CPU is the bottleneck, It's that the hardware isn't used as efficiently as it could be with DirectX.

[Juliogol]

So you think that upgrading to Ivy bridge would not make much of a diffrerence?

[FellTheSky]

So you think that upgrading to Ivy bridge would not make much of a diffrerence?

 

Not really. There is 300 mhz difference between sandy and ivy. So a 4.3ghz ivy would be around 4.6 ghz sandy. There are some apps where ivy is a bit faster, but for the most part that is the difference.

 

It seems to be something wrong with the system if you only reach 70 fps on medium with a r9 290. I would advice to crank up the details on DX11 and to check cpu, gpu usage and gpu clock speed during the game session

[7850OC]

Hi guys!

 

I've just received my brand new R9 290 Sapphire Tri x today (happy face ) and I got a question for you guys. I'm kinda noob when it comes to CPU.

 

I play Battlefield 4 the most, and when I plugged the card and fired up BF to test it, I get ok performance on DirectX (70-100 FPS on medium settings) but when I switch to Mantle I get AMAZING performance (100-120 FPS, Ultra with no AA).

 

From what I've read, Mantle aliviates the CPU load, and utilizes more GPU power right? So, if I'm getting much better performance on Mantle, does this mean that my proccesor is bottlenecking the card when on directX?

 

Testes Tomb raider too, and I get 50-70 FPS, max settings.

 

Here are my specs

 

Core I5 2500k @4.3 GHZ (with Hyper 212 cooler)

Asrock Extreme 4 Z77 motherboard

120 GB SSD

12 GB RAM

Cooler Master Silent Pro 700w PSU.

 

Should I start thinking about getting a better CPU? Witohut changing the motherboard it could be Ivy bridge only, because of the Socket 1155.

 

Would love to here your thoughts on this!

 

Thx in adavance!!!

you have a great cpu friend (no need to change it)

 

the thing is that, you also have a super monster gpu (that top all cpus)

 

and in some point, all cpus are bottlenecks

 

not much we can do at this point on that friend

 

well... Mantle can 

 

 

(people hope DX12 do the same, but... is all in shadows at this point) 

 

(oh, by the way, congrats for the Tri-X i personally think is one of the best cards out there)

[Juliogol]

 

you have a great cpu friend (no need to change it)

 

the thing is that, you also have a super monster gpu (that top all cpus)

 

and in some point, all cpus are bottlenecks

 

not much we can do at this point on that friend

 

well... Mantle can 

 

 

(people hope DX12 do the same, but... is all in shadows at this point) 

 

(oh, by the way, congrats for the Tri-X i personally think is one of the best cards out there)

 

 

I get a very solid FPS with Mantle, but as said before I will try to crank up the details on DX 11 and see what happens.

 

Thanks! I am also very happy with my purchase!

[<boss2939>]

I've had many bottlenecking issues in the past my advice is download msi afterburner than put all your settings low if the gpu usage is below 80-99% there is a bottleneck, but bf4 can get pretty cpu intensive in 64 man server so mantle helps with that by lowering the cpu usage.

[gilang01]

the only game that i want to support Mantle is planetside 2. cpu intensive to the max

[LOST TALE]

Open Windows Task Manager.  (search it in start menu OR press CTRL-ALT-DELETE and open it from security screen)

Now that your the matrix god of the digital univers ein your PC (not quite)

 

Go to performance Tab. And look at the graph if one of your cpu cores hit 100% while playing BF4, thats a wall bottlenck (caus eI happen to call it that way B) )

If all cores hit 100%, its a full bottleneck. (then you can tell us and yourself if there's a cpu bottleneck) ( its actually more complicated then that but its ok)

 

Go to processes tab.

 

There you will find BF4.exe along with all the other crap running ony our computer.

See how much % of your cpu BF4 uses. If its 70% and your cpu is at 90% this means everything else in your computer uses 20%. so if you want more fps because your a ''tearing no-sync madman'' You can clean some of the other stuff running on your computer, like the microsoft spyware along with other crap my prebuilt PC OS had.

 

If you terminate something you shouldn't you can always restart it or restart your computer as it doesnt delete files.

 

If you use teamspeak, using a channel without voice encryption will lower cpu usage from 2-3% to 0-1%

If you use skype, then you should consider to move away because skype is stone age and TS is electrical age.

[I/O]

I get perceivabley more smoothness when running mantle over directXin BF4.

[Sonefiler]

I think you should be doing better on DirectX with a 290. But... Maybe I'm thinking of a 290x.

Dude pretty much everything benefits from mantle. I have a r9 290 OCed to be faster than a 290X and I'm still seeing performance gains.

[Buckley]

My 290x & 8350 runs better on mantle rather than dx11

[xAcid9]

Maybe something is utilizing your CPU in the background. Did you disable Core Parking? If not, try that. 

[Juliogol]

Open Windows Task Manager.  (search it in start menu OR press CTRL-ALT-DELETE and open it from security screen)

Now that your the matrix god of the digital univers ein your PC (not quite)

 

Go to performance Tab. And look at the graph if one of your cpu cores hit 100% while playing BF4, thats a wall bottlenck (caus eI happen to call it that way B) )

If all cores hit 100%, its a full bottleneck. (then you can tell us and yourself if there's a cpu bottleneck) ( its actually more complicated then that but its ok)

 

Go to processes tab.

 

There you will find BF4.exe along with all the other crap running ony our computer.

See how much % of your cpu BF4 uses. If its 70% and your cpu is at 90% this means everything else in your computer uses 20%. so if you want more fps because your a ''tearing no-sync madman'' You can clean some of the other stuff running on your computer, like the microsoft spyware along with other crap my prebuilt PC OS had.

 

If you terminate something you shouldn't you can always restart it or restart your computer as it doesnt delete files.

 

If you use teamspeak, using a channel without voice encryption will lower cpu usage from 2-3% to 0-1%

If you use skype, then you should consider to move away because skype is stone age and TS is electrical age.

 

Nice post haha

 

Yep, on task manager BF4 is taking up to %90 of CPU usage, mostly if i am playing on a 64 server, and depending on the map. But I am almost sure that a core i5 2500k maybe a little low for this game as I am not having any problem with Tomb Raider for example.

[LOST TALE]

Maybe something is utilizing your CPU in the background. Did you disable Core Parking? If not, try that. 

First, I would check if there are parking issues ( if the cpu runs idle frequency when it shouldnt)

Because you are killing 90% of cpu  power consumption which translates into heat throughout the day + voltage and life span

 

Unparking cores is stupid.

[LOST TALE]

Nice post haha

 

Yep, on task manager BF4 is taking up to %90 of CPU usage, mostly if i am playing on a 64 server, and depending on the map. But I am almost sure that a core i5 2500k maybe a little low for this game as I am not having any problem with Tomb Raider for example.

Your not having any bottlenecks, that is very good performance. In fact it is tremendous.

 

In good software,

Aslong as the cpu doesn't touch 100%, any bottleneck is trivial.

 

Also you might hear that BF4 can reach ram bandwith bottlenecks ( your ram frequency and amount of channels determine  RAM bandwith)

But thats almost certainly outdated benchmarks. ( The BF4 patch history is the most tremendous optimisation I have ever seen, I bet its gone )

I would infact put my word on it.

 

Ofcourse oyu have the craps that drop FPS before any hardware capability saturation. usually they need a cas latency of 0 or some frankenstein bottleneck I don't know about. Something that makes it go slightly faster with every order of magnitude faster hardware.

[Humbug]

to OP, no point changing your CPU. It's very fast. And also mantle has had a big effect on the industry so the next versions of other graphics APIs like directX and opengl are attempting to optimize in similar ways to What mantle has demonstrated. This means that CPUs will be used more efficiently in the next few years and we will see fewer CPU bottlenecks.

[xAcid9]

First, I would check if there are parking issues ( if the cpu runs idle frequency when it shouldnt)

Because you are killing 90% of cpu  power consumption which translates into heat throughout the day + voltage and life span

 

Unparking cores is stupid.

 

If he didn't touch the Min Proc State, the CPU will still downclock to minimum, power consumption with Unparked core is marginal compare to Parked. It is not stupid if this solve his problem.

[LOST TALE]

If he didn't touch the Min Proc State, the CPU will still downclock to minimum, power consumption with Unparked core is marginal compare to Parked. It is not stupid if this solve his problem.

If It's a last resort, sure. I went one sided. The power consumption is not marginal. The power consumption of 0% usage at parked frequency vs unparked triples with both my GPU and my CPU. Because the differenc ein frequency happens to be 3.3x

I have wattage sensor on both GPU and CPU (multiple on CPU)

 

Marginal? definitly not. Unless 300% is marginal.

The frequency alone definitly scales power consumption. its almost % on % thats because theres voltage and the actual usage of it.

I just tested my cpu by hovering over interactive icons until the frequency would double from 1.6Ghz to 3.5Ghz.

the wattage doubled from 7Watts to 16 Watts

I think voltage would be included too.The result is very consistent and instantenous.

I did not add wattage up, I looked at the package wattage.

I used ASUS AI suite to look at cpu frequnecy and Hardwaremonitor for wattage.

The IA cores went from 2.5Watts to 8 watts and then with same usage ( well % usage halves once frequnecy doubles)

reached 14 watts.

 

Further more my motherboard may drop voltage/frequency ratio when % usage is low. Other motherboards I think dont drop voltage outside parked frequency I would think.

And so the difference would be even bigger as you add the unnecessary voltage increase.

 

I've never actually verified this hypothesis this far. No regrets.

 

Do you have a different result?

[LOST TALE]

If he didn't touch the Min Proc State, the CPU will still downclock to minimum, power consumption with Unparked core is marginal compare to Parked. It is not stupid if this solve his problem.

Well instead of doing that he could

first actually verify if its a cpu issue and if parking is an issue.

Then I would try the min cpu proc state in power options(right?).

 

I think most people who used this never had their cpu in parked mode when playing (because that would get them like 10x or 3x FPS) instead of just gaining 10 fps

because the game wasnt optimised I assume.

[xAcid9]

If It's a last resort, sure. I went one sided. The power consumption is not marginal. The power consumption of 0% usage at parked frequency vs unparked triples with both my GPU and my CPU. Because the differenc ein frequency happens to be 3.3x

I have wattage sensor on both GPU and CPU (multiple on CPU)

 

Marginal? definitly not. Unless 300% is marginal.

The frequency alone definitly scales power consumption. its almost % on % thats because theres voltage and the actual usage of it.

I just tested my cpu by hovering over interactive icons until the frequency would double from 1.6Ghz to 3.5Ghz.

the wattage doubled from 7Watts to 16 Watts

I think voltage would be included too.The result is very consistent and instantenous.

I did not add wattage up, I looked at the package wattage.

I used ASUS AI suite to look at cpu frequnecy and Hardwaremonitor for wattage.

The IA cores went from 2.5Watts to 8 watts and then with same usage ( well % usage halves once frequnecy doubles)

reached 14 watts.

 

Further more my motherboard may drop voltage/frequency ratio when % usage is low. Other motherboards I think dont drop voltage outside parked frequency I would think.

And so the difference would be even bigger as you add the unnecessary voltage increase.

 

I've never actually verified this hypothesis this far. No regrets.

 

Do you have a different result?

 

Look pretty similar to me. I've only tested it when i still have my kill-o-wat meter last time. I monitored my system power consumption and the difference between Idle with core parked and unparked only around 6-10 watts on my system. It may look huge when use % but if u look at overall system consumption, it's pretty small. 

 

Well instead of doing that he could

first actually verify if its a cpu issue and if parking is an issue.

Then I would try the min cpu proc state in power options(right?).

 

I think most people who used this never had their cpu in parked mode when playing (because that would get them like 10x or 3x FPS) instead of just gaining 10 fps

because the game wasnt optimised I assume.

 

That's the reason, the game was't really optimized. My problem with BF4 when enabled Core Parking was micro stutters. 

[Juliogol]

Cores are unparked, just tested again with Directx on ultra I get 50-70 FPS. ON mantle, same settings 90-120 FPS. In both cases with AA off.

 

It's almost double the performance with mantle. Weird.  :wacko:

 

EDIT: Added a capture of task manager while running BF4 on Mantle