In Vega 20's defense, it's a pretty darn small die compared to Volta, TU104, and TU102 so AMD could strap two Vega 20 GPU's onto one card, drop the clocks 20% to get the much better perf/w, and likely have a good competitor to Volta's successor.
I'm not sure either. Given the time scales, likely R&D budget and their starting point they've probably done a decent enough job of making it into a compute card.
It is still likely to end up in a nasty place when Volta's successor rolls up but a huge & growing market so pie for all
The software stuff they talked about working really well is probably more important in the medium /long term anyway.
Why this talk of gaming in a thread concerning a data center GPU?
It's just a higher clocked Vega, so gaming-wise it will perform as expected with slightly higher clocks and inclusive of all the other quirks of the architecture, perhaps with the memory bandwidth no longer being a limitation when it comes to scaling of the GPU core with frequency.the GPU itself should be capable of playing games just fine, and could be the basis for a regular Radeon
That is not true at all.
Alot of people were jumping on the 20tflop bandwagon from WCCFtech a while back.
Some people at overclock.net were claiming 44 billion transistors, 30tflop cards based on 2.8 increase in transistor density.
Heck, even AMD slides which initially stated 1.35x performance at the same power have shrunk down to 1.25x performance at the same power.
At a 1.25 increase in tflop count only, this card will be slower than a gtx 1080 ti in gaming considering the bottlenecks of the Vega architecture at higher frequencies and would thus lose to a rtx 2080 which got slammed for it's performance not that long ago.
With only this much tflops, it makes absolute zero sense for AMD to release this as a consumer gaming card which many people wanted. You would be looking at speeds at best 20% faster than vega 64 which put it slight faster than an rtx 2070 but much more difficulty being able price near that region because of the 7nm die and the 4 stacks of HBM2.
This is a new node and you think 25% better than what peoples were expecting considering Nvidia was crapped on for only a 40% performance jump with essentially the same node and was crapped on. Why do people rewrite history, so we can prop one company up and tear another down. It creates a more hostile forum environment because it turns things into videocard politics. Don't lie to yourself. Using the same power while getting 25% more performance is at the bottom tier of peoples expectations considering the move to the new node.
Just to illustrate this further, people said, the rtx 2080 was a disappointing jump with a 40% performance jump from the rtx 1080 to 2080 because the gtx 980 to 1080 was a 66% jump. They call it turding after it was launched because of this disappointing performance jump. So now 25% performance jump on a new node is better than what people were expecting?
IF links have latency of 60-70ns!
This is NOT a gaming card. Vega Frontier was actually very competitive with Volta, even though it was an older GPU compared to volta and done on 16nm compared to 12nm.Volta is 12nm, Vega 20 is freaking 7nm. Volta @7nm will stomp all over it in style, especially when you ramp it's clocks up!
Nope, it was barely competitive with GP102. GP100 outclassed it in every metric. Volta is in a different league all together.Vega Frontier was actually very competitive with Volta
Nope, your wrong because i say your wrong. Notice how you have done this and nothing else this entire thread?Nope, it was barely competitive with GP102. GP100 outclassed it in every metric. Volta is in a different league all together.
It's just a higher clocked Vega, so gaming-wise it will perform as expected with slightly higher clocks and inclusive of all the other quirks of the architecture, perhaps with the memory bandwidth no longer being a limitation when it comes to scaling of the GPU core with frequency.
Otherwise there is nothing to be excited about from the point of view of gaming with this launch.
We're looking at about 20% higher clock speeds and double the memory bandwidth for Vega 20, compared to Vega 10. It's not unreasonable to think this would mean at least 20% more performance across the board. That would put a hypothetical gaming Vega 20 about on par with the RTX 2070, albeit without the ray-tracing functionality and probably a higher TDP.