How to test a used GPU?

[throwaway60540]

I went to buy a used GPU but I'm not sure how to do it, what parameters, etc. Can anyone suggest me ways, software, etc to test used GPUs?

[Levent]

Play a game on it. Run 3dmark and see how it performs other systems identical yours. Monitor temperatures with GPUZ.

[throwaway60540]

5 minutes ago, Levent said:

Play a game on it. Run 3dmark and see how it performs other systems identical yours. Monitor temperatures with GPUZ.

What if I physically can't come? Can I ask the seller to record a video with the GPU clearly in the PC case, him showing the specs on the monitor itself, and then running some software to test it to confirm it's fine?

[Galf]

11 minutes ago, throwaway60540 said:

I went to buy a used GPU but I'm not sure how to do it, what parameters, etc. Can anyone suggest me ways, software, etc to test used GPUs?

- a 3dmark run, the free version on Steam is fine. Monitor temps with GPUz (Ideal temps may vary on architecture! So first research what's considered normal.)
- play a game that is heavy but well optimized, i.e. Cyberpunk, or Naraka Bladepoint if you want something that is f2p. Crank the settings obviously.

and also a visual inspection helps: check if the screws have been tampered with if you're buying with hand-to-hand exchange, ask why the gpu was opened if it looks like it was. Check that all fans spins freely and silently. Check as much as you can if the pads look fine and the cooler isn't crooked (easy mistake when reassembling a GPU is to over-crank some screws and then you bend the cooler, which can do all sorts of bad damage.

[throwaway60540]

2 minutes ago, Galf said:

- a 3dmark run, the free version on Steam is fine. Monitor temps with GPUz (Ideal temps may vary on architecture! So first research what's considered normal.)
- play a game that is heavy but well optimized, i.e. Cyberpunk, or Naraka Bladepoint if you want something that is f2p. Crank the settings obviously.

and also a visual inspection helps: check if the screws have been tampered with if you're buying with hand-to-hand exchange, ask why the gpu was opened if it looks like it was. Check that all fans spins freely and silently. Check as much as you can if the pads look fine and the cooler isn't crooked (easy mistake when reassembling a GPU is to over-crank some screws and then you bend the cooler, which can do all sorts of bad damage.

This is the 3DMark the seller has attached (it's a ROG Strix 1080ti), the seller is open that the card was used for mining for 1,5 years. Is this a fine score for a 1080ti and a G4400? It seems to be slightly lower than the usual graphics score for 1080ti but maybe it's because the system is heavily bottlenecked? The seller also has images of the gpu itself attached (I'll attach them later).

[Galf]

6 minutes ago, throwaway60540 said:

This is the 3DMark the seller has attached (it's a ROG Strix 1080ti), the seller is open that the card was used for mining for 1,5 years. Is this a fine score for a 1080ti and a G4400? It seems to be slightly lower than the usual graphics score for 1080ti but maybe it's because the system is heavily bottlenecked? The seller also has images of the gpu itself attached (I'll attach them later).

cut

Hey so without considering the gpu score itself as I'd have to deep dive to find a reply to that, I never owned a 1080ti myself... the seller seems pretty straightforward. Bear in mind it's an old gpu, with no DLSS support, that has been used heavily. 1.5 years 24/7 mining is A LOT of use.

So if it's REALLY cheap then consider it, otherwise at this point it's a pretty power hungry card that doesn't perform particularly good with modern titles

[throwaway60540]

9 minutes ago, Galf said:

Hey so without considering the gpu score itself as I'd have to deep dive to find a reply to that, I never owned a 1080ti myself... the seller seems pretty straightforward. Bear in mind it's an old gpu, with no DLSS support, that has been used heavily. 1.5 years 24/7 mining is A LOT of use.

So if it's REALLY cheap then consider it, otherwise at this point it's a pretty power hungry card that doesn't perform particularly good with modern titles

The 1080ti costs 150$. I'm fine with no DLSS because the most similar card to it (being RTX 3060) would have blurry image if I used DLSS on it (plus, FSR 3(.1)). I'll attach the pics of the card itself later. Is there any other software I should get the seller to run?

[kosturko92]

4 hours ago, throwaway60540 said:

The 1080ti costs 150$. I'm fine with no DLSS because the most similar card to it (being RTX 3060) would have blurry image if I used DLSS on it (plus, FSR 3(.1)). I'll attach the pics of the card itself later. Is there any other software I should get the seller to run?

Trust me bro, go with RTX 3060 if you can. Newer architecture, will support drivers for longer, you will easier re-sell it later, outperforms 1080ti in some titles, and most likely didn't mine ( minning can burst GPU's to the point of failing very soon after you buy it )

1080 Ti came out in 2017, 3060 came out in 2021, its huge diffrence

[Galf]

5 hours ago, throwaway60540 said:

The 1080ti costs 150$. I'm fine with no DLSS because the most similar card to it (being RTX 3060) would have blurry image if I used DLSS on it (plus, FSR 3(.1)). I'll attach the pics of the card itself later. Is there any other software I should get the seller to run?

up to you, but DLSS at its quality setting looks better than native at this point and it's at least equal even at 1080p.
FSR... I can't fathom why it's so bad. I WANT TO USE IT because I have a 3000 series gpu so no frame generation, but I swear to god FSR 3 in The First Descendant feels like DLSS in its first iteration on a 1080p screen. It's borderline unusable. 

[emothxughts]

From my own experience buying used cards, I'd ask the seller to send you a GPU-Z screenshot and videos of the card running benchmarks like 3DMark, Heaven and Superposition, as well as a Furmark fuzzy donut endurance run. If the card can run all these tests without artifacting or crashes, it's good to go.

 

Also worth asking the seller if the card ever crashed or artifacted or otherwise do funny things while it was in their possession.

[Stahlmann]

On 7/4/2024 at 1:16 AM, Galf said:

up to you, but DLSS at its quality setting looks better than native at this point and it's at least equal even at 1080p.
FSR... I can't fathom why it's so bad. I WANT TO USE IT because I have a 3000 series gpu so no frame generation, but I swear to god FSR 3 in The First Descendant feels like DLSS in its first iteration on a 1080p screen. It's borderline unusable. 

DLSS is still considered by far the best upscaling algorithm. At the "Quality" setting, it consistently delivers almost indistinguishable image quality compared to native resolution with TAA anti-aliasing. Even "balanced" is still very usable at 4K output resolution. But at 1440p or 1080p, I wouldn't recommend dropping below "quality". Still, with "quality" you'll get a 30-50% performance boost, basically for free. DLSS is downright magic, imo.

 

FSR on the other hand, even the newer FSR 3, is still pretty bad at anything other than 4K output resolution. Be it shimmering, performance uplift, ghosting, blocky textures, and detail reconstruction, it lags behind DLSS in every way.

 

So imo a used 3060 is the better choice for upscaling alone. Especially in newer games, having access to DLSS basically guarantees that the GPU will hold up better in the future. Some games also have lightweight enough raytracing implementations that you might even be able to get some of that eye candy, or you only turn on reflections for example. A 1080Ti will go to <10 FPS as soon as you turn on any kind of ray tracing.

 

If you buy used, you should be good to go with a 3D mark screenshot showing the score and a GPU-Z screenshot. Of course, the best case is that you can pick it up in person and test it in a game or two. If you have to have it shipped, no amount of benchmark screenshots and videos will 100% guarantee that it will work when the package arrives.