probability of GPU being phased out
Moderator: Site Moderators
Forum rules
Please read the forum rules before posting.
Please read the forum rules before posting.
probability of GPU being phased out
I want to upgrade my GPU (current one will not fold). 4800 series
When this chip craziness subsides, I plan on building a new computer.... But, at the moment I will keep runing this old one.
Sytem I7-4770 MSI gaming MB - not running 24/7 but my PPD is around 28K (this is just sad) It seems 8hrs is a rough estimate per WU
The computer is older PCI3, and I dont want to invest much to get a GPU that will fold (with todays prices) if it has a short folding life span.
1.How long would you estimate the current low end GPUs will work with folding? I would be bummed if I get a card and next month it is black listed!
2. I just looked and it seems an RTX2060 is available with PCI 3, How hindered is a PCI3 card compared to a PCI4 if I move it to a PCI4 MB later??
3. I know this can almost not be answered but by me, but it could take a month to look at all the cards available and try and figure which ones will fold and life expectancy and PPD numbers (I just looked at a RX580 thread listing 400K PPD). If you needed a PCI3 GPU what would you look at today?
When this chip craziness subsides, I plan on building a new computer.... But, at the moment I will keep runing this old one.
Sytem I7-4770 MSI gaming MB - not running 24/7 but my PPD is around 28K (this is just sad) It seems 8hrs is a rough estimate per WU
The computer is older PCI3, and I dont want to invest much to get a GPU that will fold (with todays prices) if it has a short folding life span.
1.How long would you estimate the current low end GPUs will work with folding? I would be bummed if I get a card and next month it is black listed!
2. I just looked and it seems an RTX2060 is available with PCI 3, How hindered is a PCI3 card compared to a PCI4 if I move it to a PCI4 MB later??
3. I know this can almost not be answered but by me, but it could take a month to look at all the cards available and try and figure which ones will fold and life expectancy and PPD numbers (I just looked at a RX580 thread listing 400K PPD). If you needed a PCI3 GPU what would you look at today?
Re: probability of GPU being phased out
this question assumes all GPUs are made by NVidia, not by other folks. At the present time, FAH supports AMD and is preparing to support Intel. I'm comfortable with that assumption, but your prossing is incomplete and may (or may not) be expanded at a future date. So you're asking me to predict what NV is going to do in the near future.
Here is my prediction. BTW, my crystal ball is a bit hazy.
FAH software uses mixed precision, so GPUs which do not support FP64 hardware instructions are not supported. The lowest GPUs are the GT1030 and some models of GT7x0.
*Will those GPUs discontinue support for Double Precision: No.
*Will FAH decide to use Half Precision? Probably not, since it doesn't really help scientific research although they might find a use for it. Perhaps, and if they do, it will be offered as an enhancement to scientific computing and be optional.
I do not recommend you spend good money on a weak device. Anything that is considered weak, today, will appear to be increasingly weak. It'll still work in the future, but you won't be happy wit it.
Here is my prediction. BTW, my crystal ball is a bit hazy.
FAH software uses mixed precision, so GPUs which do not support FP64 hardware instructions are not supported. The lowest GPUs are the GT1030 and some models of GT7x0.
*Will those GPUs discontinue support for Double Precision: No.
*Will FAH decide to use Half Precision? Probably not, since it doesn't really help scientific research although they might find a use for it. Perhaps, and if they do, it will be offered as an enhancement to scientific computing and be optional.
I do not recommend you spend good money on a weak device. Anything that is considered weak, today, will appear to be increasingly weak. It'll still work in the future, but you won't be happy wit it.
Posting FAH's log:
How to provide enough info to get helpful support.
How to provide enough info to get helpful support.
Re: probability of GPU being phased out
Given the multitude of questions you stated above, i would guess you are not tech savvy. If that's the case, before pulling the trigger on any purchase you should consider any other uses you will/could gave for your pc, say gaming/rendering,etc. That factor should influence your decision. An RX 580 will be enough for you to game in 1080p settings and double as a folding machine when you are away from the computer. It won't be enough if you want to edit or render videos, so you could pick an RTX 2060 for example.
Study your present and future needs, and base your decision on that.
As for the PCIe 3.0 to 4.0 migration you will never notice a difference.
Study your present and future needs, and base your decision on that.
As for the PCIe 3.0 to 4.0 migration you will never notice a difference.
GPU only
RTX 3060 12GB Gigabyte Gaming OC [currently mining]
Folding since 14/02/2021
-
- Posts: 2522
- Joined: Mon Feb 16, 2009 4:12 am
- Location: Greenwood MS USA
Re: probability of GPU being phased out
Welcome to F@H!
I find myself telling folks with 10 year old GPUs that time has passed them by.
While Kepler and Maxwell cards still work, I would not recommend cards older than Pascal (GTX 10x0)
Newer cards are fine, but blowing money on older gear may have a shorter lifespan.
Currently, the amount of RAM is not important, the PCIE bus on the card is not important. (Sometimes it is important on the motherboard)
F@H will be gain more points with more cores.
I find myself telling folks with 10 year old GPUs that time has passed them by.
While Kepler and Maxwell cards still work, I would not recommend cards older than Pascal (GTX 10x0)
Newer cards are fine, but blowing money on older gear may have a shorter lifespan.
Currently, the amount of RAM is not important, the PCIE bus on the card is not important. (Sometimes it is important on the motherboard)
F@H will be gain more points with more cores.
Tsar of all the Rushers
I tried to remain childlike, all I achieved was childish.
A friend to those who want no friends
I tried to remain childlike, all I achieved was childish.
A friend to those who want no friends
Re: probability of GPU being phased out
People have already done work to determine PPD for a range of GPUs. -> https://docs.google.com/spreadsheets/d/ ... edit#gid=0
Read it as a guideline, I've noticed it lacks the specific setup used and any restriction placed on the cards themselves.
My own GTX 1080 Ti and RTX 2070 Super score in the 2-3 million PPD range with the power limit set to 60-70%.
I wouldn't get hung up on the PCIe 4.0 either. Graphics cards and processors which support it only came to market last year and they are in shortage everywhere.
Besides the move to PCIe 4.0, just like the earlier move to PCIe 3.0, was about future proofing the interface to ensure ample bandwidth for future devices.
USB 2.0 and SATA are about the only interfaces I can think of which needed upgraded long before they actually got it. Technically SATA still doesn't, for mechanical hard drives, but the explosion of SSD performance just swept that away.
Read it as a guideline, I've noticed it lacks the specific setup used and any restriction placed on the cards themselves.
My own GTX 1080 Ti and RTX 2070 Super score in the 2-3 million PPD range with the power limit set to 60-70%.
I wouldn't get hung up on the PCIe 4.0 either. Graphics cards and processors which support it only came to market last year and they are in shortage everywhere.
Besides the move to PCIe 4.0, just like the earlier move to PCIe 3.0, was about future proofing the interface to ensure ample bandwidth for future devices.
USB 2.0 and SATA are about the only interfaces I can think of which needed upgraded long before they actually got it. Technically SATA still doesn't, for mechanical hard drives, but the explosion of SSD performance just swept that away.
Re: probability of GPU being phased out
Thanks for the reply's - aetch thanks for the spreadsheet it was helpful
A bit of useless history, I used to work at HP in the IT department, in 2000 life shifted and I now own my own business. I used to run SETi on several systems. I remember talk of using that concept for other projects but then got away from keeping up with computer tech. I dont play games (dont have the time), but have two systems running on my desk right now, never have purchased a complete system,I built all of my computers and some for family, but up till this point havent needed to upgrade much to pull email and do a web search.
This thread was a bit of a Hello, and I tend to over think things so was looking for 'yes thats what I would do or no you missed a detail' , It will drive me crazy folding at this rate waiting for prices to come down on components (if they would very much for the foreseeable future).
I ordered an RTX 2060 - Although expensive to what used to be normal, the price to get a better GPU right now makes no sense.
A bit of useless history, I used to work at HP in the IT department, in 2000 life shifted and I now own my own business. I used to run SETi on several systems. I remember talk of using that concept for other projects but then got away from keeping up with computer tech. I dont play games (dont have the time), but have two systems running on my desk right now, never have purchased a complete system,I built all of my computers and some for family, but up till this point havent needed to upgrade much to pull email and do a web search.
This thread was a bit of a Hello, and I tend to over think things so was looking for 'yes thats what I would do or no you missed a detail' , It will drive me crazy folding at this rate waiting for prices to come down on components (if they would very much for the foreseeable future).
I ordered an RTX 2060 - Although expensive to what used to be normal, the price to get a better GPU right now makes no sense.
Last edited by Erik on Tue Jun 08, 2021 5:59 pm, edited 1 time in total.
Re: probability of GPU being phased out
Well, the 2060 with do just fine! Welcome aboard mate!
GPU only
RTX 3060 12GB Gigabyte Gaming OC [currently mining]
Folding since 14/02/2021
Re: probability of GPU being phased out
Prices are kind of high. The 2060 being at the bottom of the list for me.
But it'll fold decently for the next 4 years or so!
If it wasn't for the GPU shortage of the past 3 to 4 years (mainly miners, I'd say, but now also chip shortage), I would have recommended an RTX 3000 series GPU.
Both Nvidia and AMD make capable GPUs.
Nvidia having the slight upper hand.
GTX 1660Ti, 1650 super, RTX 2060, 2070, 2080 are now considered mid-class GPUs.
Anything RTX 2080Ti, 3060 or up is high end stuff!
But it'll fold decently for the next 4 years or so!
If it wasn't for the GPU shortage of the past 3 to 4 years (mainly miners, I'd say, but now also chip shortage), I would have recommended an RTX 3000 series GPU.
Both Nvidia and AMD make capable GPUs.
Nvidia having the slight upper hand.
GTX 1660Ti, 1650 super, RTX 2060, 2070, 2080 are now considered mid-class GPUs.
Anything RTX 2080Ti, 3060 or up is high end stuff!
Re: probability of GPU being phased out
As an owner of a 3060, it is in no way "high end stuff" a 3060 ti, perhaps!MeeLee wrote:Prices are kind of high. The 2060 being at the bottom of the list for me.
But it'll fold decently for the next 4 years or so!
If it wasn't for the GPU shortage of the past 3 to 4 years (mainly miners, I'd say, but now also chip shortage), I would have recommended an RTX 3000 series GPU.
Both Nvidia and AMD make capable GPUs.
Nvidia having the slight upper hand.
GTX 1660Ti, 1650 super, RTX 2060, 2070, 2080 are now considered mid-class GPUs.
Anything RTX 2080Ti, 3060 or up is high end stuff!
GPU only
RTX 3060 12GB Gigabyte Gaming OC [currently mining]
Folding since 14/02/2021
Re: probability of GPU being phased out
... in theory, and often in practice.JimboPalmer wrote:F@H will be gain more points with more cores.
Analysis of a protein (or any other scientific calculation) for a given protein has certain percentage of the calculation that can be performed in parallel and a portion that must be performed serially.
The parallel portion can generally be speeded up by running it on more shaders. The serial segments don't change. The proportions of the two types of calculations depends on the analysis being done and on the number of atoms where the same type of calculations are performed.
Theoretical speed "GPS" depends directly on the number of shaders times the clock speed, because you can always test with code that is easily parallelized, making the GPU look good. In practical terms, you're limited by the segments of code that must be run serially.
Newer GPUs feature their AI capabilities and their real-time enhancements to shadows in games with things like tensor cores and half-precision calculations, neither of which are used by FAH. Most of FAH's heavy calculations are traditional FP32 (Single Precision floating point) plus a small fraction of Double Precision (FP64). Any modern CPU can do those calculations, but not in a highly parallel manner.
Posting FAH's log:
How to provide enough info to get helpful support.
How to provide enough info to get helpful support.
Re: probability of GPU being phased out
You are correct.iero wrote:As an owner of a 3060, it is in no way "high end stuff" a 3060 ti, perhaps!MeeLee wrote:Prices are kind of high. The 2060 being at the bottom of the list for me.
But it'll fold decently for the next 4 years or so!
If it wasn't for the GPU shortage of the past 3 to 4 years (mainly miners, I'd say, but now also chip shortage), I would have recommended an RTX 3000 series GPU.
Both Nvidia and AMD make capable GPUs.
Nvidia having the slight upper hand.
GTX 1660Ti, 1650 super, RTX 2060, 2070, 2080 are now considered mid-class GPUs.
Anything RTX 2080Ti, 3060 or up is high end stuff!
The 3060 would be classified as mid class (like the 2080).
I think the high end gear starts from a 2080 Super / 3060 Ti.
Re: probability of GPU being phased out
Within each generation (first digits) the speed of the nV GPUs is identified by the "tens" digit and modified by a suffix such as Ti.
2080 ia faster than a 2060.
2080 ia faster than a 2060.
Posting FAH's log:
How to provide enough info to get helpful support.
How to provide enough info to get helpful support.
Re: probability of GPU being phased out
Just a quick update ....
The 2060 has been in for about 10 days, working well. The CPU is an I7-4770 running 44K PD and the 2060 is doing 906K PD. I dont have the system on full time, maybe average 80-90% up time.
Does the PPD calculate actual folding time or just 24hrs?
The 2060 has been in for about 10 days, working well. The CPU is an I7-4770 running 44K PD and the 2060 is doing 906K PD. I dont have the system on full time, maybe average 80-90% up time.
Does the PPD calculate actual folding time or just 24hrs?
Re: probability of GPU being phased out
The PPD (points per day) is an estimation, nothing more, of the points you could generate if you run your system 24/7.
It's noteworthy that the estimated PPD will rise and fall depending on the work unit(s) you're working on at that moment.
You need to go to the stats pages for the points you have actually accrued
Folding at Home -> https://stats.foldingathome.org/
Extreme Overclocking -> https://folding.extremeoverclocking.com/ (not a suggestion to overclock, they just do a fabulous breakdown of the stats and are updated every 3 hours)
It's noteworthy that the estimated PPD will rise and fall depending on the work unit(s) you're working on at that moment.
You need to go to the stats pages for the points you have actually accrued
Folding at Home -> https://stats.foldingathome.org/
Extreme Overclocking -> https://folding.extremeoverclocking.com/ (not a suggestion to overclock, they just do a fabulous breakdown of the stats and are updated every 3 hours)