Discussion in 'Videocards - AMD Radeon' started by Stukov, Jun 29, 2016.
Just when I thought it actually had 4gb
That happens when there is no competition. Why do you care anyway you will never buy one so I don't see why you care.
Turns out OCUK were in fact out of stock at the time of ordering, although the card was listed as in stock.
Going to wait for the GTX 1060 to launch before ordering again.
Oh sorry, it didn't have two horses in the back, it had a Fiat 125. Better now? :infinity:
I can't believe that people think that thinking that this kind of thing from a company is "not a big deal". It's literally like selling you a car with false specs. Look how nice that went for VW.
But all card reviewed were the same as the ones shipped. There was no real bait and switch. VW is a bait and switch. You plug the inspection plug in the DLC3 and the car runs clean. In the case of the 970 it was the same card no matter what. The specs were misleading/misrepresented but the card from launch to now has not changed.
It wasn't only the memory. It was the amount of L2 cache, ROPs and Texture Units too. And when you get two of those for SLI, you don't expect a 3.5GB performance. Why are you defending this? Seriously. The VW cars didn't change either. The only difference between them is that VW could cheat on the emissions, while NVIDIA couldn't on the memory bandwidth test, but they cheated where they could (on the spec sheet), just like VW.
So tell me how that spec sheet affected how the card performed in the end? Do you really buy a card because of the number of ROPs?
Didn't ATI pull a VW a long time ago with altering bench performance?
Look I'm not defending Nvidia here over misrepresenting their card. All I'm saying is that there was no bait and switch. There was never a 970 with the specs originally listed. And the card does have 4GB that is fact. Just because a 6950 could not effectively use 2GB did not mean it had 1.2GB
Sigh... I hate stupid car analogies that don't even fit. It's more like one storage room that you can fill/empty quickly vs. one bit smaller room that you can fill/empty quickly but which also has attic which makes the storage space equal but it's slower to fill/empty that.
I say its more like false advertising if the card is not utilizing the hardware as advertised. If you bought an 8 GB card but only 6 could be used, you would be furious once you tried to play anything that could use that much vram, especially in 4k where it would be needed. In the same token, Fury X has less ram and performs on scale with other cards using more vram, so performance is the measure which we should give the attribute to. Nvidia misled to some degree.
So tell me where the 480 can fill all 8GB available and stay above 60FPS or even 30FPS.
It will run out of gas before it fills 3.5 Gbs. :wanker:
I will definitely be able to tell you once afterburner or another software can show it on display in game. But I can say this, at 4k on project cars with some of the settings turned down the card plays in that range easily. As far as 4GB is concerned, that is easily needed by today's top games and every mb counts.
Even if 480 cant use all 8gb vram effectively before runs out of power but still can use 5-6gb which is always better than 4gb.
This is something I cant figure out why its so hard for some people to understand.
Running out of GPU power is far better then running out of Vram.
With their belief(thought), even my Fury X was a pile of shyte because it couldnt maintain 60fps in games @ 4k before running out of Vram. Thats pretty silly IMHO
Even then, a 4GB RX 480 is still a great budget gpu.
It only has 3.5gb vram.
You probably thinking of 970.
VCE anymore useful in these? How about VSR? Thanks for any answers.
I thought the same, but that really doesn't work that way. Imagine the VRAM as the trunk of your car, the CPU RAM as your garage, and the GPU as another house.
You are not able to unload all the contents of your trunk simultaneously in the new house, but if your trunk can take everything in a single trip, you don't waste time with new trips.
It's exactly like that for GPU VRAM. The GPU doesn't have to refresh or read all of it for every frame, it just needs to keep things it uses frequently within reach. 8GB is the new standard for very good reasons, mostly that it can mirror the memory contents of console ports (which have even weaker GPUs and yet use such large memory pools).
VCE should be updated to version 4.0 instead of 3.0 from Fury/Nano but as for what that entails I do not know though the driver has registry settings for x265 / HEVC or what it's also called plus it might be a bit faster but I don't think there's any specs available.
(I don't think there's much for 3.0 either, it does x264 but it's slower than Nvidia or Intel if I understood a post by Unwinder in the Afterburner section of these forums correctly when someone asked about slow encoding speed.)
Similarly GCN 4th doesn't say much either, not sure what the changes are unless it's all hardware and it's still using the GCN 1.2 specs with the new hardware generation improvements, don't know if there's any details on that either but I haven't found anything.
I do not think Polaris is a big improvement over Fiji or Tonga. I think it having lower gmc then Fiji says a lot imo 8.5 vs 8.1. Carrizo has higher smc (I have no clue now what smc is actually). Polaris VCE is 3.4 ain't it?
I wonder how much Vega changes things tho since there should be more updates. GMC up to 9.0 and maybe other updates.