Topic: Acorn M.2 FPGA based GPU Accelerator - page 75. (Read 73507 times)
Quick question: do you have to have an accelerator for each GPU, or will 1 work with all GPUs in the rig?
I will take some.
I will offer assistance to op at no charge.
If op needs a pre order run I will do it at no fees for me.
I no longer do escrow but for this project I offer free escrow other then the tx fee.
At op do we need intel CPUs ?
Are higher end CPUs better?
I will offer assistance to op at no charge.
If op needs a pre order run I will do it at no fees for me.
I no longer do escrow but for this project I offer free escrow other then the tx fee.
At op do we need intel CPUs ?
Are higher end CPUs better?
Can you clarify what you’re offering?
Higher end CPUs can be interesting, but system memory matters more if you’re trying to use the accelerator with CPU***
sonds tempting, but cannot open website, it`s simply doesnt load.
will it be available for worldwide shipping?
will it be available for worldwide shipping?
What url/ browser?
sonds tempting, but cannot open website, it`s simply doesnt load.
will it be available for worldwide shipping?
will it be available for worldwide shipping?
I am interested in trying one to start with, same boat as others w/ Onda b250s tho good to know slot 0 can be set to 4x. Also curious about CPUs, I'm just G4400s mostly.
I have a lot of good cpus. Would be nice if this tech likes them.
I am interested in trying one to start with, same boat as others w/ Onda b250s tho good to know slot 0 can be set to 4x. Also curious about CPUs, I'm just G4400s mostly.
Let me see if I’ve got this right... this device helps GPUs get a higher hashrate, but only works for GPUs in a PCIe slot and not on risers?
Not asking for performance numbers, but what algos are you targeting?
Not asking for performance numbers, but what algos are you targeting?
Risers are fine but less gain. All algorithms in general can run on it, it is as flexible as a GPU as far as programming, but of course not all are profitable and FPGAs have different strengths.
Sorry that’s vague, if I specify one algorithm as good profit but don’t post all the info suddenly people are parroting the information incorrectly and people get confused. I’m not taking anyone’s money so patience is the name of the game. Most of the algorithm details are in the first post anyway.
Detailed performance numbers by configuration, GPU, and part are coming - it’s a lot of testing and people are already confused without adding partial information to the mix.
Thank you for the response. I did the survey on your site, but some questions I didn’t know how to answer because I don’t know enough about the product yet.
I was asking about the algos because it seems a lot of FPGA talk is about what I call “off the wall” algos, and while some are profitable, the coins they support don’t interest me. I’m looking for something for the Big 3... Ethash, Equihash and CN variants.
I will take some.
I will offer assistance to op at no charge.
If op needs a pre order run I will do it at no fees for me.
I no longer do escrow but for this project I offer free escrow other then the tx fee.
At op do we need intel CPUs ?
Are higher end CPUs better?
I will offer assistance to op at no charge.
If op needs a pre order run I will do it at no fees for me.
I no longer do escrow but for this project I offer free escrow other then the tx fee.
At op do we need intel CPUs ?
Are higher end CPUs better?
Let me see if I’ve got this right... this device helps GPUs get a higher hashrate, but only works for GPUs in a PCIe slot and not on risers?
Not asking for performance numbers, but what algos are you targeting?
Not asking for performance numbers, but what algos are you targeting?
Risers are fine but less gain. All algorithms in general can run on it, it is as flexible as a GPU as far as programming, but of course not all are profitable and FPGAs have different strengths.
Sorry that’s vague, if I specify one algorithm as good profit but don’t post all the info suddenly people are parroting the information incorrectly and people get confused. I’m not taking anyone’s money so patience is the name of the game. Most of the algorithm details are in the first post anyway.
Detailed performance numbers by configuration, GPU, and part are coming - it’s a lot of testing and people are already confused without adding partial information to the mix.
Let me see if I’ve got this right... this device helps GPUs get a higher hashrate, but only works for GPUs in a PCIe slot and not on risers?
Not asking for performance numbers, but what algos are you targeting?
Not asking for performance numbers, but what algos are you targeting?
What an effort to write such a long wall of text. The problem is that it really doesn't mean much and is probably just a bunch of incoherent babble about fake hashrates. Just take a look at all the comments on this thread, all newbies who are seemingly bumping this post the ground.
You clearly didn’t read the blob of text.
How important is pcie 3.0 4x bandwidth for this FPGA? Because I use 8 gpu riserless Onda mobos; the only way it'll connect is via m2 key host boards on Pcie 1x lanes.
i have an onda B250 D8P v1, and you can set the link width for slot 0. in the bios i just set PEG 0:1:0 (which is the 1st slot near CPU) to x4 link, and gpuz confirms a x4 link to the card. its under Max Link Width which is default AUTO which leaves it at x1. can be set to x1, x2 and x4. so we can use a x4 pcie->m2 adapter in slot 0 at full efficiency.
so i will definitely take one of these for my rig.
BTW did the survey.
What an effort to write such a long wall of text. The problem is that it really doesn't mean much and is probably just a bunch of incoherent babble about fake hashrates. Just take a look at all the comments on this thread, all newbies who are seemingly bumping this post the ground.
I did the survey, let's see now.
-Will it increase hash of gpus that are on pcie x1.
-Is there any limitation to how many gpus it can handle and is there any decrease in performance if more gpus are used
-if two of these are used do we expect double the results
Also is there any MoBo that has more then one Pcie x16 that runs at full x16 speed.
-Is there any limitation to how many gpus it can handle and is there any decrease in performance if more gpus are used
-if two of these are used do we expect double the results
Also is there any MoBo that has more then one Pcie x16 that runs at full x16 speed.
It is better to use fewer GPUs, as long as your PCIe bandwidth andsupports it. You must be able to have at least 4x PCIe 2.0 lanes of bandwidth to the GPUs you are accelerating. That is 8 GPUs in 1x, which is not the ideal.
The maximum lift total per accelerator is around 30MH normally, as stated in post. There are some algorithms where there is 60. In all non-standalone cases the PCIe bandwidth becomes bottleneck before accelerator performance.
very interesting stuff , but some algo performance gains imo is must have info
I'll be interested in 4, whenever they are ready to order.
I'd be interested in a few and would be willing to join pre-order. Haven't rolled the dice in a while. This could be fun.
How important is pcie 3.0 4x bandwidth for this FPGA? Because I use 8 gpu riserless Onda mobos; the only way it'll connect is via m2 key host boards on Pcie 1x lanes.
I was looking at those mobos and wasn't sure the size of the gap between gpu's is it sufficient for cooling or do use a fan to help I run them in server cases with 3500rpm fans. But they should be fine in cooler climate with 1070ti or lower end cards. Spacing is better than the colorful or biostar riserless 8gpu boards.
How important is pcie 3.0 4x bandwidth for this FPGA? Because I use 8 gpu riserless Onda mobos; the only way it'll connect is via m2 key host boards on Pcie 1x lanes.
Unfortunately if you don’t have any place with at least 4x PCIe 2.0 it will be difficult to use in the GPU hybrid role. You would be limited to standalone algorithms.
Noted, that's unfortunate.
Jump to: