RX 480 Results

Gavin
Gavin
Joined: 21 Sep 10
Posts: 191
Credit: 35,034,710,485
RAC: 865,350

All this excessive power draw

All this excessive power draw stuff is brewing into a storm in a tea cup, in my opinion. Red team versus Green again and quite frankly I don't care! Its just another poo, poo battle between rival GPU manufacturer's, this sort of scaremongering happens every time both parties release new kit.

Although all but two of my hosts now use (old) NVidia GPU's I actually have no affiliation to AMD and at the moment it would be fair to say NVidia are the leading crunch card maker and what I would give to have numerous 980Ti's... However I absolutely refuse to pay the massive premium for their cards in pursuit of my hobby.

I'm hoping that what I have written doesn't sound like the start of some lunatic rant!

I'm sure that many reviewers out there have biased opinions on each manufacturer's offerings and will pick holes in the opponents line up for whatever reason, potential customers simply need to read between the lines.
The only way to see what works here at Einstein is for early adopters to buy and test to see what works best (hopefully) for minimal cost. People just need to bear in mind that crunching isn't gaming at the bleeding edge and power requirement is much likely to be LESS doing what we do. :-)

... I will of course come back and eat humble pie if I'm wrong...

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

RE: All this excessive

Quote:

All this excessive power draw stuff is brewing into a storm in a tea cup, in my opinion. Red team versus Green again and quite frankly I don't care! Its just another poo, poo battle between rival GPU manufacturer's, this sort of scaremongering happens every time both parties release new kit.

Although all but two of my hosts now use (old) NVidia GPU's I actually have no affiliation to AMD and at the moment it would be fair to say NVidia are the leading crunch card maker and what I would give to have numerous 980Ti's... However I absolutely refuse to pay the massive premium for their cards in pursuit of my hobby.

I'm hoping that what I have written doesn't sound like the start of some lunatic rant!

I'm sure that many reviewers out there have biased opinions on each manufacturer's offerings and will pick holes in the opponents line up for whatever reason, potential customers simply need to read between the lines.
The only way to see what works here at Einstein is for early adopters to buy and test to see what works best (hopefully) for minimal cost. People just need to bear in mind that crunching isn't gaming at the bleeding edge and power requirement is much likely to be LESS doing what we do. :-)

... I will of course come back and eat humble pie if I'm wrong...

One does need to be careful when filtering information. I saw one poor person get eviscerated on the AMD boards when he came for advice after his first PCI-E slot stopped working after an 8 hour gaming session on his new RX-480. Exactly why it happened, who can say, but the paranoia and outright hostility on the part of some users was frightening; he was accused of lying, doctoring pictures, not being a natural-born U.S. citizen, killing ambassadors in north Africa, and more.

I want both brands to succeed; monopolies stifle innovation.

cal_grufti
cal_grufti
Joined: 9 Feb 05
Posts: 12
Credit: 20,233,994
RAC: 0

I am seeing greater stability

I am seeing greater stability and more than 10% elapsed time improvement with the updated driver - no other changes - ceteris paribus in economist speak. I suspect that the particular motherboard in this PC was not happy with the power draw on the PCI-E bus. I'll post more details later.

It's crazy that even large companies feel the need to launch products when many important pieces are not yet finished.

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

Deleted

Deleted

Gavin
Gavin
Joined: 21 Sep 10
Posts: 191
Credit: 35,034,710,485
RAC: 865,350

RE: One does need to be

Quote:

One does need to be careful when filtering information. I saw one poor person get eviscerated on the AMD boards when he came for advice after his first PCI-E slot stopped working after an 8 hour gaming session on his new RX-480. Exactly why it happened, who can say, but the paranoia and outright hostility on the part of some users was frightening; he was accused of lying, doctoring pictures, not being a natural-born U.S. citizen, killing ambassadors in north Africa, and more.

I want both brands to succeed; monopolies stifle innovation.

This is exactly the point I was attempting to make when I said you need to read between the lines, there is so much misinformation and apparent hate fuelled nonsense attached to the latest and greatest graphics card releases that the best way to discover the true facts is to buy one for yourself and do your own testing. Which leads me back to the original topic of this thread.

Overnight testing with my new Wattman tweaks was again a bust, tasks still taking many hours to complete on the RX 480. I spotted the new driver release this morning and upgraded, all was running smoothly again until about 10 - 15 minutes into the run when I noticed that the tasks progress ground to a near halt and the 480's fan speed dropped. I immediately fired up GPU-Z to see what was happening...

Core clock at 1266 MHz
Memory clock at 2000 MHz
GPU load at 100%
Memory controller load at 0%
GPU temp at 51C
Fan speed at 50%

Watching the output for 5 or 10 minutes revealed that GPU load never fluctuated from 100%, clocks, temperature and fan speed also all remained the same but memory controller load would raise to 1 or 2% in time with a .004% progression of a BRP6 task. Something is causing my card problems!

I have now reduced the task concurrency from 3 to 2 and for the last hour the card has been behaving properly so I'm now hoping to get some real crunch times in the bag but if I encounter any more issues I will be RMA'ing...

Gav.

tbret
tbret
Joined: 12 Mar 05
Posts: 2,115
Credit: 4,679,229,271
RAC: 555,726

RE: I have now reduced

Quote:

I have now reduced the task concurrency from 3 to 2 and for the last hour the card has been behaving properly so I'm now hoping to get some real crunch times in the bag but if I encounter any more issues I will be RMA'ing...

I've been a spectator while this thread has been lengthening.

I know you are completely competent. I know you are very experienced.

Your results are "bothering me," though, so I thought I would ask this one question:

How sure are you that your driver installation is "completely, absolutely, clean?"

I'm not suggesting you've done it wrong. I just know how much trouble I had getting a good driver installation on a Win 10 installation of a 390X (way more than I ever had before - multiple cleanups, multiple directory deletions, multiple Registry "fixes" etc.).

The Vulkan runtimes just really messed me up and they did *not* want to go-away and leave a clean "path". I have no idea what got installed where. I just know that whatever it was took a lot of effort to make it go away.

cal_grufti
cal_grufti
Joined: 9 Feb 05
Posts: 12
Credit: 20,233,994
RAC: 0

Even with the new Radeon

Even with the new Radeon Crimson 16.7.1 driver performance degrades after a few tasks have completed. I am running GPU 0.33 / CPU 0.66.

The elapsed times are fastest after a restart of the system. Further into the run they have "decayed" by as much as 14%. The slower elapsed time runs are plagued by phases where the Memory Controller Load drops to 0 and then recovers all while the GPU continues at close to 100%.

I tested briefly with GPU 0.50 / CPU 1.00 - same issues.

Fast tasks at GPU 0.33 / CPU 0.66 have taken about 42 minutes (1/3 of 126 minutes), slow tasks have taken up to 48 minutes.

Both fast and slow tasks appear to validate just fine. I have had no invalid results with the new driver. All are either valid or waiting.

cal_grufti
cal_grufti
Joined: 9 Feb 05
Posts: 12
Credit: 20,233,994
RAC: 0

Now I have two recent

Now I have two recent inconclusive validation results. Those will probably turn into invalids once the third result comes in. Two out of about 40.

AgentB
AgentB
Joined: 17 Mar 12
Posts: 915
Credit: 513,211,304
RAC: 0

I'm not seeing any problems,

I'm not seeing any problems, but it's not clear how power or clock settings can be adjusted in Linux. "What no wattman?" i hear you say....

So i thought i'd go onto extreme x8 just to see if i could force some invalids. some results are in.

GPU temp 79C-
Wall power 235W (220V)
Tasks being validated ok.
Run times stable at ~23300s
CPU time ~570s

and sadly a black screen (but still crunching) needing a restart.

No errors or invalids as yet.

i'll leave it at x8 for a few hours, but i wil reverse back to x3 after the next few tasks

I have noticed the Xorg log fills up with many errors during running (a few Mb per hour) so i suspect there is work yet on the AMDGPU drivers.

[169699.139] (WW) AMDGPU(0): amdgpu_dri2_flip_event_handler: Pageflip completion
 event has impossible msc 11739231 < target_msc 11739232

[204988.219] (WW) AMDGPU(0): get vblank counter failed: Invalid argument
[204988.219] (WW) AMDGPU(0): first get vblank counter failed: Invalid argument

Edit: this is independent of whether crunching or not.

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

May i sugest you do a small

May i sugest you do a small table starting with 1 WU at a time...

Number of concurrent WU x Avg Number of secs to crunch a WU and if you can, the temp, GPU usage and the Total Watts used.

Just rememeber leave some CPU cores free to feed the GPU so we could eliminate the CPU efect on the table.

And BTW someone else could do the same homework on a 1060/1070/1080 GPU´s

That will helps us to follow the real production of the GPUs.

Thanks in advance

lHj2ixL.jpg

 

Comment viewing options

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