• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Best bang for the buck F@H card.

D

Desslok

Diamond Member
Looking to upgrade and wanted to get the collectives opinion on which card is the best bang for the buck now? I do not care if it is Nvidia or AMD.

Thank you.
 
Desslok said:
Looking to upgrade and wanted to get the collectives opinion on which card is the best bang for the buck now? I do not care if it is Nvidia or AMD.

Thank you.
Click to expand...
Are you taking power usage into consideration? the GTX 1060 gives quite a bit of compute power for the buck, but the higher-end cards do a lot more work per kWh. If the true cost of ownership is to be known, it has to be calculated out over the expected lifetime of the card, knowing the cost per kWh, etc.
 
The GTX1070 is definitely the $400 winner. A few months back I was able to get some used 1080's though, for an average of $430 each on ebay. EVGA brand let's you transfer the remaining warranty, so that takes the worry out of a used GPU if you're interested in taking that path.
 
Hmmmm. Good question. I last folded in December-Jan, so the current work units have likely changed since then, and the old data may no longer apply. Also, each workunit has a different 'pay-scale.' If the (then) almighty 1080 downloaded one of the 'bad' workunits, points per day suffered for that day.

So perhaps dual cards would cut the odds in half of getting 'bad' or low paying workunits....or would it double the chances of getting one? 😀

https://forums.anandtech.com/threads/11th-annual-folding-home-holiday-season-3-way-race.2489602/

Take a peek at the first two pages at least, lot's of good information that is much more accurate than my memory. 🙂 Should be a chart in there somewhere too, of different cards and different work units, made by Markfw. Short answer though is that a 1080 could go from 900k ppd to 300k ppd based on the kind of work downloaded. Dual 1060's might just be better/more consistent. Also, I am convinced the the CPU can make a huge difference too.
 
Two 1060s would be faster than one 1070, almost certainly. Just less efficient, and will need a case with good air flow.

Edit: Actually, I don't know about the 3GB versions. My 1060s are 6GB, but the differences extend beyond just the memory. The 3GB cards have fewer compute resources as well.
 
OCN's PPD datatbase has a reasonable entry for GTX 1070 (sample size 128), but unfortunately only an unreliable entry for GTX 1060 (sample size 10, which is too few due to the wide variability between F@H workunits). Furthermore, the entry is for the GTX 1060 6 GB version with 1280 shaders, not the 3 GB version with 1152 shaders.

So, from this database:
GTX 1070 - 704 k PPD
GTX 1060 6 GB - approx. 384 k ppd

If we had the theoretical case of 1 big card vs. 2 small cards, with the big card having exactly double the shader count than each of the small cards, then
  • the big card should tend to bring more than double the PPD than the two small cards because of F@H's nonlinear credits, which reward not only the amount of work done but also the speed at which that work was done,
  • OTOH the big card might miss to bring (more than) double the PPD because, with the larger card, overhead spent in the CPU and in network transfers take a bigger proportion of the entire duration from sent time to report time of a job.
 
I have that exact card. It was getting about 350kppd last year.
 
So I'd say then, two 1060 3GB cards would equal a single 1070, but fall short of a single 1080, assuming they are all being fed the same tasks.

One thing you may want to consider, @Desslok , is that if you go with a more powerful single card now, you will have that 2nd slot open still, and can add a 2nd 1070/1080 later once they age and the prices fall even further.
 
My 3 1070's are currently averaging about 620k ppd. My 1080's about 690k ppd. My 1080TI's about 925k ppd. It really varies. Just get the 1070 I think....
 
I have 3 GTX 1060 6GB cards. Don't know the average PPD, but right this minute, 1 is doing 470kppd, 2 is doing 374kppd, and 3 is doing 346kppd. This averages to 397kppd, which seems about right to me from experience. My lone 1070 is doing 604kppd right now, which seems below its potential. It's definitely harder to keep the bigger card fed, for instance, if I run any other DC projects at all, F@H will take a hit on the 1070 rig, but the 1060s are much less affected. Also I was able to run a 1060 on an Ivy Bridge Celeron system with no loss of performance. To get the above 1070 number, I had to stop other DC projects, even though 2 entire Ryzen cores had been dedicated to the 1070.
 
OK, I think its easier if I just paste my current farm picture here. A 980, 4 980TI's, 3 1070;s, 2 1080;s and 4 1080TI's. The card is in the description, and the PPD is a column. As you can see, its all over the place for the same cards. 542 to 648 for 3 identical 1070's
iPVmogK.png


Edit: and BTW the Ryzens has 2 cores, and the TR's have 3 cores allotted to the vid cards. Not sure on the 2683's, but I think I said 8% of the 28 threads get reserved for video cards (92% for BOINC). Looks like also 3 threads for video cards, BOINC is using 25.
 
Last edited:
You must log in or register to reply here.

TRENDING THREADS

Back
Top