Then do you want to explain why Rise of the Tomb Raider doesn't kill the Fury when using 5+ GB of RAM?
http://www.hardocp.com/article/2016/02/29/rise_tomb_raider_graphics_features_performance/13
It scaled even better than the 12 GB Titan X, even though obviously both were unplayable frame rates, the 4GB HBM wasn't holding it back even though 50% of it was in system RAM on Fury X vs dedicated on Titan X (7.8-9.4GB dedicated Titan X vs 4 dedicated + 2.5 -3.8 dynamic Fury X)
And that's part of the confusion. You see with DX11 games, memory management can be tuned via the software driver. AMD hired two engineers to tackle this with their Fiji lineup. Here's what they're able to do:
In instances where a game requires more than 4GB VRAM...
- Dynamic Memory is used (system memory) in order to store data that is not required for a given frame.
- Data that is required for a frame is stored in the HBM framebuffer. So you're always pulling data from HBM and not system memory even if you go over 4GB. That means no slow downs.
- Replicated data is removed as well because a lot of games replicate textures out of laziness.
- The PCIe bus is used to transfer data back and forth from system memory (dynamic memory) and the HBM framebuffer.
End result is that even if you spill into system memory, performance doesn't tank.
That's exactly my point, you can't simply say performance doesn't tank because you can swap assets in and out of system ram, because if that were always the case, there would never be a need for more vram and my 2GB 680's would have seen a much longer life span.
Now we have people using complete guesswork about how HBM isn't affected because it's faster. If you're having to fallback on a 40GB/s pipe with higher latency, HBM's bandwidth advantage is almost completely negated.
Then do you want to explain why Rise of the Tomb Raider doesn't kill the Fury when using 5+ GB of RAM?
http://www.hardocp.com/article/2016/02/29/rise_tomb_raider_graphics_features_performance/13
It scaled even better than the 12 GB Titan X, even though obviously both were unplayable frame rates, the 4GB HBM wasn't holding it back even though 50% of it was in system RAM on Fury X vs dedicated on Titan X (7.8-9.4GB dedicated Titan X vs 4 dedicated + 2.5 -3.8 dynamic Fury X)
Please stop stating this obvious fact. People are incapable of understanding that vram stored isnt vram used.Very easily, this has been mentioned many times, not only in this thread but countless others. Ram used doesn't equal ram needed. Black Ops 3 uses nearly all of my 6GB on my 980Ti but still plays fine on my 7970 which only has 3GB. Both are GDDR5. If VRAM usage worked the way you think it worked, my 7970 would have to have HBM to achieve such a feet since you're of the opinion HBM is immune to performance drop when vram is used up.
- Data that is required for a frame is stored in the HBM framebuffer. So you're always pulling data from HBM and not system memory even if you go over 4GB. That means no slow downs.
Every frame requires certain pieces of data (textures etc) but not the entirety of the data stored in the framebuffer or Dynamic Memory/System memory.I'd like to hear a further explanation on this point.
Every frame requires certain pieces of data (textures etc) but not the entirety of the data stored in the framebuffer or Dynamic Memory/System memory.
So long as that data is kept in the framebuffer, then the game will never pull the data from system memory.
If a frame, coming up, requires data stored in the system memory then the driver prefetches the data from system memory and places it into the framebuffer.
So no matter what you're rendering, the required data always resides in the framebuffer and not in the system memory.
What happens when you have 8gb of system memory and the game/operating system uses 7.2gb's and your running @ 4k with 4gb of hbm? Where does it get the memory to stream then or does it just tank?
There is a few game that now recommend 16gb of system memory.
http://wolfsgamingblog.com/2015/10/...nt-system-requirements-recommend-16gb-of-ram/
What happens when you have 8gb of system memory and the game/operating system uses 7.2gb's and your running @ 4k with 4gb of hbm? Where does it get the memory to stream then or does it just tank?
There is a few game that now recommend 16gb of system memory.
http://wolfsgamingblog.com/2015/10/...nt-system-requirements-recommend-16gb-of-ram/
If someone can afford a 4GB HBM card + a 4K monitor, while not taking advantage of the cheap Ram price to buy at least another 8GB stick, that's really user stupidity.
The 980 TI struggled more than Fury X @ 4k SSAA since it capped out memory @ 4k without AA.
If a core i3 user who brags about getting a cheap GTX 960 can afford 16GB of system memory, I'm sure GTX 980Ti and Fury X users can.What happens when you have 8gb of system memory and the game/operating system uses 7.2gb's and your running @ 4k with 4gb of hbm? Where does it get the memory to stream then or does it just tank?
There is a few game that now recommend 16gb of system memory.
http://wolfsgamingblog.com/2015/10/...nt-system-requirements-recommend-16gb-of-ram/
If a core i3 user who brags about getting a cheap GTX 960 can afford 16GB of system memory, I'm sure GTX 980Ti and Fury X users can.
Did you seriously just ask the question "What if I need more system memory?"(In a nutshell this is what you asked).
Buy more!
It's cheap.
And that's part of the confusion. You see with DX11 games, memory management can be tuned via the software driver. AMD hired two engineers to tackle this with their Fiji lineup. Here's what they're able to do:
In instances where a game requires more than 4GB VRAM...
- Dynamic Memory is used (system memory) in order to store data that is not required for a given frame.
- Data that is required for a frame is stored in the HBM framebuffer. So you're always pulling data from HBM and not system memory even if you go over 4GB. That means no slow downs.
- Replicated data is removed as well because a lot of games replicate textures out of laziness.
- The PCIe bus is used to transfer data back and forth from system memory (dynamic memory) and the HBM framebuffer.
End result is that even if you spill into system memory, performance doesn't tank.
All cards use system memory as cache, not just AMD.
Did you guys ever see the triple 4K showdown thing over at Tweaktown (I know)
http://www.tweaktown.com/tweakipedi...re-triple-4k-eyefinity-11-520x2160/index.html
Obviously settings need to be reduced for playability to remain anywhere decent, but the results are still quite surprising I'd say.
No, you just imagined a hypothetical 4K gaming desktop with 8GB of RAM.I never asked "what if I need more system memory".
Every frame requires certain pieces of data (textures etc) but not the entirety of the data stored in the framebuffer or Dynamic Memory/System memory.
So long as that data is kept in the framebuffer, then the game will never pull the data from system memory.
If a frame, coming up, requires data stored in the system memory then the driver prefetches the data from system memory and places it into the framebuffer.
So no matter what you're rendering, the required data always resides in the framebuffer and not in the system memory.
He dumbed it down so the average poster here could understand it. While you may want to split hairs about his choice of the word framebuffer, his overall explanation will suffice.
What happens when you have 8gb of system memory and the game/operating system uses 7.2gb's and your running @ 4k with 4gb of hbm? Where does it get the memory to stream then or does it just tank?
There is a few game that now recommend 16gb of system memory.
http://wolfsgamingblog.com/2015/10/...nt-system-requirements-recommend-16gb-of-ram/