Question Nvidia Quadro RTX 3000 vs A3000

[oliveria]

hello good people, I'm interested in buying either a dell precision 7750 or 7760 with either Quadro RTX 3000 or A3000, I've read extensively about these two cards , I see the A3000 is somehow rich in some features but I'm heavily worried about the slow clock speed of 600mhz, I don't want to believe that maybe having more cuda cores can compensate for the slow clock speed, in terms of speed I'm leaning towards the rtx 3000 but I don't know of this route is suicidal since I'll also be leaving the tensor cores that are available in the A3000, I would want to settle for something that I won't regret picking since I'm treating this as an investment that will help me do heavy architectural drawings plus renderings, just this clock speed of 600mhz is reducing my morale for the A3000, I'll really appreciate your suggestions, Thanks

 

[BFG-9000]

Both RTX 3000 and A3000 have a base clock of 600MHz and the boost clock differs by only 15MHz (1215MHz vs 1230MHz). The biggest difference is RTX 3000 has 2304 shaders and A3000 has 4096 more powerful ones.

RTX 3000 is Turing and has 288 tensor cores. A3000 is Ada which only has 128 tensor cores but they are up to 4x as fast as the Turing tensor cores.

 

[oliveria]

I read a comparison of these two graphics card of both 6GB on technical city they list the core clock of RTX 3000 as 945mhz and of A3000 as 600mhz, despite the A3000 having many cores and other specs higher than the rtx 3000, could this base core clock speed result in slow operations of the heavy softwares for renderings and CAD that I intend to run on it, in your opinion is this 600mhz just more than enough to run normal rendering operations, or there's something hidden that I'm not getting?

 

[BFG-9000]

I don't think the base clock makes any difference because no work is done at the base clock, unless you have installed some software specifically to lock the GPU clocks to prevent them from boosting under a workload.

I will say that in a laptop, the 70w A3000 is much more likely to boost to and maintain higher clocks than an 80w RTX 3000 purely for thermal reasons.

Apparently the RTX 3000 also comes in a power saving 60w Max-Q variant which is what has the 600MHz base clock. The 80w version is the one with the 945MHz base clock and can boost to a whopping 1380MHz. The A3000 is more than 25% faster than the Max-Q version but only about 12% faster than the 80w RTX 3000... not counting the difference in tensor performance.

Both 600MHz and 945MHz base clocks are closer to the speeds of desktop Kepler cards from more than 10 years ago, which were the very first nVidia generation to have boost clocks at all. It makes sense to use a lower idle clock in a laptop to keep the fan noise down when not doing much, but in a modern laptop when idling at the desktop the GPU is actually powered off and its duties switched to the IGP in the CPU instead. So technically 600MHz vs 945MHz base makes no difference whether under load or at idle as it is only the default clock the GPU starts up/turns on at.

This is kind of like considering two trucks where one idles at 800rpm and the other at 1145rpm, and concluding the 1145rpm one must be better for heavy workloads because that's faster. It's not the idle RPM nor the redline that matters, but the performance at the RPM you can actually use.