Why bother? Wait for the next gen GPUs and use 8 PCIe lanes for each. Much easier and cheaper. Heck, using TR would be cheaper. Even ignoring cost, you have to find someone wanting to use 2x GPU + wants to upgrade to the new CPU and mobo, keeping old GPUs.
But a 16x gen 3 GPU will still only get 8x gen3 bandwidth on 8x gen4 slot, even if the theoretical maximum bandwidth is the same: Unless you have a switch instead of wires the translation is missing.
Raises hand. I certainly would run a flagship graphics card with PCIe 4.0 x8. Why wouldn't you? A Radeon VII or 2080TI barely even uses PCIe 3.0 x8 bandwidth, so PCIe 4.0 x8 is overkill in the extreme already for even the highest end graphics cards.
You should be well aware that PCIe 3.0 16x card will still use 16x PCIe lanes in a Gen3 mode (3.0) wasting, for all practical purposes, the theoretical bandwidth. The Ryzen 3, as far as current Gen3 devices goes, is still the same old regarding PCIE bandwidth. We won't be seeing Gen3 devices run any faster or taking any less PCIE lanes. Practically there is no change for this hardware.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
307 Comments
View All Comments
Billy Tallis - Monday, May 27, 2019 - link
IO bandwidth doubled throughout the platform and you're saying it's becoming *more* IO starved?!Church256 - Monday, May 27, 2019 - link
I assume they mean, an x16 GPU with 3.0 lanes would use up all 16 CPU lanes and provide no benefit.Take 16 PCI-E 4.0 lanes into a switch and have 32 PCI-E 3.0 lanes come out. Dual 3.0 GPUs without any bandwidth drop.
Or what will actually happen is 2 cards get 8 lanes each limited by the 3.0 on the GPU side and they run in x8 3.0 mode.
This is only a problem until motherboard lanes match card lanes but it's still an issue.
Zizy - Monday, May 27, 2019 - link
Why bother? Wait for the next gen GPUs and use 8 PCIe lanes for each. Much easier and cheaper. Heck, using TR would be cheaper. Even ignoring cost, you have to find someone wanting to use 2x GPU + wants to upgrade to the new CPU and mobo, keeping old GPUs.schujj07 - Monday, May 27, 2019 - link
With PCIe Gen 4 they have effectively doubled the lanes from X470 since it will have double the bandwidth.sor - Monday, May 27, 2019 - link
Well, they haven’t doubled the lanes, unless flagship GPUs are going to ship with x8 PCIe 4.0. The math always comes down to the device count.levizx - Monday, May 27, 2019 - link
You do realize x16 GPUs will work on x8/x4 links as long as physical slots fit, right? Yes, I said x4 because even x4 links are usable now.abufrejoval - Monday, May 27, 2019 - link
But a 16x gen 3 GPU will still only get 8x gen3 bandwidth on 8x gen4 slot, even if the theoretical maximum bandwidth is the same: Unless you have a switch instead of wires the translation is missing.sor - Monday, May 27, 2019 - link
Sure, but who wants to run a x16 PCIe 4.0 flagship card in a x8 or x4 slot? How many people do that today with their PCIe 3.0 x16 cards?SaturnusDK - Monday, May 27, 2019 - link
Raises hand. I certainly would run a flagship graphics card with PCIe 4.0 x8. Why wouldn't you? A Radeon VII or 2080TI barely even uses PCIe 3.0 x8 bandwidth, so PCIe 4.0 x8 is overkill in the extreme already for even the highest end graphics cards.naxeem - Monday, May 27, 2019 - link
You should be well aware that PCIe 3.0 16x card will still use 16x PCIe lanes in a Gen3 mode (3.0) wasting, for all practical purposes, the theoretical bandwidth.The Ryzen 3, as far as current Gen3 devices goes, is still the same old regarding PCIE bandwidth.
We won't be seeing Gen3 devices run any faster or taking any less PCIE lanes. Practically there is no change for this hardware.