Geforce GTX460: Overclocking and 3DMark results

[cowie]

goodluck with the cards GT
should make a good gamig rig if thats what floats your boat

NCspecV81 helped me get 40 more on the core
big thanks
here 1.14v repeatable runs of 03 and 06
quick test
[/URL]
after 1.18v real it blackscreens

[GoldenTiger]

Doh, that's still one hell of a clock Cowie . I had it in my head that I'd get higher than I have with my cards in SLI... so I feel disappointed even though it games amazingly well . My main purpose for them is gaming, so it works nicely... secondary is development, with benchmarking being about as far back as you can go other than to compare my own rig to itself vs. the single 470 I had.

[NCspecV81]

Have you tried the MSI bios or just the evga one?

[GoldenTiger]

If you meant that question to me (not sure if it was addressed to cowie or not), I tried the MSI one only.

[xdan]

I first tried sandwiched (note that the fans actually end up taking the full 2 slots even though the pictures don't look it) but the top card was running a full 25c higher than the bottom (around 55c vs. 80c top/bottom during a game). So, I am trying them with the x4 speed, and the temps are virtually identical on each card (within 2-4c now) at full load after gaming awhile.

The good news for me is they ARE very quiet.... I have it at 82% fan speed and it is as quiet as the rest of my case fans for this Gigabyte 1GB pair (dual fans). Color me impressed on that. The other good news is that they run like gangbusters in the only game I've tested so far, BattleField Bad Company 2. Also, I was able to flash them with an MSI BIOS that was modded to allow up to 1.212v (thank you NCspecV81 of XS) instead of the 1.087v normal cap in Afterburner.

Running at 2560x1600 8x CSAA, 16x HQ Anisotropic Filtering (AF), max in-game settings with HBAO on in DX11 mode *AND* Transparency MSAA enabled in the control panel, my overclocked Gigabyte 1GB GTX 460's in SLI are pulling off around 55minimum (usually 60), and 75-80 average FPS for me across the different maps of BFBC2. This is, mind you, with them running at PCI-E 2.0 x4 instead of the full x16, which means around a 10% performance loss from what I have read.

The bad news is I can't nail down a stable overclock yet like I was hoping to that is sky-high. So far I am running 880mhz core on both cards with 3900mhz on the memory... not quite as high as I was looking for, and this is with 1.100v on the cores. They currently max out at around 76c for the pair at these speeds after an hour+ of gaming in BFBC2 with the 82% fan speed (note again that it REALLY is quiet, these fans don't get loud at all ).

1. Try each card single on PCIE 16X to see what it can do each card on their one.
2. I think that the 4X electrical slot cand reduce the capability of overclocking, i mean at 4x i think the power through the slot is reduced too so instead the card might used more power from the power conectors and it might have a negative efect.
3. From what is known by now all chips can do 830-840Mhz/4000(core/memory) with stock voltage. 1.1v it's much for 24/7. Try 1.06v it's more acceptable. Try to use more fans in the case(than you have allready)
Unfortunatelly i think that when they are sandwhiched the bottom card intake all the hot air from the first card...

[GoldenTiger]

Xdan, your thoughts are very possible and interesting... I'm trying them sandwiched again and will up the fan speeds (which I hadn't the first time), and then test them individually if the overclock does not improve. Especially #2 piques my interest as a possibility.

[GoldenTiger]

Thanks Xdan... I tried both of them individually, one is crashing at stock speeds when played with for awhile while the other is already going 875+ core with 1.037v without issue and 3850 mem. I'm RMA'ing the defective one... replacement would be here tomorrow but they're out of stock for cross-ship right now.

The good card has been working at 922core/4000mem with 1.100v, will tweak it further then try SLI again once my replacement comes.

[BETA911]

1.1v it's much for 24/7. Try 1.06v it's more acceptable. Try to use more fans in the case(than you have allready)

I don't think +1.1V on a 40nm GPU die is much for 24/7. I'm not 100% what the stock voltage was on these card under 3D mode, but 1.1 is easy in mine opinion.

[Ket]

Dang that sucks I'll have a GTX460 to play with in a few days. Any chance you could upload that MSI vBIOS you have Golden?

[xdan]

I'm RMA'ing the defective one... replacement would be here tomorrow but they're out of stock for cross-ship right now.

Can you replace it with another brand of GTX 460? I think that you would have better temperatures with one card as reference, because reference cooler drops less hot air inside and much more outside the case than Gigabyte cooler.
Using the reference card as top card and the Gigabyte as bottom i think can improve the temperatures in SLI. And from lab 501 the reference cooler( 60-62C with fan at 60%) is better even than Gibabyte's with 4-5C.
On neweeg Sparkle 1gb reference is on stock.
http://www.newegg.com/Product/Produc...82E16814187113

The fins are oriented to conduct the air outside the case and the pass from thin fins and more dense fins to larger fins and less dense has the role to make air turbulence good for cooling.

[GoldenTiger]

The BIOS I have was edited by NCspecV81 and can be downloaded from his post over on OCN as an attachment: http://www.overclock.net/nvidia/7819...ap-unlock.html

I'm both happy and disappointed... because I couldn't fully stabilize any real overclock with one of the cards being defective (which I'm disappointed with it coming in bad), but I'm happy that I figured out the issue and will get more speed in the end plus a fully stable system .

EDIT: Yeah, Xdan, I could do a different brand... prefer to have two matching, though, and I'm unsure of Sparkle's warranty quality... tempting though.

[xdan]

The waranty is lifetime, of course that i would have recomended Evga EE(all air external exhaust) but it seems that from 2-3 days there are not any Evga.s GTX 460 in stock.

[Ket]

The BIOS I have was edited by NCspecV81 and can be downloaded from his post over on OCN as an attachment: http://www.overclock.net/nvidia/7819...ap-unlock.html

I'm both happy and disappointed... because I couldn't fully stabilize any real overclock with one of the cards being defective (which I'm disappointed with it coming in bad), but I'm happy that I figured out the issue and will get more speed in the end plus a fully stable system .

EDIT: Yeah, Xdan, I could do a different brand... prefer to have two matching, though, and I'm unsure of Sparkle's warranty quality... tempting though.

Champion If there was a cookie system in place, you good sir would earn a cookie from me

[GoldenTiger]

I mean as far as how their actual customer service is... they aren't a well-known brand in the US . I also know how two of the Gigabyte ones sound in SLI already and really liked the silence, plus it's better for future resale for next-gen to have 2x matching.

[xdan]

Than just raise the fans speed to 70-80% if it doesn't disturb you.

[GoldenTiger]

Champion If there was a cookie system in place, you good sir would earn a cookie from me

Welcome ... FYI the modded BIOS, at least on my Gigabyte card, allows me to change fan speed inside of Afterburner from 0-100% with proper, full functionality .

[Ket]

Sweet. Deffinately going to have more fun OCing than what I thought would initially be possible thanks to the modded vBIOS

[revenant]

looks like Nvidia has a Celeron 300a here.

[Ket]

You cant say Celeron next to a 460 you'll make it cry Even though back in the day I saw quite a few 300 cellys hit some insane speeds

[GoldenTiger]

looks like Nvidia has a Celeron 300a here.

Funny thing is, I was thinking of making a thread for my OC result with the something similar to the following title right as I was testing my pair of 1GB 460's:

"nVidia GTX 460A: The celeron 300a of video cards ringside show!"

My non-defective card is sitting very happily at 920mhz core/4000mhz ram so far, haven't upped voltage further than 1.100v for that and it's rock-stable. It's loading at around 72c, so I probably have a bit more room if I really wanted. Still, I think I'll leave it at that for now likely and try to get my 2nd card for SLI to those speeds once it's arrived... then push harder if I still have thermal room.

[Ket]

I'd be interested to know if increased GPU voltage helps with memory OCs. One thing I'm concerned about with the 460 is it being bandwidth stricken. I've noticed it a lot the mem not OCing well and if I'm not mistaken with nVidia GPUs (I'm rusty) the IMC is integrated on the GPU, correct?

[GoldenTiger]

I'd be interested to know if increased GPU voltage helps with memory OCs. One thing I'm concerned about with the 460 is it being bandwidth stricken. I've noticed it a lot the mem not OCing well and if I'm not mistaken with nVidia GPUs (I'm rusty) the IMC is integrated on the GPU, correct?

It did with my GTX 470... voltage helped raise core and/or memory, but there was a definite correlation where one would cause the other to need to be lowered to become stable again... and yes that is correct as far as I know, the IMC is integrated on the GPU for nV chips.

[Ket]

There we have it I say. Do some indepth playing with core/shader/mem clocks at various voltages. I'd dare say a card that got fitted with good vRAM and given a voltage bump is quite capable of outpacing a stock 470

[GoldenTiger]

You got me curious, Ket... so I dropped core to 900mhz, threw RAM up to 4400mhz, and for about 15 minutes it's been stable so far... definitely a performance boost over 922/4000, too . I'll do some benching later if it stays stable.

EDIT: 1.112v seems stable at 900c/4400m, 35 minutes of BFBC2 so far, will run it thru awhile and make sure all's well with Kombustor/etc. later. If it stays stable I'll do some benches.

[GoldenTiger]

So I did some Just Cause 2 testing on "The Dark Tower" bench to see how clocks affected things. The benchmarks each last 2 minutes. Here's some data.

TEST RIG: i7 920 @ 3.72ghz HT enabled, 6GB DDR3 @ 8-8-8 2T 1563mhz triple-channel, 256GB SSD (Indilinx Barefoot w/ Samsung NAND), X-Fi XtremeGamer, Gigabyte GTX 460 1GB.

All tests @ 2560x1600 with 8x CSAA w/ Transparency, 16x HQ AF, maxed in-game settings with Bokeh filter off.

CORE ---------- MEM ---------- FPS
_______________________________
900 ---------- 3600 ---------- 23.98
900 ---------- 3900 ---------- 23.95
900 ---------- 4200 ---------- 24.48
900 ---------- 4400 ---------- 24.25

940 ---------- 3600 ---------- 24.63
940 ---------- 3800 ---------- 24.84
940 ---------- 4000 ---------- 25.05

For grins, 940c/3800m @ 2x AA resulted in a framerate of 27.17/second.

Well... I'm sort of surprised. For the 940 tests I raised voltage to 1.125, for the 900 tests all ran at 1.112v. It seems that I may be CPU bottlenecked, since the results are all SO close together. I may try dropping my CPU speed to something like 3.2ghz and see how things are affected... or Vantage on Extreme pre-set (I hate how long it takes ). Temperatures peaked at ~74c during a long session (1hr) of BFBC2 @ 1.112v and 900c/4400m, high fan percentage but still very quiet (just a light, gentle whoosh of air, I can barely hear it above my other fans which are ~25-27db ones, this is at 92%!!).