- Sep 18, 2024
- 42
- 2
- 35
Yesterday I got an Asus RTX 5080 card. I played games for hours, the only one that crashed was Forza Horizon 5, but I know that game has a bug and crashes with any card. It crashed when I had a 4060 Ti 16 GB, then it crashed with a 5070 Ti, and it crashes at random with the 5080 just the same. No other game crashed.
Besides that, I spent hours in Blender refining a model that is very heavy in everything, all fine. Then left that rendering overnight, and that went well.
But like every new component, I wanted to stress test it, so this morning I ran the VRAM module in OCCT. I left it running, and about five minutes later, it showed me that it got about 45000 errors. I wasn't happy at all, since I bought it straight from Asus and it wasn't going to be an easy exchange like with Amazon, especially with Asus using Fedex.
But I remembered that in GPU Tweak (the Asus app to configure and monitor the card), I had it set to the OC profile. This profile is a joke as far as OC goes, because it just changes the clock from 2700 to 2730 Mhz, and the power target from 100 to 112%. But I set to the default profile, and rebooted the computer.
I ran the VRAM module again, and after 20 minutes, I had no errors. So I wanted to see what happened if I ran the test again with the OC profile. I opened GPU Tweak, set it to OC and rebooted. Then I ran the test again, and this time I ran it for the full hour it's set to. No errors.
So it seems to me that the card is perfectly fine. If the VRAM was faulty, there would errors in every pass, correct? Since there aren't, it might seem that something corrupted the VRAM during that session, and when I rebooted and the VRAM was cleared, then the corruption went away. Is that a fair assessment? Or is OCCT just not too reliable and gave these errors when there were none?
If OCCT is not 100% reliable, what other tool can I use to test the card properly? So far I have used Furmark latest version, Superposition, and several of the Blender demo projects.
Besides that, I spent hours in Blender refining a model that is very heavy in everything, all fine. Then left that rendering overnight, and that went well.
But like every new component, I wanted to stress test it, so this morning I ran the VRAM module in OCCT. I left it running, and about five minutes later, it showed me that it got about 45000 errors. I wasn't happy at all, since I bought it straight from Asus and it wasn't going to be an easy exchange like with Amazon, especially with Asus using Fedex.
But I remembered that in GPU Tweak (the Asus app to configure and monitor the card), I had it set to the OC profile. This profile is a joke as far as OC goes, because it just changes the clock from 2700 to 2730 Mhz, and the power target from 100 to 112%. But I set to the default profile, and rebooted the computer.
I ran the VRAM module again, and after 20 minutes, I had no errors. So I wanted to see what happened if I ran the test again with the OC profile. I opened GPU Tweak, set it to OC and rebooted. Then I ran the test again, and this time I ran it for the full hour it's set to. No errors.
So it seems to me that the card is perfectly fine. If the VRAM was faulty, there would errors in every pass, correct? Since there aren't, it might seem that something corrupted the VRAM during that session, and when I rebooted and the VRAM was cleared, then the corruption went away. Is that a fair assessment? Or is OCCT just not too reliable and gave these errors when there were none?
If OCCT is not 100% reliable, what other tool can I use to test the card properly? So far I have used Furmark latest version, Superposition, and several of the Blender demo projects.
Facebook
Twitter
Instagram