![]() That said, even at lower resolutions we have instances where 4 GBs is not enough (Shadow of Mordor anyone?) At 4k it’s still slower overall than the GTX 980 ti so it’s not exactly like it’s going to do the job fine by itself, maybe in another generation or two a single card will be good enough but not now. – on memory capacity, you’re missing the point since I said it’s not future proof, if it does fine right now has nothing to say about it being future proof. If that’s the case, then I’m wondering how the results will actually be.Īlso, it would be very nice to see if the lack of VRAM is actually the fault of the irregular frame times. Part of me thinks a lot of this can be solved through driver updates (e.g. To me that is not good enough to convince me. But, I haven’t seen any hard evidence of that actually being the culprit (except for gaming performance). HardOCP blames it on the lack of ROP’s and lack of memory. I’ve been reading all the major reviews and reading god knows how many forum pages for answers, but thus far everybody is makes guesses. Only time will tell if they can improve it by updated drivers, or if they actually need to get new silicon to solve all the frame time problems. I think it took them this long to get the performance where it is now. CGN is a known architecure, but HBM isn’t. Personally I think the reason for the performance in the 99th percentile is coming from HBM. Also, is the lack of increase in ROP size, the culprit of it not beating the 980 Ti? I hope they can shed some light on it. Hope it will finally answer if 4GB is enough or not for 4k gaming. Hell, FP64 is massive.Įagerly awaiting the anandtech review. Then there's financial analysis which is FP64-based too. Whoa, I did a quick google search on FP64 uses and aparrently the whole Crypto scene (Encryption/Decryption) and Cryptocurrency (mining) is dependent on FP64. The same silicon sells for 5-10x more when it's FP64 unlocked (Quadro/Tesla/FirePro) because there's a very real market for it and companies buy that stuff by the truckload despite the huge costs. There's a reason Nvidia intentially cripple FP64 performance on their consumer cards, but leave it fully operational on their Quadro cards. Hundreds of people I work with daily require FP64.Īlmost all scientific compute requires FP64Īlmost all CAD/Render operation requires FP64Įven some AE plugins (Adobe) need FP64.Īny kind of coordinate work (Geo, OS, AEC) requires FP64, because it's sub-milimeter accuracy from a universal orgin hundreds of Kilometers away.Īny kind of simulation work (solar, fluid, weather) requires FP64 because the results are practically useless with FP32.Īs far as I can tell FP32 is pretty much only of use for things like media encoding and low-accuracy rendering. That’s all, folks! We’ll see you on the next episode. ![]()
0 Comments
Leave a Reply. |