Sunday, August 16th 2020
NVIDIA GeForce RTX 3080 Surfaces on Userbenchmark, Rocks 19Gbps Memory Clock
NVIDIA's second fastest "Ampere" graphics card to launch this year, the GeForce RTX 3080, surfaced on the Userbenchmark database. Hardware Leaks (aka @_rogame) fished out several juicy details about the card that will be positioned right below the flagship RTX 3090 (RTX 2080 Ti successor) that's been in the news lately. The RTX 3080 succeeds the RTX 2080. On the Userbenchmark database, the purported RTX 3080 is shown bearing a device ID "10DE 2206." Among its readable specs leaked are a GPU frequency of up to 2.10 GHz, possibly frequency capped just like "Turing," and 10 GB of GDDR6X memory across a 320-bit wide memory interface, and a memory clock speed of 19 Gbps (GDDR6X effective), which works out to 760 GB/s of memory bandwidth.
Sources:
_rogame (Twitter), Userbenchmark database
25 Comments on NVIDIA GeForce RTX 3080 Surfaces on Userbenchmark, Rocks 19Gbps Memory Clock
It shows a 10DE 1467
That is if 3080 is around $699, then indeed it "suceeds" 2080.
And if it is not, it's not really a "successor" either, just marketing "level up" trick. Could we call it "bandwidth" instead?
The only thing I can imagine 8GB+ being useful for is if the SSD tech that consoles use being "emulated" on PC with VRAM instead of a dedicated SSD. In that case, the more VRAM the better. But then we're probably talking needing tens of gigabytes, if not hundreds. The PS5 and Xbox Series X are almost a terrabyte of blazing fast storage (825GB for the PS5, 1TB for the Series X), I don't think 16GB would suffice for that sort of task
As for capacity though, I think its not the big worry. If you look at the real differentiator for Nvidia's top end, its not so much capacity - past TWO generations are stuck at 11GB. What we are getting now, is tremendous increases in throughput/bandwidth which makes sense if you consider RT's much more dynamic resource requirement. GDDR6X is starting to catch up to HBM2.
The vast, if not overwhelming majority of games can do just fine within 10GB (and has lots of headroom, even at 4K we see games cap out (=/= actually required, but 'in use') around 9-10GB, as they did the past years. We now know that Micron can only supply 1GB chips so 16GB is probably not happening soon, but may very well happen in a refresh as GDDR6X gets improved.
So Navi is supposed to bring 16GB. It will be interesting to see how they do that, but if they do it with slower memory... you might have your capacity but not your performance. I think you're stuck waiting for a refresh with higher capacities I'm afraid, if you want to really make a dent for the future.
The more obvious route however is that Flight sims adjust the way they load and serve from VRAM. Its not particularly efficient as they do it now and you can't expect GPU to scale with that inefficiency either. Another solution is to not jump on early adopter resolutions and then stack the heaviest texture load imaginable on top, expecting all will be well :) It obviously won't and you can't expect the market to happily cater to the 1% of your niche. OTOH... at least you've got something to look forward to, I suppose :P