Folding Forum

The most recent info I can find is a GTX 1050 Ti on Windows running PCIe 2.0: viewtopic.php?f=38&t=34923

PantherX wrote:The most recent info I can find is a GTX 1050 Ti on Windows running PCIe 2.0: viewtopic.php?f=38&t=34923

Nice!

If PCIe 2x 3.0 is enough for a 1050 Ti so 8x 1.1 should be enought too since the transfer rate are almost identical.

*Fingers crossed*

Please do update us if you decide to go down that road as it will be helpful to the wider community now and in future

ZePompom wrote:

PantherX wrote:I believe that kiore might be talking about Gen 3. Do note that OS matters too. IIRC, on Windows Gen 3 x8 is the lowest you can go without any impact but for Linux, it is Gen 3 x2. You can read more about it here:
Brief recent discussion: viewtopic.php?f=38&t=34932
Lengthy previous discussion: viewtopic.php?f=38&t=28847

Thanks.

Just try to figure out is 8x PCIe 1.1 is enough, so "4x PCIe 3.0 is enough" doesn't help me, sadly

I do plan to use Lubuntu on my folding machine. I already installed it and made some tests. For SSH, remote desktop and FAH control remote connection.

Sorry, missed the 1st generation bit, I was referring to 3rd gen. It does seem under these circumstance that Linux your best bet.

PantherX wrote:Please do update us if you decide to go down that road as it will be helpful to the wider community now and in future

No problem


The story explaining why I do that, for the ones interested :



Since few months, I do folding on my main PC GPU (MSI GTX 1660), in addition of World Community Grid running on the CPU.

During lockdown I had the idea to put to use an old PC I have , I made it around 2009, for my sister, and she gave it back to me once she replaced it.
As it is, the PPD per watt (and PPD in general) is garbage, It's a intel E5300 with 2GB DDR2 ram and a ATI HD4770.

The CPU struggle to do few thousands PPD and the GPU is too old and is not supported.

My idea was "What if I can only replace the GPU, and do GPU folding on it". Greatly improving its efficiency.

At start I looked for a GT 1030, because I didn't want this project to cost me too much, but it's really not a great GPU and it wasn't THAT cheap.
Then looked toward GTX 1650 Super, much better value for the price, should be ok on PCIe 1.1 (My 1660 would barely bottleneck with PCIe 16x 1.1 for folding, so a 1650 Super, being a little less powerfull, should dojust fine) but much more expensive.

Then I found P106-90 for 45€ with delivery fees, and ordered one.

Still waiting for it, and in the meantime I did some OS install (after looking for a good candidate) and testing to be able to manage and monitor the old PC without needing a screen.
+ doing some fixing on the old PC, installing a fan just in front of GPU slot for better cooling, etc ...


In this context, I would have a full PCIe 16x 1.1 slot for the GPU, so I was looking for a GPU who could fold "full power" without the PCIe bandwith being an issue.


With the P106-90 being such a bargain, I was considering buying a second one, but the old PC I have only have one PCIe 16x slot, and my main PC couldn't host a P106-90 in addition of its GTX1660 (2nd PCIe slot too cramped, PCI riser would be either too short or too expensive)
Luckily it appears a familly member also have an old PC from the same gen, and would be ok for a MotherBoard swap, allowing me to have a Asus P5K-E for my folfing rig, with Two PCIe 16x 1.1 slots.


The problem is, on this motherboard, if you use both slots at once, they are reduced to 8x speed.

So now I try to figure out if the P106-90 would fold "ok" only having a 8x PCIe 1.1 connection, to figure out if I could do this Motherboard swap and do a dual-GPU folding rig

I know that they sold Bitcoin mining GPUs for scraps a few months ago.
Those P106-100 GPUs had no display out, but were discarded by the Bitcoin mining community.
They're basically 1060 GPUs, with no display out, for the sole purpose of compute.
I bought one, but it wasn't compatible with my motherboard.
I think most motherboard manufacturers only updated compatibility with the mining motherboards (those with multi PCIE ports), few standard motherboards.

The best (fastest) GPU that will work in most PCIE 2.0 motherboards, is a 1060 6GB. It's a bit faster than the 3GB.

I tried to get my hands on a P106-100 on ebay but prices were running to high, so I took a P106-90 (GTX 1050Ti-like without display output, faster ram but with less capacity).

Didn't heard about compatibility issues with these. So I hope I won't have a bad surprise once I got it

But it's sure they can be handled by most if not all "normal" motherboards.


But I don't understand why you say "The best (fastest) GPU that will work in most PCIE 2.0 motherboards, is a 1060 6GB. It's a bit faster than the 3GB."

When I do the maths, even a GTX 1660 can be handled by a PCIe 16x 1.1 (in a folding context), so why a GTX 1060 6GB couldn't and would need at least PCIe 2.0?

It's really not true that mmx PCI N.0 is REQUIRED with a certain GPU. Change the word REQUIRED to RECOMMENDED. maybe.

I look at it this way. If' I put my fastest GPU is the fastest slot, that's good. The same goes for the next slot, but what to do with the slowest slot? If I decide to add another GPU and my m/b slot can only support nx speed, I may not get 100% of the expected speed out of the GPU but I'll still be getting some lower percentage that I wouldn't be getting if I didn't add the GPU in the "slow" slot..

ZePompom wrote: But I don't understand why you say "The best (fastest) GPU that will work in most PCIE 2.0 motherboards, is a 1060 6GB. It's a bit faster than the 3GB."

When I do the maths, even a GTX 1660 can be handled by a PCIe 16x 1.1 (in a folding context), so why a GTX 1060 6GB couldn't and would need at least PCIe 2.0?

What I mean to say, is that motherboard manufacturers in general don't update their bioses beyond 2 years.
So you're stuck with a board that can read GPUs that are created about 1 or 2 years after the board is.
PCIE2.0 standard is very slow, and very old.
The GTX 1060 range of GPUs are the last ones to be recognized by a bios running PCIE 2.0.
While a 1660 or RTX could run on PCIE 2.0, chances are (quite large) that your motherboard doesn't recognize the GPU.

MeeLee wrote:

ZePompom wrote: But I don't understand why you say "The best (fastest) GPU that will work in most PCIE 2.0 motherboards, is a 1060 6GB. It's a bit faster than the 3GB."

When I do the maths, even a GTX 1660 can be handled by a PCIe 16x 1.1 (in a folding context), so why a GTX 1060 6GB couldn't and would need at least PCIe 2.0?

What I mean to say, is that motherboard manufacturers in general don't update their bioses beyond 2 years.
So you're stuck with a board that can read GPUs that are created about 1 or 2 years after the board is.
PCIE2.0 standard is very slow, and very old.
The GTX 1060 range of GPUs are the last ones to be recognized by a bios running PCIE 2.0.
While a 1660 or RTX could run on PCIE 2.0, chances are (quite large) that your motherboard doesn't recognize the GPU.

Now I want to try, to plug my GTX 1660 on the PCIe 1.1 MB to see what happens

But I think I'll wait for the P106-90 instead

If the P106 doesn't work in the 2009 PC, I'll try to get a more recent one (maybe from work) or I'll see if I find a decent PCI riser (and not costing more than the graphic card itself) to put it somewhere in my main PC.

Got it, sadly, MeeLee was right, the old motherboard doesn't support it

So I ordered a PCI riser to get it a room in my main PC next to my GTX 1660, hopefully it'll work.

Considering buying this then. To uograde just enough my old pc to make it compatible with the P106-90 (would keep old pc case, psu, hard drive)

Trying to figure out if compatibility is ok.

For pcie 2.0 motherboards, you're well past 6 years ago, in the technology world.
It would be wise if someone could tell you the oldest Nvidia GPU supported by fah, not sure if that's the GT 600 series, or 500 series, or...?
Then look up the release date of such GPU, and the release date of your motherboard.
If the GPU is more than 2 years younger, just give up on GPU folding on your motherboard.

MeeLee wrote:...It would be wise if someone could tell you the oldest Nvidia GPU supported by fah, not sure if that's the GT 600 series, or 500 series, or...

GPU with OpenCL 1.2 and Double Precision Support - Nvidia GeForce GT 710 or newer (Kepler Architecture)

Of course, the low-end GPUs might struggle with some Deadlines. If in doubt, check the TechPowerUp GPU Database: https://www.techpowerup.com/gpu-specs/

The 700 series, it is then, the oldest generation Nvidia GPUs.
It doesn't have to be a 710.
It can be any of the following:

GeForce GT 705
GeForce GT 710
GeForce GT 720
GeForce GT 730
GeForce GT 740
GeForce GTX 745
Mid-range
GeForce GTX 750
GeForce GTX 750 Ti
GeForce GTX 760 192-Bit
GeForce GTX 760
GeForce GTX 760 Ti
High-end
GeForce GTX 770
GeForce GTX 780
GeForce GTX 780 Ti


The oldest GPUs of this series are:
GTX 780 Nov 2013,
GTX Titan Feb 2013.
Most of the rest is from 2014.
That's well beyond 5 years after a 2009 motherboard.
If you can still check your motherboard online, for an updated bios beyond 2013/14, maybe a few of these might work.
But chances are slim.