Eh if you get this GPU be sure to run
[email protected] or some other FP64 application, maybe
something with over 11GB VRAM use. Not much value otherwise. The misreporting of 128 ROPs is a disappointment. GTX 1080 TI has 88 ROPs and so does RTX 2080 Ti, so this is essentially surpassing GTX 1080 and RTX 2080 only. With any luck it will match GTX Titan X Pascal in pro workloads that don't use CUDA.
Also judging by the fact that only VEGA FE had pro drivers, I don't think it'll exceed VEGA64 by over 27% in pro viewports such as Solidworks or Siemens NX.
CATIA isn't overly dependent on pro drivers for AMD cards. For Nvidia cards it's a different story. Also for Blender: that's an ideal use case for the VRAM.
Going off Gamersnexus' review, +27% would make it still slower than GTX 1080 Ti in Maya although for Solidworks it'd be faster than GTX Titan XP provided it does scale +27%. +27% in Autocad would make it semi competitive , whereas Creo / NX both require pro drivers on AMD. Of course scaling isn't linear since these benchmarks are also partly CPU bound so I don't expect it to make any significant improvements over VEGA 64 and VEGA FE.
201.88 FPS Cautilus
Titan V @ 1980/1026MHz <-- R7 2700X @ 4.35GHz might be holding this back since it is a quicker GPU than GTX 1080
149.06 FPS MrGenius
Radeon RX Vega 64 @ 1690/1175MHz <-- Ivy Bridge @ 5.3GHz might be holding this back?
147.2 FPS Baddemichl
Quadro P5000 @ 1657.5/1499.8MHz <--- GTX 1080 , pro drivers
139.33 FPS rambustibi
Radeon Vega Frontier Edition @ 1377/800MHz <-- Threadripper is suboptimal CPU for this, but it was running 3.99GHz
128.68 FPS randyenergy
GeForce GTX 1080 Ti @ MHz <--- was running stripped OS on i5 haswell @ 4.9GHz
115.65 FPS CjMitsuki
GeForce GTX 1080 Ti @ 2088/1626MHz <-- GTX 1080 Ti overclocked to the wall on ambient pretty much , 4.55GHz R7 2700X
Benchmark download: ftp://ftp.spec.org/dist/gpc/opc/view...perf12_1_1.zip
(The listed cards above my $400
P2000 , which happens to use 75W)