Has anyone tried the 5970 on a PCI-e 1.0 x16 board? Just wondering what the performance would be like compared to a PCI-e 2.0 board given that the 2 GPUs share the 16x lane anyway. Would using 1.0 significantly reduce performance especially if you were gaming at eyefinity resolutions such as 5760x1200?
![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
5970 on PCI-e 1.0
- Thread starter Spree
- Start date
It's a few FPS difference. Probably not very noticeable at all. What motherboard/chip combo are you running that has only PCIE 1.0? I'd worry about the CPU speed bottlenecking the video card even more than the PCIE bus bottlenecking it.
Boards with PCIe 1.0 aren't terribly old. I have the board in sig - a P35 board with a quad at 3.6Ghz. Sure it's not an i7 but a core 2 quad at that speed is not slow. Matches my 5870 quite well.
PrincessFrosty
Supreme [H]ardness
- Joined
- May 6, 2009
- Messages
- 5,905
I believe it was designed to be backwards compatible, doesn't draw the PCIe2.0 spec'd power just 75W as per PCIe1.0
As for speed, it shouldn't make any difference, the amount of data sent back and forth to the card isn't going to be much more than a single GPU card.
As for speed, it shouldn't make any difference, the amount of data sent back and forth to the card isn't going to be much more than a single GPU card.
- Joined
- Apr 11, 2007
- Messages
- 33,455
I don't think you'd see much, if any at all, noticeable real world differences between running a modern card in PCIe 1.0 or 2.0. From what I've read vid card manufacturers don't utilize all the bandwidth from 1.0. Multi-GPU setups is the only place where more than 250MB/s of data rates would be needed.
I could be wrong, but that's what I've been told.
I could be wrong, but that's what I've been told.
Yeah it's just that the 5970 is crossfire coming through one slot. Also I game at 5760x1200 where possible so depending on the intensity of the game (eg Far Cry 2) there can be a noticeable difference, especially in minimum frames. Well that's what I'm thinking anyway. The hardware canucks graphs posted only show 2560x1200 resolution so I'm wondering if the difference would increase at a higher rate when you move to an eyefinity resolution.
PrincessFrosty
Supreme [H]ardness
- Joined
- May 6, 2009
- Messages
- 5,905
The CPU isn't really doing work per pixel like the graphics card is, it's mostly just syncing cards through the drivers and supplying relavant game world data to the GPU to be renderd, for example positions of objects in a scene, that doesn't scale with the number of pixels you're showing.
Remember that the cards use generic game engine data to build a scene, then it's renderd on the card and sent straight to the display port, the bulk of the data flow is between the GPU and the VRAM the only time a lot of data is really pushed through the PCI-E lanes is when the VRAM is being filled for the first time and thats during the loading of the game engine and whatever level you're playing on.
You can stream in textures to fill the VRAM during gameplay usually if the textures being used by the engine are constantly swapped but at that point you'd get stuttering or performance issues on any card because it'll likely be pulling them textures from the hdd which is slow. To avoid that most games have texture budgets which they never exceed for 1 level.
Even so with crossfire the memory is shared windows only maps 1 set of memory so I think what happening is VRAM is filled once and internally duplicated, isn't that done via the crossfire bridge? Well even if it wasn't it would probably only be an issue during loading which is a non-issue really.
Remember that the cards use generic game engine data to build a scene, then it's renderd on the card and sent straight to the display port, the bulk of the data flow is between the GPU and the VRAM the only time a lot of data is really pushed through the PCI-E lanes is when the VRAM is being filled for the first time and thats during the loading of the game engine and whatever level you're playing on.
You can stream in textures to fill the VRAM during gameplay usually if the textures being used by the engine are constantly swapped but at that point you'd get stuttering or performance issues on any card because it'll likely be pulling them textures from the hdd which is slow. To avoid that most games have texture budgets which they never exceed for 1 level.
Even so with crossfire the memory is shared windows only maps 1 set of memory so I think what happening is VRAM is filled once and internally duplicated, isn't that done via the crossfire bridge? Well even if it wasn't it would probably only be an issue during loading which is a non-issue really.