Originally Posted by Topgearfan
I tried AFR 2 and did not notice a difference. I just ran valley again at 4.8ghz and did get 120.4FPS and score of 5038 but this is still way low. I ran AIDA64 for only 10min to see how the CPU OC was real fast since it failed at 4.8ghz with 1.3v and I bumped it to 1.35 and it ran 10min ok and saw a max CPU temp in one core of 79c but most of the time it was between 60-70. So increasing another 200mhz on the CPU clock did not yield another 6fps. I do not think I can run it much higher than what I have now. The crappy thing is I am limited on the GPU OC in SLI. ALone I can run my old card at 1176-1188mhz and 3554 on the memory. My new card I can get 1215/3406 but together I am limited in the memory and gpu clock to 1162/3354 and 1188/3354. If I bump the memory offset another 25mhz it will crash in scene 12-13. If I run my old card at 1176 and the new one at 1188 it will crash. I am going to send my old card in for RMA on Monday anyway due to a suspect displayport. I am going to pay full price for a new card from EVGA up front so they send it out right away and then refund me when they get my card back. Hopefully the replacement card will be more on par with the other one.
Ask them if they can guaranty a minimum ASIC while you're at it -- heck, who knows!
I just tested 4.6@ with 326.01 on Win8.1 @ 1176mhz & I got 74.1 (best I got on stock bios so far). When I tried 1189core, instead of crashing, the screen became garbled, then dark, then the image came back as if nothing had happened & the benchmark continued & finished without incident. Admittedly the score wasn't on par, but still... nice recovery! Either 326.01 has a much better recovery than 314.22 OR Windows 8 handles those things a lot better than Windows 7.
About the mysteries remaining, I thought about possible causes...
- Those uber scores might have 2 cards that OC to 1250+ / 7000+ easy -- that'd be a huge factor - which you can check by reverting to stock speeds & retesting @ whatever your CPU was then comparing.
- PCIe 2.0 vs 3.0 might make a (small?) difference.
- Sandy vs Ivy vs Haswell = roughly 5-10% increase at same clock speeds per generation.
- Computer memory OCing might give a few decimals in fps.
- Drivers used on benchmarks -- I got much better results on some drivers @ same clock speeds
- Windows 7 vs Windows 8 (seems Win8 feels smoother but I'd have to test with same drivers so I'll wait 'till they release the same for both)
- mercury is in retrograde (joke) lol
- tweaks for the benchmarks can go a lot further than just the nVidia Panel (set Valley priority to run in real time instead of normal, boot with minimal processes, OC with C1E or not, etc etc etc)
- the fps stuck at around 100 at scenes 10-15 is perplexing though (or the max you seem to hit).
Food for thoughts for sure.
hoping you get a good (at the very least) replacement card.Edited by Melk - 6/29/13 at 4:45pm