Right now - the way I have it set, it is completing
them in about 37 minutes. I 'thought' it was doing them in
35min at one point but, I'm not sure what I changed here.
Maybe nothing changed. It could be just natural variation.
This ia a page that is a couple of pages earlier than the one in my previous message. At the time of posting, the variation here is from ~2120 to ~2220 seconds, or ~35 to ~37 minutes. Earlier still, there are pages of results that are higher, just like the most recent page. Unless you're continually tweaking some setting up and down, it's probably just natural variation.
One thing I did notice is that you seem to get work in big batches with around 4 days in between. Your work cache seems to fill up to something around 4 days and then declines to a much lower value, maybe around 0.5 days. Maybe you want to have this hysteresis in the size of your work cache. That's fine if you do, but the risk is that an untimely project outage could run you out of work. If you want your cache to remain full at all times, just reverse the two values you are using. Make the 'Store at least ...' value what you want your basic cache to be (your low water mark) and the 'Store additional ...' to be the bit on top to define a high water mark. If you don't want hysteresis type variation in cache size, just make the second setting zero.
Howdy !
It bounces around between 950 - 1000mhz
Stays at about 67 Celsius.
Bill
Those are some pretty impressive numbers. I just compared your numbers against my Pitcairn. Wow! What a difference. The temp of 67 is close to my Nvidia/Amd cards during the summer months. I priced these units and they come in at a hefty $ amount. But they do perform sooooooo....
Those are some pretty impressive numbers. I just compared your numbers against my Pitcairn. Wow! What a difference. The temp of 67 is close to my Nvidia/Amd cards during the summer months. I priced these units and they come in at a hefty $ amount. But they do perform sooooooo....
Howdy Rob,
I'm beginning to wonder if I should have spent a few more bucks
and bought the FuryX instead of the Nano.
The FuryX has a self contained water cooler. Might be better for
long term crunching ....
Then again ....
The Big Guns on here like Zalster and Jeroen are all running
980Ti s and those are really fast !
Hindsight is 20 20.
Bill
-edit- My original master plan was to eventually get a 2nd Nano
and throw it in this same machine but, I might Scrap that idea )
If anyone's interested in the performance of Fury X, I have just attached to this host: https://einsteinathome.org/host/8592961/tasks.
Running 1 WU + 6 CPU tasks, so the total CPU usage is at ~90%.
Note, that the first results might not be relevant, as I'm performing several tests to improve monitoring...
If anyone's interested in the performance of Fury X, I have just attached to this host: https://einsteinathome.org/host/8592961/tasks.
Running 1 WU + CPU tasks.
Note, that the first results might not be relevant, as I'm performing several tests to improve monitoring...
Thanks Mumak !
BTW: Someone at Seti said they were having luck running 2 tasks
at a time on newer AMD cards using a new 'VulKan' Beta driver.
I guess the higher times are because of the additional 6 CPU tasks running (2 of them are vLHC under VM). On Fiji the tasks take a large amount of CPU cycles (BRP4G ~67% CPU, BRP6 ~86% CPU). Later I'll try to reduce the CPU workload to see what impact it will have.
Still running with stock settings.
GPU clock: stable @ 1050 MHz
GPU temperature: ~ 46 C
GPU Liquid temperature: ~ 47 C
GPU VRMs: 55 C (VDDC) / 48 C (VDD)
GPU usage: very fluctuating
RE: Right now - the way I
)
Maybe nothing changed. It could be just natural variation.
This ia a page that is a couple of pages earlier than the one in my previous message. At the time of posting, the variation here is from ~2120 to ~2220 seconds, or ~35 to ~37 minutes. Earlier still, there are pages of results that are higher, just like the most recent page. Unless you're continually tweaking some setting up and down, it's probably just natural variation.
One thing I did notice is that you seem to get work in big batches with around 4 days in between. Your work cache seems to fill up to something around 4 days and then declines to a much lower value, maybe around 0.5 days. Maybe you want to have this hysteresis in the size of your work cache. That's fine if you do, but the risk is that an untimely project outage could run you out of work. If you want your cache to remain full at all times, just reverse the two values you are using. Make the 'Store at least ...' value what you want your basic cache to be (your low water mark) and the 'Store additional ...' to be the bit on top to define a high water mark. If you don't want hysteresis type variation in cache size, just make the second setting zero.
Cheers,
Gary.
RE: Howdy ! It bounces
)
Those are some pretty impressive numbers. I just compared your numbers against my Pitcairn. Wow! What a difference. The temp of 67 is close to my Nvidia/Amd cards during the summer months. I priced these units and they come in at a hefty $ amount. But they do perform sooooooo....
RE: Those are some pretty
)
Howdy Rob,
I'm beginning to wonder if I should have spent a few more bucks
and bought the FuryX instead of the Nano.
The FuryX has a self contained water cooler. Might be better for
long term crunching ....
Then again ....
The Big Guns on here like Zalster and Jeroen are all running
980Ti s and those are really fast !
Hindsight is 20 20.
Bill
-edit- My original master plan was to eventually get a 2nd Nano
and throw it in this same machine but, I might Scrap that idea )
If power usage is no issue
)
If power usage is no issue for you, then yes - the Fury X would be (a bit) better.
Otherwise, that performance for 175W TDP is very good.
Also note, that water cooling is less effective at higher ambient temperatures.
-----
RE: If power usage is no
)
Aha - Thanks Mumak.
I generally cease most crunching in the Summertime.
Don't need any more heat )
Bill
The Nano finally got up to
)
The Nano finally got up to 170K rac this morning !
Took awhile ...
Heck ... a 5 year old HD-7970 can do that !
(Probably at quite a bit Higher Power consumption. )
Bill
HD 7970 and R9 280X are the
)
HD 7970 and R9 280X are the best GPUs for Einstein@Home project (and many other projects).
If anyone's interested in the
)
If anyone's interested in the performance of Fury X, I have just attached to this host: https://einsteinathome.org/host/8592961/tasks.
Running 1 WU + 6 CPU tasks, so the total CPU usage is at ~90%.
Note, that the first results might not be relevant, as I'm performing several tests to improve monitoring...
-----
RE: If anyone's interested
)
Thanks Mumak !
BTW: Someone at Seti said they were having luck running 2 tasks
at a time on newer AMD cards using a new 'VulKan' Beta driver.
http://support.amd.com/en-us/kb-articles/Pages/Radeon-Vulkan-Beta.aspx
I am reluctant to try it as I also play games on the Nano lol !
Bill
I guess the higher times are
)
I guess the higher times are because of the additional 6 CPU tasks running (2 of them are vLHC under VM). On Fiji the tasks take a large amount of CPU cycles (BRP4G ~67% CPU, BRP6 ~86% CPU). Later I'll try to reduce the CPU workload to see what impact it will have.
Still running with stock settings.
GPU clock: stable @ 1050 MHz
GPU temperature: ~ 46 C
GPU Liquid temperature: ~ 47 C
GPU VRMs: 55 C (VDDC) / 48 C (VDD)
GPU usage: very fluctuating
-----