Inconsistent Fire Strike Results

[Username197]

I just ran Fire Strike a few times on my new laptop after updating all the Drivers and Windows but I seem to be getting very different graphics scores even when running it in the exact same conditions, just minutes apart.
Here are my results:
https://www.3dmark.com/fs/25482111
https://www.3dmark.com/fs/25482295
https://www.3dmark.com/fs/25482348
All three tests were run within a 60 minute period in the same conditions.
Also, why does it say my driver isn't approved? It's AMD Adrenalin Version 20.10.18.02
Any help?

[UL_Jarnis]

Fire Strike is not a good test for modern hardware. Ryzen is extremely sensitive to cross-CCX latencies and Fire Strike is very sensitive to this. Use Time Spy or Night Raid to test this hardware.

[Username197]

I see. However, I've seen consistent results on a friend's PC which is the exact same model. If the score differences were around 500, I wouldn't be worried but I'm seeing over 4000 points difference in the GPU test where the cross-CCX latencies shouldn't matter if I'm not wrong. I've seen a thousand point difference in Time Spy as well though. Could you please advice?

[UL_Jarnis]

The driver you mention is very old, not sure why it is not on the approved list but being over six months old, not sure if it matters at this point.

One thing that can introduce variance is BIOS versions. Another is power saving features. Third one is inadequate cooling that causes throttling.

[Username197]

The three factors you mentioned have been the same while testing. I'm just a little concerned as I just bought this laptop and I shouldn't be seeing such inconsistencies. I'll try reaching out to AMD. Thanks!

[Huperspace]

how long where the pauses between the runs?
laptops are very heat sensitive. the load up with heat over an h (full load) and need around the same time to cool down (at idle)
so your first was the best because of cold start. second is bad because heated up (GPU) already and third is the heated up CPU because of the GPU.
I only see a down path to the last test.
only way to test "real" laptops performance is 100% fan option and make a warmup phase like 30min non stop test and the normal runs then the heat is "constant" over all test and so will the points only differ in small percentage. but if you want the best results only then you have to let it cool down between runs (and I'm not taking about CPU/GPU temps reading but the whole laptop and heatpipes as they are a heat storage)

[UL_Jarnis]

This is true - starting from cold vs. starting from already warmed up state definitely gives different results on a laptop that has... average laptop cooling solution, ie. sustained performance will result in some throttling vs cold.

[Username197]

Originally posted by Huperspace:

how long where the pauses between the runs?
laptops are very heat sensitive. the load up with heat over an h (full load) and need around the same time to cool down (at idle)
so your first was the best because of cold start. second is bad because heated up (GPU) already and third is the heated up CPU because of the GPU.
I only see a down path to the last test.
only way to test "real" laptops performance is 100% fan option and make a warmup phase like 30min non stop test and the normal runs then the heat is "constant" over all test and so will the points only differ in small percentage. but if you want the best results only then you have to let it cool down between runs (and I'm not taking about CPU/GPU temps reading but the whole laptop and heatpipes as they are a heat storage)

The pauses were around 30 minutes. Would this actually be a reason to see a 4000 point difference in the results? I'm surprised. I should run it again and check it out.

[Username197]

Just ran it again, started from cold. Getting around 14000 graphics score vs the 18000 I got earlier. Dunno what the issue could be. All other conditions are the same, I'll have to try it with different drivers maybe.