Nvidia Pascal and AMD Polaris, starting with GTX 1080/1070, and the AMD 480

Gamboleer
Gamboleer
Joined: 5 Dec 10
Posts: 173
Credit: 168,389,195
RAC: 0

RE: Got my brand new 1080

Quote:
Got my brand new 1080 already installed and crunching happily.
Not sure about the settings though. I have four Einstein WUs running at the moment that yield around 90% core usage.

If you're at 90%, you can probably get one more going. 90% is a decent test, though, and won't be much faster than 5 at 95% or so. Looking forward to seeing some completion times!

Gamboleer
Gamboleer
Joined: 5 Dec 10
Posts: 173
Credit: 168,389,195
RAC: 0

RE: Got my brand new 1080

Quote:
Got my brand new 1080 already installed and crunching happily.
Not sure about the settings though. I have four Einstein WUs running at the moment that yield around 90% core usage.

I'm hoping those four completed @ 8000 seconds weren't with the 1080; that's working out to sub-200k a day. Since you know your GPU usage, I assume you have software installed that shows your core and memory clocks. What are they running at? I'm curious to see if the 10 series is affected by the same clock state issues as Maxwell.

archae86
archae86
Joined: 6 Dec 05
Posts: 3,021
Credit: 5,013,383,606
RAC: 3,028,110

RE: Got my brand new 1080

Quote:
Got my brand new 1080 already installed and crunching happily.
Not sure about the settings though. I have four Einstein WUs running at the moment that yield around 90% core usage.


Thanks for the early report. We were all hoping for higher throughput, but it is very good to have some actual data.

During the time the returned Einstein BRP6 WUs were running on your machine, was there any other work running on the GPU? Was there any BOINC CPU work running (presumably from another project, as your task list does not list any that you have downloaded from Einstein). Any other significant CPU usage (games, antivirus scans, ...)?

The one hope I had that a 1080 might be worth having over a 1070 was that little "X" in the memory type employed. As on earlier nvidia cards the current Einstein BRP6 code has cared quite a bit more than games do about memory throughput, I thought there was a chance that might give the 80 a biggish bump over the 70. That now seems unlikely.

As your early reports seem likely to fulfill any early information value my Founders 1080 purchase might have brought here, I think I may cancel my B&H order and wait for the 1070s to arrive

Dave, if you are running GPU-Z or any one of the usual monitoring/overclocking utilities, could you take a look and let us know:
1. Power consumption reported--as percent of TDP
2. Average GPU core clock rate
3. Average GPU memory clock rate
4. Average memory controller load
5. Fan speed %
in addition to the GPU load % you have already reported?

Also what is your subjective impression of the GPU fan noise? Is it loud, and is the character of it obnoxious (tonal, frequent changes in pitch...buzzy) or more like white noise?

Thanks for any and all information.

juan BFP
juan BFP
Joined: 18 Nov 11
Posts: 839
Credit: 421,443,712
RAC: 0

Actual Temp will be good to

Actual Temp will be good to know too.

lHj2ixL.jpg

 

Dave
Dave
Joined: 7 Dec 12
Posts: 14
Credit: 7,276,378
RAC: 0

RE: RE: During the time

Quote:
Quote:

During the time the returned Einstein BRP6 WUs were running on your machine, was there any other work running on the GPU? Was there any BOINC CPU work running (presumably from another project, as your task list does not list any that you have downloaded from Einstein). Any other significant CPU usage (games, antivirus scans, ...)?

No, nothing at all besides some regular surfing the web. No other BOINC tasks, no higher CPU load.

Haven't had much time to run plenty of WUs so far (gaming is more important right now ;) ), but I will have the PC run through the night so tomorrow we'll see what the GPU is capable of.

Right now with four WUs GPU-Z shows:

1. Power 65-75% TDP
2. Core 1835 MHz
3. Mem 1251 Mhz (shouldn't that be somewhat higher? Even x4 it's not near the 10GHz mark).
4. Mem controller load 78%
5. Fan 53% or 2100 rpm, 79°C.

It's not a very noisy card; more a deep tone woosh, no high pitched sound. It is noticeable from 1m away, but in my opinion not very obtrusiv.
I would say that anything around 2100 rpm is ok, anything below 1800 rpm rather silent and below 1500 rpm very quiet.

archae86
archae86
Joined: 6 Dec 05
Posts: 3,021
Credit: 5,013,383,606
RAC: 3,028,110

RE: 1. Power 65-75% TDP 5.

Quote:
1. Power 65-75% TDP
5. Fan 53% or 2100 rpm, 79°C.


It appears that the default fan control is not aggressive at trying to keep the temperature down, with only 53% fan speed commanded at 79C.

Some plausible reports assert that the Founder Edition as distributed does some form of throttling at 83C at default settings. While your card would appear not to be throttling, it seems rather close. On the other hand, if the fan curve were more aggressive, in your conditions it would get rather noisy, it seems. This might be an argument in favor of Einstein users considering a non Founders Edition card reputed to have more favorable fan noise vs. cooling capability, but as you seem right about at the threshold of this mattering, perhaps not.

Could you comment on your room ambient? Specific temperature, or subjective hot, cold, ordinary?

Could you comment on your case as regards cooling capacity? Plenty of us are very far from having game addict class cases, with a single 120mm entry and another 120mm exit fan perhaps being par for the course, though some of us run with the side panel off.

I, personally, just bought my first many-fan case in anticipation of installing 10x0 card(s) in it, with a total of eight fans, but most running very slowly. This cost less than it might seem, as four of the eight fans are very quiet (e.g. Gentle Typhoon), very low-speed (500 and 800 rpm) 120mm fans I had placed in storage as moving too little air to serve in my two-fan cases.

Reported memory speed by different applications and in marketing literature is a vexing subject. As the 970 was reported to go to a P2 state which reduced the memory clock compared to gaming conditions, perhaps you might glance at the same reporting application (GPU-Z?) next time you have BOINC turned off and are running some game and see whether the reported speed (by the same reporting path) is appreciably higher. My guess is that to get to marketing numbers we are supposed to multiply by 8, in which case your 1251 reported is spot on for "10 GHz".

Thanks again for your generous sharing of information.

Dave
Dave
Joined: 7 Dec 12
Posts: 14
Credit: 7,276,378
RAC: 0

1) It's pretty warm in my

1) It's pretty warm in my room at 23°C (73°F). After running Einstein for some time now, the fan has gone up to 53%, 2170rpm, running at 81°C. I would say anything above that is getting obtrusive now.

I have had mixed experience with custom coolers. The MSI GTX 980 Gaming 4G wasn't very quiet either, running GPGPU tasks. Those cards depend heavily on case airflow while the reference cards don't need much airflow at all. I'm just a big fan of those direct heat exhaust designs.
Throtteling has always been an issues with reference cards once they hit that temperature ceiling. But so far it's been running at 1835 Mhz no problemo.

2) Tried a different monitoring tool and it shows 5005 MHz memory which means that memory is running at its specified frequency of 10 Ghz. Good.

3) I used to have a reference GTX 980 (and the above mentioned MSI Gaming) and I'd say that the 1080 FE is a bit quieter, albeit not by much. It is however a lot (!) quieter than my old GTX Titan (classic). Coil whine is not an issue but I do get some lower pitched buzz when being close to the exhaust grille.

I've also noticed a problem with the fan as it keeps spinning up from time to time for no apparent reason. It doesn't show in GPU-Z either. Guess it's driver related (happens in games, too).

4) My case is a Silverstone FT03 µATX case with a 120mm bottom fan at 750 rpm, a front intake at 550 rpm and an exhaust at 350 rpm. CPU is running semi-passively-

5) I have not yet tweaked the card at all so far (fan curve, setting lower powre limits).

archae86
archae86
Joined: 6 Dec 05
Posts: 3,021
Credit: 5,013,383,606
RAC: 3,028,110

RE: I've also noticed a

Quote:
I've also noticed a problem with the fan as it keeps spinning up from time to time for no apparent reason. It doesn't show in GPU-Z either. Guess it's driver related (happens in games, too).


The Reddit Nvidia forum has a great deal of 1080 material right now. One moderately active topic is one on annoying fan speed changes. Some people posting there claim that setting a fan curve in one or another of the usual suspect monitoring/control programs fixed them up. I'm not sure which reports there to trust, but it appears that at least some of what they are discussing resembles the issue you have reported.

archae86
archae86
Joined: 6 Dec 05
Posts: 3,021
Credit: 5,013,383,606
RAC: 3,028,110

I've cancelled my

I've cancelled my (backordered) FE 1080 order. Thanks to Dave we have early data on 1080 performance here at Einstein. My current plan is to watch how the non-Founder 1080 cards roll out to help choose which 1070 card to order.

It is possible that my hope/guess that the "X" in the memory used difference between the 1080 and 1070 might give the 1080 a special advantage here on Einstein BRP6 code, in which case the 1070 may yet disappoint my adjusted expectations, but finding that out would itself have some value.

Gamboleer
Gamboleer
Joined: 5 Dec 10
Posts: 173
Credit: 168,389,195
RAC: 0

RE: I've cancelled my

Quote:

I've cancelled my (backordered) FE 1080 order. Thanks to Dave we have early data on 1080 performance here at Einstein. My current plan is to watch how the non-Founder 1080 cards roll out to help choose which 1070 card to order.

It is possible that my hope/guess that the "X" in the memory used difference between the 1080 and 1070 might give the 1080 a special advantage here on Einstein BRP6 code, in which case the 1070 may yet disappoint my adjusted expectations, but finding that out would itself have some value.

I'm leaning towards cancelling also. Assuming the FE 1080's continue to show ~190k per day performance, I can't justify $700 for a card that I really would only use for BOINC tasks. I play older games on a 1080p monitor, so I don't need the latest & greatest.

Rumors indicate the 1080ti will be substantially more powerful than the 1080, so that may be worthwhile later. And we have new AMD cards coming as well!

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.