News Nvidia RTX 5090 laptop GPU undergoes multiple benchmark tests — flagship mobile GPU yields inconsistent results

[Admin]

The first RTX 5090 laptop GPI benchmark runs have appeared on Geekbench 6, but they’re wildly inconsistent.

Nvidia RTX 5090 laptop GPU undergoes multiple benchmark tests — flagship mobile GPU yields inconsistent results : Read more

 

[evdjj3j]

Windows probably just decided it would be a good time run Software Compatibility Telemetry.

 

[JamesJones44]

If I'm not mistaken all of these scores are worse than a rtx 4090 laptop. Seems like they might be suspect

 

[atomicWAR]

I picked up a lappy running a 4090, 7945HX3D with 64GB of ram and 1440P Gsync screen on sale back in September. I was a little worried I might regret it but with every thing I am seeing with the new 50 series parts, I think my choice was a good one. Besides I needed my lappy right then and didn't really have time to wait. My biggest regret is I didn't get more VRAM... All in all this new generation RTX cards hasn't left me particularly impressed as of yet. MFG looks interesting but I worry about latency and the fact actual raster increases will be closer to 25-30% if Nvidia charts are close to correct (FC6 and Plagues Tale).

 

[JarredWaltonGPU]

If I'm not mistaken all of these scores are worse than a rtx 4090 laptop. Seems like they might be suspect

As with all Geekbench results, take any performance figures with a bag of salt. They're basically worthless. It's a synthetic benchmark that runs in a minute or so, and minor changes with drivers and such can cause huge swings in performance.

My take: Geekbench results are barely worth the ink and paper used to print them. And in the internet age where ink and paper aren’t used, that means they’re effectively worthless. It shows that the mobile 5090 exists, which we already know, and not much else.

Core specs have 82 SMs on 5090 with 24GB of GDDR7 VRAM, compared to 76 SMs on 4090 with 16GB of GDDR6 VRAM. Based on the stated "AI TOPS" (FP8 and FP4), 5090 could be about 33% faster. We'll need to see actual gaming benchmarks and real application tests to find out for certain how they compare, not Geekbench leaks.