4 ms·

> The FP32 doubling, is one of the most important bits here. But fortunately for programmers, this doesn't really change how you do your code. Early benchmarks

> The FP32 doubling, is one of the most important bits here. But fortunately for programmers, this doesn't really change how you do your code. Early benchmarks are showing games under-performing quite a bit in the worst cases. The crux of the issue is that it's not /exactly/ a no-compromise doubling of FP32. Each data path per SM can either do 2xFP32 or 1xINT32/1xFP32 per clock cycle. So if your game or application has any significant INT32 operations scheduled, all of a sudden you're back to the number of FP32 cores you had last generation, though you get the benefit of parallel INT32 execution. It's a pretty cool architecture overall though.

PaulKeeble 6y ago
Its not uncommon for GPU workloads in games to max out about 20% INT32 calculations, but alas its enough to drop the FP32 performance quite a bit. I suspect Nvidia next time will probably separate out the INT32 and 2x FP32 units and gradually move towards going towards a better ratio of hardware that better suits the usual workload split.
- _kbh_ 6y ago
  Due to the lower amount of INT32 in game loads as you stated, I don't think that separating INT32 and FP32 hardware makes a lot of sense, because you can share a substantial amount of the hardware between the two overall leading to space savings.
  - peterhj 6y ago
    On the contrary, "dark silicon" instead suggests that separating fp32 and int32 (now in GA102/104, fp32 and int32/fp32) data paths at the cost of more die space usage currently makes excellent sense. (See also: tensor cores, ray tracing cores.) Jensen Huang very briefly alluded to this when during the GA102/104 announcement he mentioned the end of Dennard scaling.
    - _kbh_ 6y ago
      But the GA102/GA104 doesn’t have seperate execution units for INT and FP32 because the INT also does FP32. So I don’t see how that shows that separating FP32 and INT hardware makes sense.
kllrnohj 6y ago
> though you get the benefit of parallel INT32 execution Parallel INT32 was added with the last generation, in Turing. See page 13 of https://www.nvidia.com/content/dam/en-zz/Solutions/design-visualization/technologies/turing-architecture/NVIDIA-Turing-Architecture-Whitepaper.pdf https://www.nvidia.com/content/dam/en-zz/Solutions/design-vi... So nvidia split out the INT32 from FP32 last gen to make them independent, then re-added FP32 to the INT32 but kept it as 2 datapaths.
Guthur 6y ago
I've been thinking that's why we are seeing the true doubling in full RTX like quake and minecraft but not on more traditional rendering engines. From my understanding int is often used for lookups, and I'd presume a lot of that is some sort of environment mapping which adds some contention as int is more limited and "steals" from the doubling of FP.

PaulKeeble 6y ago

Its not uncommon for GPU workloads in games to max out about 20% INT32 calculations, but alas its enough to drop the FP32 performance quite a bit. I suspect Nvidia next time will probably separate out the INT32 and 2x FP32 units and gradually move towards going towards a better ratio of hardware that better suits the usual workload split.

_kbh_ 6y ago
Due to the lower amount of INT32 in game loads as you stated, I don't think that separating INT32 and FP32 hardware makes a lot of sense, because you can share a substantial amount of the hardware between the two overall leading to space savings.
- peterhj 6y ago
  On the contrary, "dark silicon" instead suggests that separating fp32 and int32 (now in GA102/104, fp32 and int32/fp32) data paths at the cost of more die space usage currently makes excellent sense. (See also: tensor cores, ray tracing cores.) Jensen Huang very briefly alluded to this when during the GA102/104 announcement he mentioned the end of Dennard scaling.
  - _kbh_ 6y ago
    But the GA102/GA104 doesn’t have seperate execution units for INT and FP32 because the INT also does FP32. So I don’t see how that shows that separating FP32 and INT hardware makes sense.

kllrnohj 6y ago

> though you get the benefit of parallel INT32 execution Parallel INT32 was added with the last generation, in Turing. See page 13 of https://www.nvidia.com/content/dam/en-zz/Solutions/design-visualization/technologies/turing-architecture/NVIDIA-Turing-Architecture-Whitepaper.pdf https://www.nvidia.com/content/dam/en-zz/Solutions/design-vi... So nvidia split out the INT32 from FP32 last gen to make them independent, then re-added FP32 to the INT32 but kept it as 2 datapaths.

Guthur 6y ago

I've been thinking that's why we are seeing the true doubling in full RTX like quake and minecraft but not on more traditional rendering engines. From my understanding int is often used for lookups, and I'd presume a lot of that is some sort of environment mapping which adds some contention as int is more limited and "steals" from the doubling of FP.