[8051]

AFAICT Volta videocards were all Nvidia only designs. Why didn't Nvidia allow the AIB's access to Volta GPU's? Was it because of the HBM2 memory requirements?

 

[Sir Beregond]

AFAICT Volta videocards were all Nvidia only designs. Why didn't Nvidia allow the AIB's access to Volta GPU's? Was it because of the HBM2 memory requirements?

Probably because it wasn't a full generation of professional/gaming cards like say Turing and Ampere. You basically only had the $3k Titan V and the like $10k Quadro GV100. Oh and I guess the Tesla V100 as well. I doubt there was a large enough market for any card as it's a very specific demographic buying these. There were also no GeForce cards based on Volta, so AIB's were probably better served building/selling Pascal cards (at the time), and Nvidia was equipped to handle market for these Volta cards.

That's my read on it.

 

[WannaBeOCer]

As mentioned they never made a second die for the Geforce/Quadro line since Turing was already in the pipeline. They only had a GV100 die with FP64 cores. As we've seen they make a second die for example GP102, TU102, and GA102 for Geforce and Quadro lines.

 

[UltraMega]

He's just doing this to fill up the site with random content guys. Look at the OPs discussion history.

 

[8051]

UltraMega's thread crapping is getting old as is his idiotic speculation on my motivations. If you don't have anything meaningful to contribute don't reply.

 

[UltraMega]

UltraMega's thread crapping is getting old as is his idiotic speculation on my motivations. If you don't have anything meaningful to contribute don't reply.

I don't think so. Your threads are completely random, and you don't spend a lot of time replying to most of them so it's clear you're not looking for an engaged discussion.

The most logical explanation is you are trying to create content here. I've seen multiple fake accounts that were trying to create content here get banned, so it's definitely not crazy to think that's happening here. The only real difference I can find between your patterns and theirs is your account is older.


But if that's not the case, just tell me why do you make so many random threads?

 

[8051]

Probably because it wasn't a full generation of professional/gaming cards like say Turing and Ampere. You basically only had the $3k Titan V and the like $10k Quadro GV100. Oh and I guess the Tesla V100 as well. I doubt there was a large enough market for any card as it's a very specific demographic buying these. There were also no GeForce cards based on Volta, so AIB's were probably better served building/selling Pascal cards (at the time), and Nvidia was equipped to handle market for these Volta cards.

That's my read on it.

The Titan V had better performance than the 2080Ti according to TPU's GPU database:
NVIDIA TITAN V Specs

Could the lack of RT cores be one of the reasons Nvidia abandoned the Volta arch.?

 

[WannaBeOCer]

The Titan V had better performance than the 2080Ti according to TPU's GPU database:
NVIDIA TITAN V Specs

Could the lack of RT cores be one of the reasons Nvidia abandoned the Volta arch.?

The Volta architecture wasn't abandoned. It was used for over 2 years until the GA100 was released. These dies are designed for high-performance servers and data center racks to power AI and HPC workloads.

RT cores are definitely the reason we didn't see Geforce/Quadro based GV100 and why the GA100 is practically a very difference architecture than the GA102.

 

[Sir Beregond]

The Titan V had better performance than the 2080Ti according to TPU's GPU database:
NVIDIA TITAN V Specs

Could the lack of RT cores be one of the reasons Nvidia abandoned the Volta arch.?

Sure, but as Volta never had any gaming centric GeForce branded products, I just don't see why AIB's would be involved. Thinking about the market at that time (not what it is now), there was probably not a big demand for a $3k Titan V outside certain prosumer demographics.

And WannaBe is right. GV100 was in use until GA100 replaced it 2 years later.

So yeah for GeForce, Nvidia opted to go from Pascal to Turing for RT.

 

[8051]

The Volta memory bandwidth isn't too impressive for HBM2 -- only 40GB/s greater than a 2080ti.
Is it still a worthwhile GPU for cryptocurrency mining? Or still too expensive to be feasible?