I encountered this problem as well. I thought the card was faulty. However the problem disappeared after I updated the video driver, so I guess it was a problem with the default Windows 7 driver.

Errius wrote on 2022-04-13, 15:37:

I encountered this problem as well. I thought the card was faulty. However the problem disappeared after I updated the video driver, so I guess it was a problem with the default Windows 7 driver.

Thanks! That is exactly the problem I've noticed (mostly in web browsers) but can't duplicate the conditions where/when/why that happened. I know I tried the card on two different NM10/ICH7 boards and same things happened and both in Win and Linux NV proprietary drivers 390.x latest version. I was thinking to have bought a faulty card but I suppose this might be some compatibility problem. Can you give me some technical info about your revision card? The bridge chip is the Pericom PI7C9X? And what about the bios performance levels? And which drivers did you tried? Also which vbios do you have?

I'm sure this didn't happen with a PCI-EX GT610 while they both share the same analog VGA noise on both card outputs but that is not a big problem. I've tried running now the Unigine Heaven heavy test and nothing happens.

I have two of these cards, a ZT-60604 and ZT-60604-10L which appear to be identical apart from the model number. BIOS for both is 75.19.1B.00.01.

They both had this problem when they first came to me, and in both cases it disappeared after updating video drivers. (I also replaced the old thermal paste and, as they both run hot, I had an external fan blowing on them as well. However, I don't believe the problem was temperature related.)

Test rig where I noticed problem was GA-X58A-UD5 machine running Windows 7 Professional 64-bit with default Windows video driver 344.11, which was then updated to latest Nvidia driver for this system, which at the time was 391.35.

These cards were later also tested in Windows XP Professional 32-bit on same machine running the latest Nvidia driver for that OS, and no problems were noticed.

Cards were extensively tested in 3DMark03, 3DMark05, 3DMark Vantage and 3DMark 11. Problem did not recur.

I have a picture of one of the cards with HS removed

Errius wrote on 2022-04-13, 16:32:

I have two of these cards, a ZT-60604 and ZT-60604-10L which appear to be identical apart from the model number. BIOS for both i […]
Show full quote

I have two of these cards, a ZT-60604 and ZT-60604-10L which appear to be identical apart from the model number. BIOS for both is 75.19.1B.00.01.

They both had this problem when they first came to me, and in both cases it disappeared after updating video drivers. (I also replaced the old thermal paste and, as they both run hot, I had an external fan blowing on them as well. However, I don't believe the problem was temperature related.)

Test rig where I noticed problem was GA-X58A-UD5 machine running Windows 7 Professional 64-bit with default Windows video driver 344.11, which was then updated to latest Nvidia driver for this system, which at the time was 391.35.

These cards were later also tested in Windows XP Professional 32-bit on same machine running the latest Nvidia driver for that OS, and no problems were noticed.

Cards were extensively tested in 3DMark03, 3DMark05, 3DMark Vantage and 3DMark 11. Problem did not recur.

I have a picture of one of the cards with HS removed

I've got the ZT-60604-10L built in 2014 and I think the same bios version (EDIT: it's confirmed bios is 75.19.1b.00.01). When I noticed the problem anyway I think I was using the latest driver in Win 8.x even if I'm not entirely sure it might have been also the 390.x version which I'm sure I've tried it on Linux. Now I'm testing it with older 340.x driver to see if this happens again with these in Linux.

I agree it might not be something related to temps, while I was thinking I might use an active cooler but I suppose the power demand of this card is already at the PCI limit and that 1 or 2 watts more might be a problem. In Win I did try to change the GPU clock and it can be lowered but not the gpu voltage or the memory clock. It's interesting that in Linux with the open driver it try to boot the card @ the lower 0,90 volts but the o.s. randomly crash at that voltage I suppose cause the clocks are high for that value and fixed (810/666Mhz).
The GPU indeed run at very high temps but the passive heatsink is imho too much heavy for this card and also quite boring to install when changing the paste. With a low power Atom the GT610 gpu anyway runs from 38°C to 70°C depending on the 3D stress and with a low speed fan above it on the case.

Passive video cards make me nervous. Even if they are designed to run hot, I always mount a fan near them to keep things cool. This will extend their lives too.

Errius wrote on 2022-04-13, 17:12:

Passive video cards make me nervous. Even if they are designed to run hot, I always mount a fan near them to keep things cool. This will extend their lives too.

I agree too that these cards seems like quite too much for any passive heatsink design and same thing for the passive PCI-EX GT610 that without a case fan and with the opened case on Unigine Valley test I've seen the internal gpu temps reaching 100°C.

Anyway I am a bit less worried now that I know other cards display this problem too and I hope it's not a bad sign but more related to the driver not probably designed for this strange and very specific PCB layout. Cause reading around some user opinions I read about some of these cards didn't last much for some and I wonder if there might be a problem in the hw design.

Do your cards automatically change clocks/voltage or just like my card has only one performance level?

Mine run at run at 810/1620/1334. I didn't check voltages. I assume if they run too hot, then the speed will be reduced, but I never encountered this issue with these cards. (I did have this problem with a GeForce 9500 GT which had improperly installed HSF.)

Errius wrote on 2022-04-13, 17:32:

Mine run at run at 810/1620/1334. I didn't check voltages. I assume if they run too hot, then the speed will be reduced, but I never encountered this issue with these cards. (I did have this problem with a GeForce 9500 GT which had improperly installed HSF.)

I suppose it should have some automatic safe temp protection but usually these card has also the classic NV PowerMizer control panel tool to set the Adaptive or Performance config performance levels but not this card; there are no steps but a single one. The only GT610 PCI-EX x16 card I've got has two performance levels (0,1,..), even the Geforce 210 PCI has three performance levels and when in idle it clock down to the first low clock/voltage and increase to the highest once needed. I wonder if the GPU was designed to run at this value only "on demand" more than the fixed value we see here.

I just left everything at stock speed and didn't try changing anything.

Errius wrote on 2022-04-13, 17:57:

I just left everything at stock speed and didn't try changing anything.

I've left every settings at default too but I was expecting it to slow down its clock in idle like other cards. But I wonder it might be a reading problem of the control panel about the actual real time clocks who knows, cause reading the total system power watts the values are more or less similar in idle/stress to the Geforce 210 that automatically at default slow down quite a lot its clocks (like to 135Mhz for the GPU) in idle. Anyway thank you really, this helped a lot to consider again using this card I was thinking faulty but I'll try to search more info about this and update here. 😉

I've a few of these Zotac boards (both the 512MB & 1GB versions) in some older server systems - seem to work OK in that environment.

Apparently there are some clone boards out there as well, based on the same Zotac PCBs / BIOS - https://dogemicrosystems.ca/wiki/GT610_PCI

Yes it's interesting that no other bios were working beside the original one. But probably as I suspect from the different performance levels driver settings I suppose it is optimized for the PCI bridge chip solution. Chip that as tested in the past is different from the usual PLX chip on other cards. There were also other chips to do this task and each should have different features/speed while probably in their times not anymore oriented to vga but more less demanding cards. Latency and memory bandwidth usage seems like quite a complex task involved in such different bus. And looking not necessary at the frames rate but more the syntethic benchmarks it seems like a big impact can be seen on the polygon/s values which in my config stop at 10Mt/s in both cards I've got with two different bridge ICs. And I think I've seen the same value on a much more powerful Core2 3,3Ghz board.