Hello.
I want to buy an RTX 4000 ADA graphic cards like in the link:
So I can play with CUDA programming a liitle bit…
Nevertheless, I don’t see that particular card here on the list:
Is this card not officially compatible with CUDA or there is some other reason it’s not on the list?
Thanks
Probably the list is not up-to-date or not complete.
You can be quite sure, that the GPU is fully CUDA compatible and supported (perhaps somebody from Nvidia can confirm) and that it has compute capability 8.9.
Running CUDA kernels on it, probably is its main intended use after all.
As @Curefab says, NVIDIA often lags in updating these lists of GPUs. Every GPU NVIDIA has shipped for the past 15 years is CUDA capable.
The one curious thing I noticed about the RTX 4000 Ada is that according to the specifications in the TechPowerUp database it provides less bandwidth than the Ampere-architecture predecessor model RTX A4000 (360 GB/sec vs 448 GB/sec). Even less than the Turing-architecture pre-predecessor RTX 4000 (416 GB/sec).
Even so the FP32 FLOPS roughly doubled from RTX 4000 to RTX A4000 , and roughly doubled again from RTX A4000 to RTX 4000 Ada. Therefore I would expect the low memory bandwidth of the RTX 4000 Ada to become a frequent choke point for performance.
Wikipedia (List of Nvidia graphics processing units - Wikipedia) lists it even with only 1750 MHz memory clock, which (with 160 bits) would be only 280 GB/s. The Nvidia datasheet says 360 GB/s (NVIDIA RTX 4000 Ada Generation Datasheet), so perhaps it is only the SFF version with even less bandwidth.
In many cases the memory bandwidth would be a limiting factor, but at least it has lots of (48 MB?) L2 cache as Ada generation card.
@bogdan.maksimovic If you want to have good memory bandwidth, look for the Ada 4500 or even the Ada 6000, but there will be a pricing difference.
Yes, the SFF (small form factor) version of the RTX 4000 Ada has reduced specs because it needs to make do with half the power budget of the full-sized version.
The signature feature of the {Quadro | RTX} 4000 lineage is that it represents NVIDIA’s top-performing single-slot professional GPUs. That is important in some environments and of no importance in others.
We are changing from the P2200 to the RTX 4000 Ada, we are concerned with EMI differences are there any different frequency clock oscillators on the RTX 4000 Ada if this isnt the right place for this questions please advise
It would be remarkable if two different products, released four years apart, were not different.
RTX GPU’s would seem to comply with the various standards outlined here.
If your EMI requirements are more specific, you will probably need to do your own testing.
The RTX 4000 Ada potentially uses higher clock frequencies than the P2200. But those partially can be set by the device/module manufacturer or by you. But even with the same clock frequencies, the emissions can happen from slightly different locations or with a different strength. So you have to test in any case.