1950x for a good (UBT) home

Having problems installing that new stick of memory? Found some great software or having issues with something? Or maybe want to chat about your PlayStation, X-Box, Nintendo, Sega, even your old Spectrum 48k....! Or maybe something you want to sell or acquire (computing related of course!). Let us know here...
Woodles
Active UBT Contributor 20+ yrs
Posts: 10834
Joined: Thu Dec 20, 2007 12:00 am
Location: Cambridgeshire

Re: 1950x for a good (UBT) home

Post by Woodles » Mon Jan 13, 2020 5:33 pm

I was mistaken :oops:

Tctl is 94 degrees
Tdie is 68 degrees

So if I'm reading the specification right, that's the limits and it's throttling.

Think I need another heatsink fan (at least)

Damien Healy
Active UBT Contributor 10+ yrs
Posts: 1120
Joined: Mon Aug 26, 2019 2:16 pm

Re: 1950x for a good (UBT) home

Post by Damien Healy » Mon Jan 13, 2020 5:56 pm

Yep, sounds very hot indeed and pushing its max temp (68 degrees according to the spec sheet). I wonder if that’s why you are generating invalid units??

Sounds like you have a Linux box but, if you do have windows on it, try grabbing the Ryzen Master utility. It will give you a much better idea of what’s going on.

Damien Healy
Active UBT Contributor 10+ yrs
Posts: 1120
Joined: Mon Aug 26, 2019 2:16 pm

Re: 1950x for a good (UBT) home

Post by Damien Healy » Mon Jan 13, 2020 6:03 pm

On the positive side, we reached #6 top producer in QuChem a couple of days back, and recently took #8 on the FB board.

Once this box is humming away, we will be in a very nice spot indeed.

Woodles
Active UBT Contributor 20+ yrs
Posts: 10834
Joined: Thu Dec 20, 2007 12:00 am
Location: Cambridgeshire

Re: 1950x for a good (UBT) home

Post by Woodles » Mon Jan 13, 2020 6:41 pm

I did wonder when the heatsink only covered about half of the heatspreader.
I don't think the invalids are related, they've all been tried by someone else and failed there as well.

Checking I get this:

Code: Select all

mark@:~$ sensors | grep "Tdie"
Tdie:         +67.9°C  (high = +70.0°C)
Tdie:         +68.1°C  (high = +70.0°C)
mark@:~$ cat /proc/cpuinfo | grep "cpu MHz"
cpu MHz		: 3187.559
cpu MHz		: 3187.813
cpu MHz		: 3187.988
cpu MHz		: 3186.614
cpu MHz		: 3188.799
cpu MHz		: 3188.590
cpu MHz		: 3188.835
cpu MHz		: 3189.073
cpu MHz		: 3184.809
cpu MHz		: 3184.932
cpu MHz		: 3184.990
cpu MHz		: 3185.087
cpu MHz		: 3185.134
cpu MHz		: 3185.270
cpu MHz		: 3185.347
cpu MHz		: 3184.911
cpu MHz		: 3189.982
cpu MHz		: 3190.178
cpu MHz		: 3190.203
cpu MHz		: 3188.323
cpu MHz		: 3190.438
cpu MHz		: 3190.226
cpu MHz		: 3190.470
cpu MHz		: 3190.741
cpu MHz		: 3186.472
cpu MHz		: 3186.635
cpu MHz		: 3186.689
cpu MHz		: 3186.782
cpu MHz		: 3186.812
cpu MHz		: 3186.963
cpu MHz		: 3187.035
cpu MHz		: 3186.597
So what's the all core turbo supposed to be? Is it actually throttling? - [edit]3.7GHz apparently so it IS throttling :(

Either way, time to wait for in progress tasks to finish, shut it down and add more fans :)
Maybe use some proper heatsink paste as well rather than what came with the HSF?

Most of the QuChem output is down to you :o

Woodles
Active UBT Contributor 20+ yrs
Posts: 10834
Joined: Thu Dec 20, 2007 12:00 am
Location: Cambridgeshire

Re: 1950x for a good (UBT) home

Post by Woodles » Mon Jan 13, 2020 8:36 pm

New heatsink paste and a second fan:

Code: Select all

mark@:~$ sensors | grep "Tdie"
Tdie:         +65.4°C  (high = +70.0°C)
Tdie:         +65.8°C  (high = +70.0°C)
mark@:~$ cat /proc/cpuinfo | grep "cpu MHz"
cpu MHz		: 3406.946
cpu MHz		: 3407.187
cpu MHz		: 3407.393
cpu MHz		: 3408.054
cpu MHz		: 3408.300
cpu MHz		: 3408.622
cpu MHz		: 3408.763
cpu MHz		: 3409.107
cpu MHz		: 3409.788
cpu MHz		: 3410.049
cpu MHz		: 3410.199
cpu MHz		: 3410.449
cpu MHz		: 3410.428
cpu MHz		: 3410.815
cpu MHz		: 3410.950
cpu MHz		: 3411.220
cpu MHz		: 3410.756
cpu MHz		: 3410.956
cpu MHz		: 3411.142
cpu MHz		: 3411.467
cpu MHz		: 3411.638
cpu MHz		: 3411.908
cpu MHz		: 3412.047
cpu MHz		: 3412.284
cpu MHz		: 3412.900
cpu MHz		: 3413.215
cpu MHz		: 3413.379
cpu MHz		: 3413.627
cpu MHz		: 3413.809
cpu MHz		: 3414.093
cpu MHz		: 3414.215
cpu MHz		: 3414.507
Just within spec but not throttling, 68 degrees and 3.4GHz :)

I think I'm going to look at AIOs or at least, a bigger HSF (although this one is recommended by AMD on their website as being able to handle threadrippers, maybe I'll try playing with fan profiles?)

Damien Healy
Active UBT Contributor 10+ yrs
Posts: 1120
Joined: Mon Aug 26, 2019 2:16 pm

Re: 1950x for a good (UBT) home

Post by Damien Healy » Mon Jan 13, 2020 9:11 pm

Yeah, that ran around 3.7GHz most of the time under good water cooling. It did fluctuate a bit here and there.

Looks like there are only a few percentage points in it.

Woodles
Active UBT Contributor 20+ yrs
Posts: 10834
Joined: Thu Dec 20, 2007 12:00 am
Location: Cambridgeshire

Re: 1950x for a good (UBT) home

Post by Woodles » Mon Jan 13, 2020 9:26 pm

I'm not sure if being right next to the GPU is affecting it either.

Some options to try anyway, there may be some BIOS values to change as well.

Woodles
Active UBT Contributor 20+ yrs
Posts: 10834
Joined: Thu Dec 20, 2007 12:00 am
Location: Cambridgeshire

Re: 1950x for a good (UBT) home

Post by Woodles » Tue Jan 14, 2020 8:18 am

Decided to go for one of these - https://www.coolermaster.com/catalog/co ... 0-rgb-tr4/

Due in on Friday so hopefully I'll be up to 5GHz by Monday :lol:

Damien Healy
Active UBT Contributor 10+ yrs
Posts: 1120
Joined: Mon Aug 26, 2019 2:16 pm

Re: 1950x for a good (UBT) home

Post by Damien Healy » Tue Jan 14, 2020 10:58 am

Nice one. Hope you have a big case :)

Woodles
Active UBT Contributor 20+ yrs
Posts: 10834
Joined: Thu Dec 20, 2007 12:00 am
Location: Cambridgeshire

Re: 1950x for a good (UBT) home

Post by Woodles » Sat Jan 18, 2020 3:33 pm

AOI arrived and spent a pleasant morning installing it
Image

Image
Seems to have made a bit of a difference:

Code: Select all

mark@:~$ sensors | grep "Tdie"
Tdie:         +47.0°C  (high = +70.0°C)
Tdie:         +47.0°C  (high = +70.0°C)
mark@:~$ cat /proc/cpuinfo | grep "cpu MHz"
cpu MHz		: 3527.446
cpu MHz		: 3527.435
cpu MHz		: 3527.425
cpu MHz		: 3527.403
cpu MHz		: 3527.395
cpu MHz		: 3527.380
cpu MHz		: 3527.374
cpu MHz		: 3527.359
cpu MHz		: 3527.303
cpu MHz		: 3527.288
cpu MHz		: 3527.280
cpu MHz		: 3527.264
cpu MHz		: 3527.257
cpu MHz		: 3527.241
cpu MHz		: 3527.232
cpu MHz		: 3527.219
cpu MHz		: 3527.257
cpu MHz		: 3527.241
cpu MHz		: 3527.232
cpu MHz		: 3527.216
cpu MHz		: 3527.212
cpu MHz		: 3527.197
cpu MHz		: 3527.188
cpu MHz		: 3527.173
cpu MHz		: 3527.116
cpu MHz		: 3527.102
cpu MHz		: 3527.097
cpu MHz		: 3527.082
cpu MHz		: 3527.072
cpu MHz		: 3527.059
cpu MHz		: 3527.054
cpu MHz		: 3527.039
3.5 GHz all cores at only 47 degrees, very impressed! :D

Caveat: It is WEP rather than QuChem as they've stopped producing work for the moment and the side was off the case.

Now to see if I can get it up to 3.7 GHz like Damien did.

Edit. Covers on and running for a couple of hours, temperatures have gone up to 50.1 degrees.

chriscambridge
Active UBT Contributor 1+ yr
Posts: 1975
Joined: Mon Aug 08, 2016 1:56 pm
Location: Duroliponte | Cantabrigia

Re: 1950x for a good (UBT) home

Post by chriscambridge » Sat Jan 18, 2020 3:59 pm

I would definitely be interested in what kind of temps you are hitting when running a 100% load of AVX tasks; for future reference.

Also as to what config you went with; push, out of the top (fans) - or pull, from the front (fans)?
Image

Damien Healy
Active UBT Contributor 10+ yrs
Posts: 1120
Joined: Mon Aug 26, 2019 2:16 pm

Re: 1950x for a good (UBT) home

Post by Damien Healy » Sat Jan 18, 2020 5:56 pm

Nice work - looking good. I didn’t overclock it. I think you’ll find it hitting 3.7 in many workloads at those temps? I used to run it under windows, but that really shouldn’t make a difference.

Woodles
Active UBT Contributor 20+ yrs
Posts: 10834
Joined: Thu Dec 20, 2007 12:00 am
Location: Cambridgeshire

Re: 1950x for a good (UBT) home

Post by Woodles » Sat Jan 18, 2020 6:32 pm

Hi Damien,

I was just going by the BIOS that had "Target CPU speed" set at 3.4 GHz. I think I'll have to change the clock ratios to get more than 3.5 GHz on all the cores.

At the moment, it's turning out WEP tasks almost twice as quickly as my 3GHz Xeons :o :)

Damien Healy
Active UBT Contributor 10+ yrs
Posts: 1120
Joined: Mon Aug 26, 2019 2:16 pm

Re: 1950x for a good (UBT) home

Post by Damien Healy » Sat Jan 18, 2020 6:59 pm

Wow, that’s great. If they are also 16 core, then I assume it has something to do with feature support (SSE / AVX / etc)??

Woodles
Active UBT Contributor 20+ yrs
Posts: 10834
Joined: Thu Dec 20, 2007 12:00 am
Location: Cambridgeshire

Re: 1950x for a good (UBT) home

Post by Woodles » Sun Jan 19, 2020 12:35 pm

Hi Chris,

Some experimenting.

I assumed PrimeGrid 321 LLR uses basic AVX as there were some posts about testing AVX 512.
It's not quite 100% load as I have a 1080 GPU in for display purposes and that's got a core reserved while it runs PrimeGrid PPS.

At 4.0 GHz, Tdie temperatures were 105.2 degrees.
At 3.8 GHz, Tdie temperatures were 69.9 degrees.
At 3.7 GHz, Tdie temperatures were 67.5 degrees.

I was very surprised it ran at 4 GHz but quickly turned it down with the temperatures. Might be worth repeating with less demanding (non AVX) tasks?
3.8 GHz was still a bit high.
I've left it running at 3.7 GHz and will check for any failed workunits.

Radiator fans are on full and definitely audible.

chriscambridge
Active UBT Contributor 1+ yr
Posts: 1975
Joined: Mon Aug 08, 2016 1:56 pm
Location: Duroliponte | Cantabrigia

Re: 1950x for a good (UBT) home

Post by chriscambridge » Sun Jan 19, 2020 1:19 pm

So just to confirm your previous results, non-OC running on AIO is: 3.5 GHz @ 44c ?

I don't suppose you did a test on non-OC while running AVX tasks?

Thats pretty crazy how high the temps were for the higher OC's; I would have thought the CPU would have shut the system down if it hit 105c. Perhaps only Intel CPUs do this.. (eg TJMax)

--

Checking my 1950x (non-OC) on the Arctic TR33 (double fans)

>lpcpu | Grep MHz

shows 3.5 MHz

Psensor

shows Tdie 55c

while running AVX/SSE2/FMA tasks

--

Sorry I don't think I explained properly what I was asking about the AIO fans.

Obviously you can setup the radiator fans to either push or pull. Also you can either pull cold air in, or push hot air out, (through the radiator).

I was just wondering which setup you went with?
Image

Woodles
Active UBT Contributor 20+ yrs
Posts: 10834
Joined: Thu Dec 20, 2007 12:00 am
Location: Cambridgeshire

Re: 1950x for a good (UBT) home

Post by Woodles » Sun Jan 19, 2020 1:32 pm

Hi Chris,

Non overclock (3.5 GHz) on WEP was 47 degrees.
Sorry, I didn't. I'll do some later when these tasks finish.
It was only at 105 degrees for a minute or two before I shut it down, maybe there's some sort of averaging or delay? I'm guessing the CPUs are binned and ran at the sweet spot for that rating. I'll try and do a wider range of non OC and OC on non AVX and AVX tasks.
I only managed 3.2 GHz on my TR33 even with two fans and running non AVX tasks. Looks like you have a good chip there.

I'm sure I responded earlier about the fan configuration but it seems to have disappeared?

Radiator was mounted to the top of the case, fans were mounted below the radiator pushing air through it and out of the top.
There's also a 200mm intake fan at the front of the case and a 120mm fan at the back exhausting.

chriscambridge
Active UBT Contributor 1+ yr
Posts: 1975
Joined: Mon Aug 08, 2016 1:56 pm
Location: Duroliponte | Cantabrigia

Re: 1950x for a good (UBT) home

Post by chriscambridge » Sun Jan 19, 2020 1:43 pm

Okay thats useful to know; don't worry too much about doing anymore AVX test for myself as its just curiosity more than anything.

By the way I just found a document about BOINC Roadmap 2019, so I did a post, which you guys may be interested in:

viewtopic.php?f=22&p=127680#p127680
Image

Post Reply