Question Nvidia Quadro RTX 3000 vs Quadro P4200 ?

[oliveria]

hello good people once again, I'm having plans of buying a dell mobile workstation and I'm in a dilema in choosing once that has Nvidia Quadro RTX 3000 or Quadro P4200, I see p4200 has 8gb and some good specs as opposed to rtx 3000 which has 6gb, currently I'm using a workstation that has Quadro p3000 of 6gb and would prefer a meaningful buying of another workstation with atleast 8gb vram, what's your advice about these two , or should I hold on and get a better one having like quadro rtx 4000, thanks in advance for your response

 

[Eximo]

Either would be a huge jump in CUDA cores, and the newer RTX3000 has higher memory bandwidth.

Really depends on your workload which would be better, faster GPU or more memory.

 

[oliveria]

I see the core clock speed of p4200 is 1215mhz and of rtx 3000 is 945mhz, while the memory clock speed of p4200 is 7132mhz and of rtx 3000 is 14000mhz, the memory clock speed of the rtx is double that of p4200, is this going to give me a boost if I go with the rtx, how about the core clock speed which is lower than that of p4200, how are they related or how will I realize its influence when rendering in heavy architectural softwares, also the bandwidth if rtx is 448gbs and of p4200 is 192, how will this bandwidth difference impact my usage or how does it come into play, thanks

 

[Eximo]

Base clock vs boost I think you are comparing. Boost clock of the RTX 3000 is 1380. As long as temperature and power allow, that is how fast it will go.

Overall performance is going to go to the RTX 3000. It is newer based on Turing rather than Pascal. It is also a 60-80W GPU, vs the 115W of the P4200. So it is more efficient as well. Which means it should boost longer before the laptop cooler heats up. More memory bandwidth helps, in pretty much everything. The only downside, as I have mentioned, is sticking with 6GB of memory. If your workload can exceed that, then you might want to seek out a larger VRAM amount. If not, then it will only benefit you to have 6GB of GDDR6 rather than 8GB of GDDR5.

You would see a similar improvement getting an Ampere card, RTX A3000, or an Ada Lovelace GPU, RTX 3000 Ada, which is even newer than Ampere. Of course these are more expensive.

 

[oliveria]

Great reply and thanks a lot for your explanation, then I think it would be better if I go for a workstation with quadro rtx 4000 because as part of the upgrade, I would admire this time to have 8gb vram, though another worry I'm having with the rtx 4000 is the tdp of 160w, I'm afraid this could result into overheating of the laptop, something I've always been keen about and which contributed towards me going with Quadro p3000 earlier with a tdp of 75w since apart from performance, I'm very much concerned about heating of the workstation, I see this tdp of 160w somehow going to force me to compromise and just go for rtx 3000, I see that overtime, with constant use if rtx 4000, internal components of the dell precision 7740 will start to wear out due to overheating hence reduced durability, I think the one with rtx 3000 might just be perfect for me as I'm not planning to replace laptop very often, what's your thoughts about this

 

[Eximo]

Laptop won't really be damaged by the heat. Just that the CPU/GPU will throttle sooner with a higher TDP. Also drain the battery faster if you are doing anything while mobile.

A compromise might be a laptop with the RTX A2000 8GB model. A smaller GPU, but it still has more CUDA cores than the RTX3000. It does have a lot less memory bandwidth though, they just doubled the memory chips up from the 4GB version.

RTX 2000 Ada 8GB would probably also be a good choice. Even more cores, but more importantly going from 4MB of cache to 24MB, which is a hugely improved amount and makes up for the relatively small bus. One of the reasons the 40 series GPUs (Ada) do so well compared to the 30 series cards (Ampere)

 

[oliveria]

Ok I got your point, allow me to ask maybe one more , so lets say if two workstations were to be doing renderings of two same size models of architecture drawings, for example the rtx 3000 of tdp 80w and p4200 of tdp 115 watts, does it mean the guy with rtx 3000 will enjoy longer turbo speeds endurability as opposed to p4200 since the 80 watts somehow can be accommodated with behemoth machines like the dell precision 7740, I really don't want to commit myself into buying lets say a big vram card laptop with a high tdp only to end up with the machine not rendering to its max potential due to heating of components and so on, I would rather be be fine with a lower tdp if it won't overheat and still do the job fine

 

[Eximo]

Most laptop coolers get overwhelmed pretty quickly. Every load type and laptop is going to settle at a different clock speed and power draw. That is why they give TDP ranges for the mobile chips. It is up to the integrator to decide how the chip behaves.

I would still take the newer RTX card, for various reasons. Ray tracing and tensor cores being additional silicon not related directly to CUDA performance.

 

[oliveria]

by saying the new rtx card you mean the ones built on ampere architecture like rtx A3000, or the arda types, does the plain rtx 3000 has tensor cores?, is it that the ampere versions have both cuda cores and tensor cores, or the cuda have been replaced by tensor cores?

 

[Eximo]

RTX 3000 is Turing
A3000 is Ampere
3000 Ada is Ada Lovelace

They all have CUDA, ray-tracing, and tensor cores. The amount and version goes up as the GPU gets newer. The one without tensor or ray-tracing cores is the Pascal card.

 

[oliveria]

I really appreciate your excellent explanation plus your time in giving me the insights, so I'll now go with a machine with rtx A3000 since I can afford it plus the tensor cores additions in them, are these tensor cores mainly just helpful in AI related items or they'll be meaningful too in the renderings?