EVGA 460 2Win, Lucid VirtuMVP, HyperSLI, BIOS

[mickleby]

A couple fairly general questions, I think:

(The JavaScript? popups to insert LIST isn't working nicely with OSX Chrome, methinks)

* BIOS "letterboxed" via mobo HDMI port
* How can I tell if SLI is working?
* How can I tell if Virtu is working?

When I connect to the gfx card the BIOS is fully visible, whereas through the ASUS P8Z77-I motherboard HDMI port I see maybe 4x3 ratio with the sides and bottom lost and unviewable. Luckily all the crucial settings can be manipulated, but I need to connect to the gfx card to completely use the BIOS. I guess Windows loads a driver that understands the issue, because as soon as the Windows video driver loads I see a full aspect ration and the screen borders are at the edge of the display. It's a Samsung 720p TV LN40A450C1D, if that is useful. At some point I will test the DVI port on the mobo, but I lack the cable so far.

This is my first experience with SLI, so an early thought was to use the EVGA Precision gfx monitor tool to see if the GPU2 core was getting load. It wasn't. Then I ran 3DMark11 and saw the score was on par with a single GTX 460, so I started hunting. I found someone who recommended HyperSLI. With HyperSLI installed I can access SLI settings in the NVIDIA Control Panel.. Cool. Did I mention it's an ASUS P8Z77-I with only a single PCIe slot? I have no idea what HyperSLI is doing (as yet), but it seems to be helping.

So then I tried to get VirtuMVP, SLI, and 3DMark11 all playing nicely. Here, I'm stumped so far. The best scores are connected to the gfx HDMI port, Virtu disabled.

* Connected to gfx HDMI with Virtu enabled 3DMark11 changes the screen resolution, but shows only a mouse pointer on a black background. I need to click the background then press Esc or it just sits there.
* Connected to the mobo HDMI with Virtu enabled 3DMark11 runs just fine, but SLI doesn't work, so I am using only 1 GTX460 and the HD4000. Not too much better than the single GTX 460.

[Activate: AMD]

I can't offer you any direct experience with Virtu or the bios issue, but as far as the SLI goes, it doesn't appear that you have SLI enabled. You shouldn't need HyperSLI, the nV control panel should recognize two GPUs and allow you to enable SLI without it. The 460 2Win is basically just two GPU's with an on-card bridge rather than the two cards with the bridge cable, so you still need to make sure that the software is set up. Be sure that you have them set to both render the game, and not one set as a dedicated PhysX card.

[mickleby]

I appreciate the reply.

NVIDIA Control Panel does not present SLI settings unless HyperSLI is running.

BTW, I am researching acceptable load temperatures. I have 85C/86C on the GTX 460 2Win, only the fan speed has changed 60min using Precision Stress Test. I have 73C/70C/69C/69C cores 73C pkg on i7-3770k 4.46GHz 100% load 60min. I have SYSTIN 55C CPUTIN 54C PCH Diode 69C on ASUS P8Z77-I same load same 60min.

Everything but the i7 seems high, but I'm n00b. I read the GTX 460 runs hot, but read differing numbers about what is normal under load. I read mobos should look to stay around 5*C above ambient, but can that be under heavy sustained load? I read to look into a NB chip fan for +40C?

I am happy the temps seem stable. They peaked about 15min into this test, and haven't risen a single degree on anything. Only the 2Win fan moves up/down a few score RPM.

----
EDIT: I tell a lie. Apparently my memory o/c isn't quite stable so my Prime95 Torture Test was not fully running for most of that test. I'll update with new numbers when I get stable o/c.

[Activate: AMD]

Ok, I think I have a slightly better handle on whats going on here. seems like HyperSLI is necessary because the nvidia driver is giving you a hard time due to the motherboard not being an SLI-certified motherboard. Its obnoxious that they would do something like that, but it sounds like classic nV. HyperSLI is basically a hack to enable SLI on non-SLI certified motherboards by exposing the SLI controls in the driver.

My guess is that Virtu is not going to play nice with that whole thing since its not a straightforward setup. I'd advise you that its not really worth the hassle. The performance boost from Virtu with a card of the 2Win's caliber will be negligible, but you'll probably lose the ability to use Intel QuickSync. If Virtu is disabled, SLI is enabled, and you go through the GPU's output, do you get load on both GPU cores?

Your temps seem alright. Are you in a small case with that motherboard?

[mickleby]

If Virtu is disabled, SLI is enabled, and you go through the GPU's output, do you get load on both GPU cores?

Your temps seem alright. Are you in a small case with that motherboard?

I think the answer is yes, tho I'm saying GPU ouput == gfx card output. I can't remember the result of mobo output with HyperSLI. But with HyperSLI, SLI can be enabled, and actually it's a bit more rosy than you describe because the PhysX can go through the iGPU effectively yielding "almost 3-way" SLI.

My 3DMark11 score this way, with the gfx o/c is up to P7900, which seems like more bang than this buck should be worth

RE temps, I got a reply from ASUS re temps. It was sort of ambiguous but agrees that my temps are fine. There is a discrepancy between the ASUS Sensor CPU Temp and the temp from AIDA64 or CPUID HWMonitor. ASUS rep says CPU temp should stay below 75C, so I assume he means the ASUS sensor because clearly the i7-3770k withstands much more.

I'm learning how to manage temps via playing with the voltage calibration. Right now I have the CPU pegged with Prime95 and the gfx pegged with the EVGA Scanner OC Stress Test. The CPU temps run around 90C at 4.6+ GHz without the gfx, a number I feel happy with. When I add the gfx stress the internal temps go up and the H100 set medium cooling system cannot dissipate heat fast enough. The CPU temps go up to Tj and the i7 nicely attenuates core ratio to keep the thermals in range, i.e just under Tj at 104/105C.

This is clearly not a real world situation. I think I will run a sensor recorder when playing heavy load games to see what readings I max at. I will also consider upping the H100 to max, but the lousy Corsair fans it came with are bloody noisy at those rpm. All in all, this is a good overclock.

Thanks for your help!

[mickleby]

To gush a bit:

I LOVE the way the CPU Frequency pops right back to 4645MHz when I drop the gfx stress test!

Already the CPU temps have dropped 6C in just about 60sec. I think I will leave Prime95 running for the 12h I'm at work. Normally, I would prefer CPU temps under 80C but I think it's ok to take the new CPU for a spin. I've been breaking in the i7 for about 10days or so.

CPU Freq 4645MHz
BCLK 103.2
Ratio 45
VCore 1.376
temps 79-86C

RAM 963.MHz (1926 on 1333C9 rated XMS3, not shabby) 1.6v

[mickleby]

So today I exchanged the VGA Ext cord I mistakenly bought for a VGA cord, and I am back to report on the ASUS BIOS to my Samsung TV issue.

Yes, I can see the entire BIOS using the motherboard DVI port. Remember, when I use the HDMI port on the mobo the BIOS screens are obscured on 3 sides by black "letterbox" like cropping.

However, the BIOS display is not so well centered as the display when coming through the NVIDIA ports. I checked the TV and any adjustment is grayed out and not available. That's not much of an issue.

More of an issue is the fact that the image is rather washed out. Even with the Windows drivers the picture is faded a bit, though the Windows driver does seem to center the image properly.

Is this signal loss in the DVI to VGA cable? It's only 6' and not passing any electrical interference. It is running about 3' along the metal case of the dishwasher though. I'll probably check all the pins and firmly reseat all the connections at some point, but the pizza has just arrived

[mickleby]

I just noticed I had mentioned the RAM overclocking.

I get a nice stable 2026MHz 9-9-9-24-1 at 1.6v I'll work on lowering the voltage at some point, but it's not seeming to cause any thermal issue so far. I'm very pleased with this $40 2x4GB 8GB kit.

[mickleby]

So I tracked down the Lucid / SLI situation: http://www.futurelooks.com/lucid-mvp...md-crossfirex/

Seems it doesn't work by design.

My experience is SLIGHTLY different than this article, in that I could access VirtuMVP by plugging into the iGPU motherboard port however this disables SLI and Lucid thinks there is only a single GTX 460--which actually I guess is true without SLI. I did tests showing that only dGPU-A of the 2Win and the iGPU were being utilized. MAYBE this opens some small chance of a hack emerging from somewhere before the next iteration of Virtu, apparently to coincide with Haswell release.

[WillRock]

The 4602Win (and the 5602Win, both of which are the same card since the 460 is pin compatible with the 560) are notorious for these issues. It's because the card isn't natively recognized by nVidia drivers (there's no 4602Win in the driver microcode). It was a cheap solution for eVGA/nVidia to get rid of their GF104 stock.

I'd just sell it while it's still worth something, and get a proper single GPU like a non-ref 660 Ti.

[3dfxrain]

I must say I was dissapoint about being told it didn't work with SLI officially, this was before the thread started. Maybe it works later. At least it works with one main graphics processor on the card.

[Activate: AMD]

So I tracked down the Lucid / SLI situation: http://www.futurelooks.com/lucid-mvp...md-crossfirex/

Seems it doesn't work by design.

My experience is SLIGHTLY different than this article, in that I could access VirtuMVP by plugging into the iGPU motherboard port however this disables SLI and Lucid thinks there is only a single GTX 460--which actually I guess is true without SLI. I did tests showing that only dGPU-A of the 2Win and the iGPU were being utilized. MAYBE this opens some small chance of a hack emerging from somewhere before the next iteration of Virtu, apparently to coincide with Haswell release.

I don't see any reason to be too upset about this. Virtu doesn't do much in a "hybrid" situation with a single GPU, and trying to balance the load between SLI + iGPU would probably be even worse. There is really no reason to use it short of deliberately injecting another potential source of error and incompatibility or doing quite a ton of encoding using Quick Sync.