LuxCoreRender Forums

Hi everyone, and sorry if this question were be posted yet.
Absolute newbie here. My problem is: I have 2 GPU, 2 Radeon VII. Installing them, one slipped from my hands and fell down on the table with the down-side sounding terrifyingly... Then I starting worring if the video card were damaged, even partially.
A friend of mine recommended me LuxMark to do a test, but for me it's incomprehensible.
I've made 2 tests:

1st TEST (Benchmark).
Lumark 3.1, OpenCL Compliler Options ALL SELECTED, Complex Benchmark (4973K triangles).
On main window I selected ONE SINGLE VIDEO CARD per time (in order to see the scores of the two cards separately), then in menu Benchmark OpenCL Selected Devices Only.
RESULTS:
GPU 1.
OpenCL GPU: 7089
58434 different pixels, 9.91%
GPU 2.
OpenCL GPU: 7104
58373 different pixels, 9.90%

2nd TEST (Stress test).
Lumark 3.1, OpenCL Compliler Options ALL SELECTED, Complex Benchmark (4973K triangles).
On main window I selected BOTH VIDEO CARDS, then in menu Stress Test.
RESULTS:
After 170 seconds:
EngineIntersect-0, Rays/sec 37735K, Prf Idx 1.00, Wrkld 49.9%, Mem 1022.2M/16368M
EngineIntersect-1, Rays/sec 37869K, Prf Idx 1.00, Wrkld 50.1%, Mem 1022.2M/16368M
After 500 seconds:
EngineIntersect-0, Rays/sec 36188K, Prf Idx 1.00, Wrkld 49.2%, Mem 1022.2M/16368M
EngineIntersect-1, Rays/sec 37412K, Prf Idx 1.03, Wrkld 50.8%, Mem 1022.2M/16368M

As you can see, more time passes more the difference between percentages of two cards increases.

What does it mean? Is it possible, from these tests, to understand if one of the two cards is partially damaged?

Thanks in advance for your kind help.

Hi,

I would bet that if the card was damaged you would not even be able to boot the computer with it, or it would not be recognized.

By the results, both cards appear to be working as intended, do you see any artifacts or somthing like that?

Thanks for your reply alpistinho!
Up to now I don't see any particular artifact... I'd say no.
My fear is that, due to the shock, some element in the card could have being disconnected, for example a little heatsink or something like that, i.e. something that at the moment is silent/hide, but in a while could become determinant. Could be?

Yes, some mechanical issue like this could have happened, altough it doesn't seem that likely. These cards do not have many moving parts.

But I think you'll have to physically inspect you card to be sure of that. Maybe shaking it to see if there is anything loose there?

I tried shaking actually, and nothing SEEMED moving wrongly...
It could be useful to know the temperature of the two cards separately (to make a comparison), but in iStat Menus I have a unique value of GPU temperature, so it's not very useful...
Perhaps do you know a method/software that shows the separate temperatures of the two GPUs in Mac OSX? (Thanks)

adrjork wrote: ↑Sat Oct 26, 2019 1:48 am Up to now I don't see any particular artifact... I'd say no.

alpistinho is right, if the card was damaged there's no way you could run a benchmark on it for 500 seconds. That's long enough for it to reach its max temp.

adrjork wrote: ↑Sat Oct 26, 2019 1:48 am My fear is that, due to the shock, some element in the card could have being disconnected, for example a little heatsink or something like that, i.e. something that at the moment is silent/hide, but in a while could become determinant. Could be?

Every component on the card is either screwed on (in the case of heatsinks) or soldered on (in the case of chips/components). It's very unlikely that just dropping it on a table would cause them to become loose. If you were to hold it up near your head and drop it on a concrete floor, for example, well that will almost certainly break something. But I think you're good.

Edit: That's not to say the cards are indestructible and you should just chuck them around. You can crack the solder, or sometimes components might be sensitive to shock forces. But as you can run the benchmark on it, I think you should be fine this time.

By the way it's quite normal to see small variances in performance in a multi-GPU system. The reason is that your operating system is using one of the cards to render your display also, and that takes some small amount of processing power. There may also be some natural tiny performance variance from the manufacturing process.

Hope this helps put you at ease!

Yes, Martini, your reply helped a lot. Really thanks.

adrjork wrote: ↑Fri Oct 25, 2019 10:59 pm What does it mean? Is it possible, from these tests, to understand if one of the two cards is partially damaged?

If the efficiency of the heating dissipation is reduced, a progressive rise of the temperature could force the GPU to reduce its clock rate (and so its performance).

Run LuxMark for sometime and check if the GPU temperature is table (and if the GPU clock has gone down or its stable).

Dade wrote: ↑Sun Oct 27, 2019 5:10 pm Run LuxMark for sometime and check if the GPU temperature is table (and if the GPU clock has gone down or its stable).

Thanks Dade,
actually it's my 2nd test:
2nd TEST (Stress test).
Lumark 3.1, OpenCL Compliler Options ALL SELECTED, Complex Benchmark (4973K triangles).
On main window I selected BOTH VIDEO CARDS, then in menu Stress Test.
RESULTS:
After 170 seconds:
EngineIntersect-0, Rays/sec 37735K, Prf Idx 1.00, Wrkld 49.9%, Mem 1022.2M/16368M
EngineIntersect-1, Rays/sec 37869K, Prf Idx 1.00, Wrkld 50.1%, Mem 1022.2M/16368M
After 500 seconds:
EngineIntersect-0, Rays/sec 36188K, Prf Idx 1.00, Wrkld 49.2%, Mem 1022.2M/16368M
EngineIntersect-1, Rays/sec 37412K, Prf Idx 1.03, Wrkld 50.8%, Mem 1022.2M/16368M

As you can see, more time passes more the difference between percentages of two cards increases.
So, Dade, are you suggesting that the increasing difference between cards percentage (49.9/50.1 to 49.2 + 50.8) is due to a damage in one card?

adrjork wrote: ↑Sun Oct 27, 2019 8:44 pm As you can see, more time passes more the difference between percentages of two cards increases.
So, Dade, are you suggesting that the increasing difference between cards percentage (49.9/50.1 to 49.2 + 50.8) is due to a damage in one card?

I was suggesting to check the GPUs clock rate and temperatures.

Anyway, it is a 2-3% difference, it is nothing, something in the 20-30% would be significative.