My GTX460 End of life benchlog in detail, this is a wall of text, but maybe an interesting read for fellow GTX460 owners who may want to send off their card with a bang before upgrading them.
With the imminent release of nVidia new GPU's over the coming months i am planning again to grab a pair mid range cards. So i thought i should take an all balls out attempting at ringing out ever scrap of performance my 460s had to offer. After my previous mild attempts at overclocking them in the past haven't really given any magnificent results this was apparently due to my testing methods. so after asking a few questions over at the evga forums i set out to have another bash at it.
Now both my GPUS are identical EVGA 01G-P3-1370-KR card's.
For testing i set my 2500k @ 4.7 & ram @ 1600 (1333 default), but this is where i did things differently. Because i was going for single card peak performance card by card rather than SLI performance normally i would take out the 2nd card. BUT i was told that to help find the max stable clocks to keep in my 2nd card and set it to dedicated physics while testing & keeping the 2nd card at stable clocks. Now i said before both my cards are identical but their individual overclocking limits are very very different. My 2nd card (bottem) previously was my primary (top) card in my previous tests and it was the card i had used last time when aiming for highest single 460 scores. Its previous highest stable clocks are 880 core running default voltage & 930 core running max 1087mV and a 2140k memory clock. But my now primary card i had never taken past 850 core/2k memory as the old in fear of binning the first card i wanted to keep the other working just in case. But now its time to ring its neck and was i in for a shock.
I grabbed the latest MSI afterburner, unlocked the volts, forced on consistent voltage and set it to 1k. Leaving the memory at 2k i started raising the core clocks 20 Mhz at a time then running msi kombuster followed by a full run of 3dmark11 if it was stable. This went on for an hour or so until i hit 940 core @ 1000 mV and i got my first fail artifacts both in kombuster and 3dmark11. i bumped up the mV to 1050 and tried again successfully until 970 gave the same failures. i bumped the mV upto max 1087 and tried again, again a successful past followed but alas this was the best i could squeeze out of the core. 980 not only crashed the nvidia drivers it gave desktop 2d level artifacting. Panicked i rest and things went back to normal and left the core at 970, but having been using the 2nd 460 for physics i then switched physics back to primary but left the 2nd card installed and ran the tests at 970c/2000m and scored a 4 points higher 3dmark11 score.
Onto memory clocks, once 970 was 100% stable, knowing memory is not as sensitive as the core i jumped straight upto 2100 with hopes it would skip about 30mins of testing, and it did with a solid return of 4 more points over the 2k score. Thinking maybe i had hit a performance limit i and looking at the average highest memory clocks other have gotten i figured id jump straight up to 2150 and that would fail and i would then have to do a few backwards runs to find the limit, but to my surprise 2150 worked, so i went to 2175 an got the same solid stable testing. Thinking ok, maybe this cards got some potential after all i kicked it up to 2200 and thought for sure this would be returning fails across the board as i've only seen watercooled 460s get 2200 & beyond. But to my utter shock kombuster & 3dmark11 ran and to my shock returned a perfect score of 4800, almost 200 points higher than EVGAs official record of 4635.
Texting a mate excitedly (facepalm i know) i had just trounced the evga record (knows i had already beaten it at 940/2k anyway) i then continued to bumped by 25mhz and repeat tests i was starting to see a clear wall in 3dmark11 of 4802 between 2200 & 2250 i decided it was time to put the physics back on the same primary card and test again, but for some unknown reason (not thinking) i set the memory clock to 2300 without realizing it and hit the kombuster button and then 3dmark11 once more, but this time getting the perfect 4800 scored again. i exit and check the gpu settings and see its set to 2300 not 2250. thinking wow.... and it was stable, but at a clear wall. I then decided going 10 mhz at a time up until i hit 2330 and got a complete driver crash fail and artifacting in 3dmark11.
At this point i sit back and look over the benchlog i was making along the way and compare it to some of the evga scores and clocks. Not seeing anyone running stock cooling and getting 2200 or above i started to Google highest GTX460 stock cooling clocks and found a few reports of 2200 & 2225 but nothing near 2300+. So knowing i had hit a clear peak i starting tweaking things, i removed the 2nd card from my system and set the nvidia image quality settings to high performance and ran 3dmark11 once more, returning a blistering score of 4815, over 1200 points above stock performance @ 720/1800.
At this point i had forgot about sli performance so i was eager to see the difference, but having never ran each card at difference clocks during SLI i was unsure if this could even be stable let along safe. But i had already committed to fact i was ready to loose 1 (or both) cards in the process of this testing anyway so i figured what the hell. I knew cards at different stock speeds could run together but i was unsure if with such high clocks and both cards running the max 1087 core voltage if it was still going to work. But so my surprise it did and return a nice score of 8221, 1000 points higher again than my previous sli score.
I also researched and attempted to raise the limitation of MSI Afterburners vcore cap asn well as flash the GPUs bios with a unlocked volt bios but this didn't seem to take. Some people reported their cards achieved up to P6000+ with volt mods (see 3dmarks11s score results) but for the most part running the normal cap of 1087, P4800 scores seem to be the cherry picked cards from premium vendors. I may in the future attempting again to unlock the voltage as im pretty confident in the 1 card hitting 1Ghz core with a vcore of 1100 or so.
In conclusion, i feel i managed to score a cherry card and it will be sad to see it and the other go in the near future. But i will say im going EVGA for my next cards, clearly they can take a beating and still live to tell the story another day. Not bad considering its equal to or only 1k points off brand new top end AU$500+ cards.
Below is the 3DMark11 benchlog.
Test conditions:
Both GPUS are: 01G-P3-1370-KR's
stock air cooling ran at 100% during testing.
Idle/load temps are GPU#1 @ 28/49 & GPU#2 @ 27/48
Ambient temp of 10 degrees, directly piped from outside and blown directly onto both cards. (<3 cold weather)
Primary GTX460 SLOT #1:
3dmark.com/3dm11/2915879 - P4615 - 930/2000 - mV 1000 - (2nd card phx 930/2000)
3dmark.com/3dm11/2915905 - P4658 - 940/2000 - mV 1050 - (2nd card phx 930/2000)
3dmark.com/3dm11/2915921 - P4693 - 950/2000 - mV 1050 - (2nd card phx 930/2000)
3dmark.com/3dm11/2915944 - P4726 - 960/2000 - mV 1050 - (2nd card phx 930/2000)
3dmark.com/3dm11/2916025 - P4770 - 970/2000 - mV 1087 - (2nd card phx 930/2000)
3dmark.com/3dm11/2916085 - P4774 - 970/2000 - mV 1087 - (Primary card phx)
3dmark.com/3dm11/2916059 - P4775 - 970/2000 - mV 1087 - (2nd card phx 930/2000)
3dmark.com/3dm11/2916110 - P4778 - 970/2100 - mV 1087 - (2nd card phx 930/2000)
3dmark.com/3dm11/2916136 - P4789 - 970/2150 - mV 1087 - (2nd card phx 930/2000)
3dmark.com/3dm11/2916151 - P4791 - 970/2175 - mV 1087 - (2nd card phx 930/2000)
3dmark.com/3dm11/2916176 - P4800 - 970/2200 - mV 1087 - (2nd card phx 930/2000)
3dmark.com/3dm11/2916198 - P4802 - 970/2225 - mV 1087 - (2nd card phx 930/2000)
3dmark.com/3dm11/2916216 - P4802 - 970/2250 - mV 1087 - (2nd card phx 930/2000)
3dmark.com/3dm11/2916244 - P4800 - 970/2300 - mV 1087 - (Primary card phx)
3dmark.com/3dm11/2916271 - P4812 - 970/2320 - mV 1087 - (Primary card phx) (Absoluse max stable)
3dmark.com/3dm11/2922170 - P4815 - 970/2320 - mV 1087 - (2nd card removed from system)
FAIL FAIL FAIL FAIL FAIL - P NIL - 970/2330 - mV 1087 - (Primary card phx) (Green artifacting)
FAIL FAIL FAIL FAIL FAIL - P NIL - 970/2340 - mV 1087 - (Primary card phx) (Afterburner max setting)
2nd GTX460 SLOT #2:
3dmark.com/3dm11/2915766 - P4642 - 930/2000 - mV 1087 -(2nd card phx)
SLI Score:
3dmark.com/3dm11/2916317 - P8221 - 970/2321 & 930/2000 - mV 1087 - (Primary card phx)
With the imminent release of nVidia new GPU's over the coming months i am planning again to grab a pair mid range cards. So i thought i should take an all balls out attempting at ringing out ever scrap of performance my 460s had to offer. After my previous mild attempts at overclocking them in the past haven't really given any magnificent results this was apparently due to my testing methods. so after asking a few questions over at the evga forums i set out to have another bash at it.
Now both my GPUS are identical EVGA 01G-P3-1370-KR card's.
For testing i set my 2500k @ 4.7 & ram @ 1600 (1333 default), but this is where i did things differently. Because i was going for single card peak performance card by card rather than SLI performance normally i would take out the 2nd card. BUT i was told that to help find the max stable clocks to keep in my 2nd card and set it to dedicated physics while testing & keeping the 2nd card at stable clocks. Now i said before both my cards are identical but their individual overclocking limits are very very different. My 2nd card (bottem) previously was my primary (top) card in my previous tests and it was the card i had used last time when aiming for highest single 460 scores. Its previous highest stable clocks are 880 core running default voltage & 930 core running max 1087mV and a 2140k memory clock. But my now primary card i had never taken past 850 core/2k memory as the old in fear of binning the first card i wanted to keep the other working just in case. But now its time to ring its neck and was i in for a shock.
I grabbed the latest MSI afterburner, unlocked the volts, forced on consistent voltage and set it to 1k. Leaving the memory at 2k i started raising the core clocks 20 Mhz at a time then running msi kombuster followed by a full run of 3dmark11 if it was stable. This went on for an hour or so until i hit 940 core @ 1000 mV and i got my first fail artifacts both in kombuster and 3dmark11. i bumped up the mV to 1050 and tried again successfully until 970 gave the same failures. i bumped the mV upto max 1087 and tried again, again a successful past followed but alas this was the best i could squeeze out of the core. 980 not only crashed the nvidia drivers it gave desktop 2d level artifacting. Panicked i rest and things went back to normal and left the core at 970, but having been using the 2nd 460 for physics i then switched physics back to primary but left the 2nd card installed and ran the tests at 970c/2000m and scored a 4 points higher 3dmark11 score.
Onto memory clocks, once 970 was 100% stable, knowing memory is not as sensitive as the core i jumped straight upto 2100 with hopes it would skip about 30mins of testing, and it did with a solid return of 4 more points over the 2k score. Thinking maybe i had hit a performance limit i and looking at the average highest memory clocks other have gotten i figured id jump straight up to 2150 and that would fail and i would then have to do a few backwards runs to find the limit, but to my surprise 2150 worked, so i went to 2175 an got the same solid stable testing. Thinking ok, maybe this cards got some potential after all i kicked it up to 2200 and thought for sure this would be returning fails across the board as i've only seen watercooled 460s get 2200 & beyond. But to my utter shock kombuster & 3dmark11 ran and to my shock returned a perfect score of 4800, almost 200 points higher than EVGAs official record of 4635.
Texting a mate excitedly (facepalm i know) i had just trounced the evga record (knows i had already beaten it at 940/2k anyway) i then continued to bumped by 25mhz and repeat tests i was starting to see a clear wall in 3dmark11 of 4802 between 2200 & 2250 i decided it was time to put the physics back on the same primary card and test again, but for some unknown reason (not thinking) i set the memory clock to 2300 without realizing it and hit the kombuster button and then 3dmark11 once more, but this time getting the perfect 4800 scored again. i exit and check the gpu settings and see its set to 2300 not 2250. thinking wow.... and it was stable, but at a clear wall. I then decided going 10 mhz at a time up until i hit 2330 and got a complete driver crash fail and artifacting in 3dmark11.
At this point i sit back and look over the benchlog i was making along the way and compare it to some of the evga scores and clocks. Not seeing anyone running stock cooling and getting 2200 or above i started to Google highest GTX460 stock cooling clocks and found a few reports of 2200 & 2225 but nothing near 2300+. So knowing i had hit a clear peak i starting tweaking things, i removed the 2nd card from my system and set the nvidia image quality settings to high performance and ran 3dmark11 once more, returning a blistering score of 4815, over 1200 points above stock performance @ 720/1800.
At this point i had forgot about sli performance so i was eager to see the difference, but having never ran each card at difference clocks during SLI i was unsure if this could even be stable let along safe. But i had already committed to fact i was ready to loose 1 (or both) cards in the process of this testing anyway so i figured what the hell. I knew cards at different stock speeds could run together but i was unsure if with such high clocks and both cards running the max 1087 core voltage if it was still going to work. But so my surprise it did and return a nice score of 8221, 1000 points higher again than my previous sli score.
I also researched and attempted to raise the limitation of MSI Afterburners vcore cap asn well as flash the GPUs bios with a unlocked volt bios but this didn't seem to take. Some people reported their cards achieved up to P6000+ with volt mods (see 3dmarks11s score results) but for the most part running the normal cap of 1087, P4800 scores seem to be the cherry picked cards from premium vendors. I may in the future attempting again to unlock the voltage as im pretty confident in the 1 card hitting 1Ghz core with a vcore of 1100 or so.
In conclusion, i feel i managed to score a cherry card and it will be sad to see it and the other go in the near future. But i will say im going EVGA for my next cards, clearly they can take a beating and still live to tell the story another day. Not bad considering its equal to or only 1k points off brand new top end AU$500+ cards.
Below is the 3DMark11 benchlog.
Test conditions:
Both GPUS are: 01G-P3-1370-KR's
stock air cooling ran at 100% during testing.
Idle/load temps are GPU#1 @ 28/49 & GPU#2 @ 27/48
Ambient temp of 10 degrees, directly piped from outside and blown directly onto both cards. (<3 cold weather)
Primary GTX460 SLOT #1:
3dmark.com/3dm11/2915879 - P4615 - 930/2000 - mV 1000 - (2nd card phx 930/2000)
3dmark.com/3dm11/2915905 - P4658 - 940/2000 - mV 1050 - (2nd card phx 930/2000)
3dmark.com/3dm11/2915921 - P4693 - 950/2000 - mV 1050 - (2nd card phx 930/2000)
3dmark.com/3dm11/2915944 - P4726 - 960/2000 - mV 1050 - (2nd card phx 930/2000)
3dmark.com/3dm11/2916025 - P4770 - 970/2000 - mV 1087 - (2nd card phx 930/2000)
3dmark.com/3dm11/2916085 - P4774 - 970/2000 - mV 1087 - (Primary card phx)
3dmark.com/3dm11/2916059 - P4775 - 970/2000 - mV 1087 - (2nd card phx 930/2000)
3dmark.com/3dm11/2916110 - P4778 - 970/2100 - mV 1087 - (2nd card phx 930/2000)
3dmark.com/3dm11/2916136 - P4789 - 970/2150 - mV 1087 - (2nd card phx 930/2000)
3dmark.com/3dm11/2916151 - P4791 - 970/2175 - mV 1087 - (2nd card phx 930/2000)
3dmark.com/3dm11/2916176 - P4800 - 970/2200 - mV 1087 - (2nd card phx 930/2000)
3dmark.com/3dm11/2916198 - P4802 - 970/2225 - mV 1087 - (2nd card phx 930/2000)
3dmark.com/3dm11/2916216 - P4802 - 970/2250 - mV 1087 - (2nd card phx 930/2000)
3dmark.com/3dm11/2916244 - P4800 - 970/2300 - mV 1087 - (Primary card phx)
3dmark.com/3dm11/2916271 - P4812 - 970/2320 - mV 1087 - (Primary card phx) (Absoluse max stable)
3dmark.com/3dm11/2922170 - P4815 - 970/2320 - mV 1087 - (2nd card removed from system)
FAIL FAIL FAIL FAIL FAIL - P NIL - 970/2330 - mV 1087 - (Primary card phx) (Green artifacting)
FAIL FAIL FAIL FAIL FAIL - P NIL - 970/2340 - mV 1087 - (Primary card phx) (Afterburner max setting)
2nd GTX460 SLOT #2:
3dmark.com/3dm11/2915766 - P4642 - 930/2000 - mV 1087 -(2nd card phx)
SLI Score:
3dmark.com/3dm11/2916317 - P8221 - 970/2321 & 930/2000 - mV 1087 - (Primary card phx)
Facebook
Twitter
Instagram