Folding Forum

I dont know what the PCIE bandwidth requirements will be for the 3080, but I really wana get one on launchday. Only issue is my PC is an X58 mobo from a decade ago(i7 960/GTX 560ti) and it's limited to PCIE 2.0 X16.
1: I'm sure this will hurt gaming performance but not sure by how much?
2: Does F@H use/need high bandwidth PCIE?

Welcome to Folding@Home!

No one is sure why, but Linux is much less PCIE speed dependent than Windows is. If this is a dedicated folder, use Linux.

And yes, there probably is a reason Nvidia went PCIE 4.0. 2.0 is going to be a limit.

A high end latest series GPU with an old mobo is likely to be a disappointment at least. If your heart set on a RTX 3xxx and willing to spend that money also spend some on a board/chip that are not so vastly different that you are guaranteed compatibility issues and bottle necks.

kiore wrote:A high end latest series GPU with an old mobo is likely to be a disappointment at least. If your heart set on a RTX 3xxx and willing to spend that money also spend some on a board/chip that are not so vastly different that you are guaranteed compatibility issues and bottle necks.

Yeah I was afraid of that. I guess I could always wait and find a 2080ti for like $250 or something. My little PC should finally break the 1mill point mark sometime tomorrow. Between the CPU and GPU running full blast 24/7 it only pulls about 35k ppd. Which is basically a waste of electricity if I can pull 3mill ppd with a 3080 lol

Because of the limitations of the motherboard & CPU in my older Windows 7 (SP1 64bit) machine, the GT 1030 I purchased is only using 4 lanes of a 16 lane slot. So, should I make the investment in a more recently built motherboard (with perhaps a Ryzen 7 or 9 series), I'd be using the full 16 lanes.

The current Win 7 machines set-up still allows me to get at least 60k a day off of that GPU though (depending on the WU it's processing). Which is a far cry from trying to get a 47.5k point WU off of my AMD powered HP laptop's integrated GPU (it's a budget system).

Paul

@psaam0001: No, you only need much pcie bandwidth with fast GPUs on Windows.

@Unibrowser: RTX 3080 still works with pcie 2.0 x16 but you may loose some PPD like 20% (?) on Windows. On Linux it gets full speed.

With Windows 10 I was seeing 2.8-3.5m PPD on a 2080Ti on a 9 year old motherboard with PCIe 2.0 x16 and a Core i5 650 if that helps.

Well within expectations of what the card is capable of.

JimboPalmer wrote:...No one is sure why, but Linux is much less PCIE speed dependent than Windows is...

From what i have read, Linux has less overhead and is more optimized than Windows. The kernel too, is more optimized. Given that, the overall result is that Linux doesn't bottleneck at lower PCIe speeds while Windows does.

Please forgive me for not fully understanding PCIe, and its varieties.... There was once a time that you bought a motherboard, a CPU, a video card, memory, and storage devices w/controllers. And then w/a little tweaking to get the settings right--they worked w/o worrying about slot generation or # of lanes used by what device.

Perhaps the rule should be: Keep Calm, and Continue Folding.

Paul

Everything keeps getting BIGGER and ?better?. I remember when I could update my pc with a pair of 2K ram chips and have a huge pc. Better resolution from gpu's requires more lanes to transmit the data from the gpu to the monitor. With resolutions exceeding 4K becoming common and 8K TVs now coming on line it won't be long before a gpu will need 32 lanes to pass the data along. We are talking about 40 frames (or more) per second having to be displayed which just means more lanes to do it with.

psaam0001 wrote:the GT 1030 I purchased is only using 4 lanes of a 16 lane slot.

Yes the GT 1030 is a x4 card, in any system.

https://www.techpowerup.com/gpu-specs/g ... 1030.c2954

The 2080Ti runs fine on a PCIE 3.0 x8 slot, even an x4 slot (with 10% lower PPD), so in theory a 3070 should work ok on a PCIE 2.0 x8 slot, and the 3080 should be ok running on a PCIE 2.0 x16 slot (both in Linux).

In practice, I would question if your Bios/motherboard is able to recognize any RTX GPU at all?
The GTX series seem more compatible with older hardware than the RTX series.

If a 2060 doesn't work, forget about the 3000 series; and stick to a 1650 Super (which I heard works on most older boards); or perhaps push your luck with a 1660, 1660 Super, or 1660 Ti?
The good thing about those GPUs is that they run pretty low in power. So you can easily run 2 or 3 in a case, without overheating.

As far as a 2080Ti becoming $250, probably not within this decade.
Look at the 1080 Ti,
That old tech is still sold for $750 cheapest on second hand sites.

We won't know the true answer for a few weeks. Once the main reviews go up some testers will run benchmarks in different pcie configs and report it.
I'd say pcie 2.0 x16 or pcie 3.0 x8 would be the minimum suggestion. Probably see 3.0 x16 on super high atom count projects or multiple projects at once.

Although we won't know whether it will work on any OS until someone tests a real one, it does seem unlikely that it will meet its full potential.

MeeLee wrote: As far as a 2080Ti becoming $250, probably not within this decade.
Look at the 1080 Ti,
That old tech is still sold for $750 cheapest on second hand sites.

1080ti now on ebay for around £300 - $360
2080ti for £500 - $600

prices are falling in expectation of the $499 RTX3070. If stock is as low as some predict, 2nd hand prices will rise again.