MilkyWay@H Benchmark thread Spring 2014 - Summer 2016

[TennesseeTony]

Mark, I'll run single tasks on host FX8320E for you overnight, but I am swamped with house remodeling, from sun up until bedtime while I'm on holiday, so I'll let you look through my history for the times. Those cards are MSI R9-280X's running at 1020MHz with a AMD FX8320E CPU.

 

[Assimilator1]

Ok cool, thanks Tony
I'll run single tasks on my Tahiti once the F@H race is over.

Have fun with house remodelling! Do you mean decorating btw?

 

[TennesseeTony]

I mean I tore out all the windows, doors, inside walls, electric, plumbing, flooring, etc., leaving the outside walls and the roof and the slab of concrete it sits upon. Now I'm putting it all back together again.

 

[TennesseeTony]

Okay I'm losing points like crazy running one task at a time, so here's a screen grab for you, if you didn't already collect the data. I'm was wrong, these task seem quite stable time-wise, and suitable for a new benchmark thread.

 

[Assimilator1]

Wow! You've totally gutted your(?) house! , good luck with the work then, hope it all goes well .

Thanks for the scrnshot Tony , looks like about an average of 54.8s, so are your 280Xs still clocked at 1100/1500?
If so then the new WUs take more than twice as long as the old MW v1.02 ones! Although, to be honest they needed to take longer!
It's interesting to see that the CPU time required for each WU has shot up too, I'm sure the old mod fit & MW v1.02 WUs only needed a few seconds of CPU time!
I'll use your & waffleironhead's times for the new thread.

Using waffleironhead's times for his HD 6950, I wanted to see what kind of improvement running multiple WUs at once vs sequentially achieves with v1.43,

101.33 seconds, running 1 at a time
179.49/2= 89.74 seconds, running 2 ....... 11.4%/WU faster
328.11/4= 82.0275 seconds, running 4 ... 19%/WU faster
612.44/8= 76.55 seconds, running 8 ....... 24.4%/WU faster

A significant improvement indeed!

Looking at your valid tasks for your HD 6950 I see these times which are presumable your single WU times (correct me if I'm wrong). 102.3, 101.3, 101.9, 102.2, 101.4, 101.2, 102.4, 102.3, giving an average of 101.2s.

 

[TennesseeTony]

...... Those cards are MSI R9-280X's running at 1020MHz with a AMD FX8320E CPU.

I think CPU speed may be a factor on these units, versus almost pure GPU except for a few seconds at the end of each task, like the older work units.

 

[Assimilator1]

Opps, sorry, forgot about that post! lol

Re CPU speed, hmm, will be interesting to see after we've had some more results.
I'm going to start a new thread, I'll link these 2 threads.
Btw, I don't suppose you know if thread titles can be edited? ....actually, Smoke should know

 

[TennesseeTony]

While I'm finishing my morning coffee, I'm letting host Red run single tasks. Red has a 280X at 1080MHz, and the CPU is a dual core Pentium G3220 @ 3.00GHz, driver 1.4.1848 (according to the host page at MW, but Catalyst 15.11 is what I have in my download folder), running Windows 7 64bit. According to Milkyway's host page(s), this dual core is about 33% more powerful in single threaded applications than FX8320E.

FX8320E is running Windows 10, 64bit, driver 21.19.164.257. It would appear this is a driver provided by Microsoft...as I don't see any catalyst downloads. lol I can't believe I forgot to install a proper driver and let Windows download it's own. Catalyst ver 16.11.5. Small hard drive, I guess I deleted the download.

Ok, coffee has been consumed, and now it's time for a screenshot from Red:

 

[Assimilator1]

New benchmark thread

Intestering! will add new data to new thread later , got to go out!

 

[TennesseeTony]

I don't see how a mere 5.5% boost in the GPU clock can account for 36% more performance, so the 33% faster CPU almost has to be the reason, especially considering the CPU time: Eleven seconds less CPU time, 14.5 seconds less task time......hmmmmm.

 

[Assimilator1]

Yea, looks like the CPU does have a large influence now, interesting! I will have to note that in the new bench requirements.
So yea, your faster time does seem to be largely down to the faster CPU, I take it the driver update made no real difference?
(Btw, if we could carry this on in the new thread )