| Widescreen Gaming Forum http://www.wsgf.org/phpBB3/ |
|
| CPU OCing and performance gain at triple wide res http://www.wsgf.org/phpBB3/viewtopic.php?f=72&t=24576 |
Page 1 of 3 |
| Author: | Wijkert [ 11 Aug 2012, 18:49 ] |
| Post subject: | CPU OCing and performance gain at triple wide res |
A couple of weeks ago I started ocing my 2500k again. Since then I have experienced some light stuttering during gaming. Running my cpu at stock stopped this stuttering, so I figured that the stuttering had to do with a bad oc. This got me thinking, what kind of advantage is ocing the cpu when gaming at very high resolutions (6048x1080)? Started testing and here are the results: [email protected] Dirt 2 Run 1: avg 72.7 fps Run 2: avg 73.3 fps [email protected] Dirt 2 Run 1: avg 75.9 fps Run 2: avg 76.7 fps [email protected] BF3 Run 1: avg 69.1 fps Run 2: avg 68.9 fps [email protected] BF3 Run 1: avg 68.2 fps Run 2: avg 68.4 fps In Dirt 2 the gain from ocing was 3.4% and in bf3 the gain was -0.7%. Keep in mind that these findings are based on just 2 games, but since I am playing them right now they were at least relevant to me. I would guess that the difference would be higher when testing cpu intensive games like Starcraft 2 or Civilization 5. Battlefield can be cpu intensive but only in multiplayer on 64 player maps for example. Doing runs on those servers would produce results which aren’t reproducible. Also, when you run more than one videocard things might be different as well. Although I really like to push my hardware to the limit, even I don’t think that ocing your cpu when gaming at triple wide resolutions is worth it. These findings indicate that your gpu will almost always be the bottleneck when gaming at higher resolutions. |
|
| Author: | Gilly [ 12 Aug 2012, 10:11 ] |
| Post subject: | Sorry to sound like an arse, |
Sorry to sound like an arse, but I think most people realised the higher the resolution the less the CPU clock speed matters... What would be interesting is seeing how FPS is related to the "FSB" speed of the Sandy/ Ivy Bridge, as that effects the speed of the PCI-e bus too. |
|
| Author: | Wijkert [ 12 Aug 2012, 11:03 ] |
| Post subject: | Gilly wrote:Sorry to sound |
Sorry to sound like an arse, but I think most people realised the higher the resolution the less the CPU clock speed matters... You would be suprised how many people keep asking this question. I already suspected what the results would be, but would have predicted a 5-10% increase instead of 0-5%. What would be interesting is seeing how FPS is related to the "FSB" speed of the Sandy/ Ivy Bridge, as that effects the speed of the PCI-e bus too. Since the release of SB, Intel made it very difficult to overclock that way. Besides I would guess that 'fsb' ocing would produce the same kind of small performance increase at similar clock speeds. |
|
| Author: | Gilly [ 12 Aug 2012, 15:55 ] |
| Post subject: | I realise that the "FSB" is |
I realise that the "FSB" is basically a clock generator for every other bus connected to the CPU, what I was meaning was if you ran it at say 105Mhz instead of 100Mhz, thus giving, theoretically 5% more bandwidth, whilst keeping the clock speed fairly similar (you could run 100x40 for 4000Mhz give or take and say 105x38 for 3990Mhz give or take), it would just be interesting to see if the increase in bus speed would help the FPS more than the large % increase in clock speed did/didn't. I personally run at 100x42 and find no problems what so ever, temperatures are super low and I use an offset of -0.27v IIRC, power consumption at the wall under idle or load is very low. |
|
| Author: | Wijkert [ 12 Aug 2012, 16:17 ] |
| Post subject: | Gilly wrote:I realise that |
I realise that the "FSB" is basically a clock generator for every other bus connected to the CPU, what I was meaning was if you ran it at say 105Mhz instead of 100Mhz, thus giving, theoretically 5% more bandwidth, whilst keeping the clock speed fairly similar (you could run 100x40 for 4000Mhz give or take and say 105x38 for 3990Mhz give or take), it would just be interesting to see if the increase in bus speed would help the FPS more than the large % increase in clock speed did/didn't. I understand what you mean, but if bandwidth would indeed increase fps, wouldn’t you see a performance increase when using pcie 3.0 instead of 2.0 (since that doubles the bandwidth)? |
|
| Author: | Haldi [ 12 Aug 2012, 18:39 ] |
| Post subject: | ...... you're doing |
...... you're doing Benchmarks with Fraps? Then at least check "Maxmimum/Minimum FPS logging" ! Especially for BF3 which hase some CPU intensive parts Minimum FPS would be interessting! Best would be a graph :) So about BCLK, FrontSideBus was on socket 775 and mainly for CPU&RAM. BCLK nowadays is directly linked to PCI RAM and some other stuff. thats why you can't OC it much, mine (Asus P9x79 Pro) goes up to 104.7mhz, any more is instable. But kinda sure it's possible to cut off PCi-E lane and clock that separatly, not sure if limited on 100mhz or variable. So about OC'ing PCi-E lanes. As stated by Wijkert there is no GPU that would use full bandwith of PCI-E 3.0 16x. The only way how you might get better performance would be in Crossfire. If i have my 2nd GPU back in the water circuit i'm gonna check out how much PCI-E overclocking capabilitys i have and how much performance i will gain. (might take some time, not sure when i have enough free time ^^) |
|
| Author: | Wijkert [ 12 Aug 2012, 21:14 ] |
| Post subject: | As Haldi already stated, |
As Haldi already stated, minimum fps is more important that avg fps. So here we go: [email protected] Dirt 2 Run 1: min/max 63/86 fps Run 2: min/max 65/92 fps [email protected] Dirt 2 Run 1: min/max 60/92 fps Run 2: min/max 69/91 fps [email protected] BF3 Run 1: min/max 57/98 fps Run 2: min/max 57/96 fps [email protected] BF3 Run 1: min/max 56/96 fps Run 2: min/max 57/94 fps All were indeed recorded using fraps. If you take the average of both runs, all results seem to be within margin of error. So it seems that my original conclusion remains the same. |
|
| Author: | Gilly [ 13 Aug 2012, 11:10 ] |
| Post subject: | His card is a PCI-e 3.0 card |
His card is a PCI-e 3.0 card being driven from a 2500K on a Z68 board, which are both PCI-e 2.0 I was personally interested if you fed the 7950 more bandwidth from the PCI-e slot if you would get more bang for buck than just overclocking the CPU, but seems you don't wanna do that! |
|
| Author: | Wijkert [ 13 Aug 2012, 12:09 ] |
| Post subject: | Gilly wrote:His card is a |
His card is a PCI-e 3.0 card being driven from a 2500K on a Z68 board, which are both PCI-e 2.0 I am not unwilling to test your hypotheses, bust based on this review it seems pointless. PCI-e 2.0 doesn't restrict the performance of my gpu. Even at double bandwidth ( PCI-e 3.0 ) the performance increase is very small to nonexistent. |
|
| Author: | Eerazer [ 13 Aug 2012, 17:43 ] |
| Post subject: | Wijkert wrote:Gilly wrote:His |
[quote]His card is a PCI-e 3.0 card being driven from a 2500K on a Z68 board, which are both PCI-e 2.0 I am not unwilling to test your hypotheses, bust based on this review it seems pointless. PCI-e 2.0 doesn't restrict the performance of my gpu. Even at double bandwidth ( PCI-e 3.0 ) the performance increase is very small to nonexistent. it's 16%. so really not very small to nonexistent. |
|
| Page 1 of 3 | All times are UTC [ DST ] |
| Powered by phpBB® Forum Software © phpBB Group http://www.phpbb.com/ |
|