What's the smoothest way to replace an older nVidia card with a newer one?

[BonzaiDuck]

the low power modes are great with these new cards, around 15W at idle.

Asus Tweak is for overclocking your GPU.

Have you tested the cards performance with benchmarks/games yet?

I have 3DMark Vantage on this system, and probably bought the nominal license and installed it in mid- to late-2011. So I am sure there are updated revisions.

After installing GPU-Tweak (Asus) and GPU-Monitor, I was stunned that the temperatures on the graphics card were so low -- the fans only running at 37% duty-cycle. This, of course, was with Lucid enabled, but the thermal and usage data on the ASUS software seemed likely to reflect the true state of affairs for the 780 GTX alone.

At this point, I wanted to goose the system with 3DMark and see what ASUS Monitor showed. Perhaps I did this with questions still open in my mind: How does it work with two monitors? How does it work with one monitor connected to the iGPU and Lucid enabled?

It ran the first test in 3DMark, and then the system froze. I would attribute this to the Lucid and multi-monitor setup. And I think that things might be different if I ran solely in "dGPU mode" -- perhaps with Lucid turned off. I chose the resolution setting in 3DMark that represents my desktop monitor -- 1680x1050. The HDTV, of course, is set to 1920x1080. I'm pretty sure I hadn't run 3DMark with both monitors connected in 2011, or for that matter -- with Lucid enabled. I'd only used it to test the dGPU after I'd finished tweaking my basic CPU OC settings back in 2011.

What part of 3D Mark I was able to watch firsthand, it seemed that the FPS record was pretty damn good -- over 100 FPS not matter what it did.

I'll want to explore this matter of 3dMark Vantage or similar benchmarks further, but with greater caution. You expect there will be BSODs and lockups when you first OC a system. I've been troubleshooting something more elusive: a freeze, random reset or occasional BSOD with the system at idle after running 24/7 for 10 to 12 days. I am beginning to suspect -- with preliminary conclusions -- that this problem had originated from my network-device Silicon-Dust HD HomeRun Prime. Neither the software nor the firmware had been updated since 2011, and the change logs at Silly-Dust suggested things like "fixes crashes . . . freezes" for earlier revisions.

I hadn't thought earlier to consider a "network device" that wasn't inside the box. But Silly-Dust has to work with Media Center. And the occasional freezes would occur when the TV was running 24/7.

I did everything one could think of: install/update every single driver I could think of (except for SiliconDust -- done as of yesterday), needlessly replaced the RAM, reviewed and re-tweaked all the clock settings and profiles -- with new stress-tests.

I also discovered today that the fan for the old GTX 570 had been clogged with kruft. Never mind that it still ran, or that cooling to the card improved when I used NVidia Inspector to change the fan's duty cycle. I could get the card to cool down to around 40C. I had thought that this was the status-quo when I first built the system. So I also began to suspect today that the dGPU's thermal situation may have had something to do with the occasional problem.

I had done everything humanly possible so far to resolve the very occasional instability. The only thing that will tell me that it's cured will be constant 24/7 running over two weeks or more. But the point of it: no matter how often I rebooted the system or just left it running 24/7 with no reboot -- this problem would occur only after a time span of a week or more. No difference. So I'm wondering if it had something to do with IP addressing for the SiliconDust device, the firmware or software. Brief review of the Silly-Dust web-site raised those sorts of issues in the change-logs.

Not only that, but there was a period late last year when the system ran with no instability for a month 24/7. It was a time when I wasn't running Media Center (hence, HD HomeRun Prime) -- getting the TV feed instead from my basic cable coax connected directly to the HDTV. I was working on "business stuff" that month, but I would still pause here or there to play games so I could "relax." So it wasn't the games. It seemed to occur only with Media Center feeding my AVR/HDTV.

 

[FalseChristian]

When I threw in my new EVGA GTX 760 2GB from 2 EVGA GTX 460 1GB SLI I had to uninstall the drivers and then reinstall the same drivers. Call me crazy.

 

[FAQdaworld]

I always uninstall with DDU,
Then go in with the newest drivers...

If it's been a while (6 months or so) I take the nuke it from orbit approach and do a fresh OS install... But I'm on RAID0 so it just makes sense (its the only way to be sure)

 

[tweakboy]

Just drop the card in the socket and reboot right into windows. You dont need driver cleaner for nVidia. Launch latest WHQL nvidia drivers and your set to go. gl

 

[BonzaiDuck]

I always uninstall with DDU,
Then go in with the newest drivers...

If it's been a while (6 months or so) I take the nuke it from orbit approach and do a fresh OS install... But I'm on RAID0 so it just makes sense (its the only way to be sure)

Well, the plot thickens -- something I've said a few times here in the last couple weeks.

I had my boot graphics adapter and primary monitor set up as the HD3000, enabled for Lucid. Then, in Windows, I had Lucid enabled. And I installed, from the ASUS disc that came with the GTX 780, the "ASUS Tweak" suite, which includes the graphics monitoring software, the graphics equivalent of AI Suite II's "Turbo EVO" for the CPU, or "Asus Tweak" for overclocking the graphics card in Windows. And it includes a version of GPU-Z. All these components have "skins" which look much like the "ROG" monitoring components.

I had an earlier version of GPU-Z ["TechPowerup"] installed, so clicking the "Info" button on "Tweak" would cause that version to open instead, and the software would crash. Only the software would stop working after the first episode -- wherein the entire system would freeze and crash.

So I uninstalled everything -- "Tweak," previous version of GPU-Z, downloaded the latest release of "Tweak," and reinstalled.

But I also moved all monitors off the iGPU to the dGPU. Under Lucid, the P-state seemed to lock on P8 -- a low power state. Now, with Lucid "off," it varies more dramatically from P2 to P8, given demand on the dGPU. which now shows 30% usage. But because the VRAM is now closer to triple that of the old GTX 570 and the dGPU is so powerful, there is plenty of the card's resources available for just about anything.

The gaming plays slightly better on the ASUS / NVidia card than it did through the iGPU and LUcid. Power consumption seems to follow along at mostly 15 watts, but spiking here and there to 30+watts, again according to demand. Clock speeds for the dGPU seem to follow along -- also on demand. And -- after installing the driver update for my HVR-2250 tuner-card and sorting out this minor software mess, things are rock stable again.

Given this state of affairs, I'm wondering if I shouldn't just let the iGPU languish unused. I had hoped that plugging in my HDMI->AVR->HDTV to the motherboard HDMI port would provide an appropriate use for the iGPU, but Intel HD3000 won't activate the TV connection unless (maybe) I used the VGA or DVI-I port on the motherboard.

Perhaps I might have done some troubleshooting with that, but it was easier to reconnect everything in dGPU mode. Meanwhile, the tech-support at FutureMark gave me a link to the latest version and instructed me about using my dated license key with it. I'll save that for tomorrow. I'm going back to sleep.