Do you have elpida vram. They have a lot less oc headroom than sammys and hynix. If u hover over gddr5 in nvidia inspector it will tell u.
Yea I didn't really do any stress testing when I OCd to 4.8, I just stuck it up there and picked a vCore to try lol, Im really lazy when it comes to stabilizing my OCs, plus I was trying to keep it as low as possible still being on a £35 4 year old Air cooler (Tuniq Tower 120 Extreme) but it battles on ok, im pretty impressed considering what Im asking it to do Might give those drivers a shot tomorrow then, tbh I didn't notice any improvements with the latest WHQL anyway It says SK Hynix Inc. --
Just tested with 331.82 and it didnt made any real difference, but it sure did in LostPlanet2 (20fps difference in cpu bound test) and Rage gpu cuda transcode benchmark form 188 to 356, old 32x.xx driver had 398 :nerd: *nvm cpu reading I had windows balanced power plan
I just ran that bench again, and my memory speeds have dropped right back to like 16k, 15k, 15, again I haven't done anything apart from run the test - something isn't right with my system if it can jump almost 1000 points between runs -- Well benchmarks are one thing, but if you're seeing real world differences in games, that's where the money is - think I`ll drop back to those drivers then
Did you run with windows balanced power plan? But then again it shouldn't affect it like that.. Also try this normal memory benchmark Run it a few times if it shows the same, maybe its just that gpgpu benchmark specific
Running with High Performance power plan This was a quick run inbetween GPU driver installs (No driver installed for this run) Interestingly, when I ran DDU in safe mode - I find this Code: DDU Version: 10.1 Detected GPU : NVIDIA GeForce GTX 760 Detected GPU : NVIDIA GeForce GTS 450 Detected Driver(s) Version(s) : 9.18.13.3193 Detected Driver(s) Version(s) : 9.18.13.3221 I`ll do some more runs with 331.82 installed in a min EDIT - Yea the normal bench seems pretty stable in terms of results, give or take a few hundred MB either way
I guess its some internal bug by gpgpu and probably small conflict with those drivers, yes that small ram variation is fine.
@-Tj-: ATTENTION: you doesn't respect my advice and you may damage again videocard due to OC* and to probably triple buffering (TB) enabled. :evil: * - If your GTX 580 hasn't 900MHz core OC version/edition then you may damage this videocard.
Cool thanks, gonna have to get a water kit before the summer, Ive a feeling 1.48v will get pretty warm under load
I know what im doing and thanks for your concern. ps, when I downclocked my old gpu vram to 1900mhz it was ok (like nothing happened) gpu freq. is fine even at 930mhz, but I RMA'd it anyway since its default 2000mhz, will see if I get the replacement or not. //Lets get back on topic :nerd:
Nice results. Btw I found that with 6508 memory I got BSODs when playing BF4 and recording it via ShadowPlay. I'm now at 6458. No problems at all
How much RAM is on that 780 Ti? You are almost double my read and write. Can't post links or images yet because my post count is under 10. 2x 780Ti 1250/7600 Read: 6377 MB/s Write: 6211 MB/s Mem. Copy: 578015 MB/s Single Prec. FLOPS: 11626 GFLOPS Dbl Prec. FLOPS: 602.1 GFLOPS 24 BIT Int. IOPS: 2390 GIOPS 32 BIT Int. IOPS: 2391 GIOPS 64 BIT Int. IOPS: 593.8 GIOPS SHA-1 Hash: 24855 MB/s Single Prec. Julia: 2063 FPS Dbl. PRec. Mandel: 156.3 FPS
Any one ever wondered why some rigs wont bench the ''same''? Im seeing some 780's and 290's benching the same as my 7950 when it comes to memory read and write speeds.(6500mb's more or less) And then there is the same models doing 12000mb's My older 6970 does the same scores on read and write 6500+- Msi Z77A-G45 i7 2600K @4600mhz RipjawsX 8gb 2133@1600 8-8-8-28 (my board dont boot at 2133mhz) 2x7950 stock for now