Ryzen and Threadripper

Message boards : Number crunching : Ryzen and Threadripper
Message board moderation

To post messages, you must log in.

Previous · 1 . . . 20 · 21 · 22 · 23 · 24 · 25 · 26 . . . 69 · Next

AuthorMessage
Profile Bill Special Project $75 donor
Volunteer tester
Avatar

Send message
Joined: 30 Nov 05
Posts: 282
Credit: 6,916,194
RAC: 60
United States
Message 2003029 - Posted: 17 Jul 2019, 17:23:29 UTC - in response to Message 2003024.  

If you don't get the three memory training cycles and a final POST beep, the memory settings are bad and will fall back to 2133. It can also lock up the machine requiring a BIOS CMOS reset to get back to square one. Ryzen has always done best with single rank memory sticks in the 2 X 8GB configuration. It also has preferred memory made with Samsung B-dies. That memory is easiest to achieve rated XMP settings of the memory sticks.

Have you tried the Ryzen DDR4 Memory Calculator by 1usmus? https://www.overclock.net/forum/13-amd-general/1640919-new-dram-calculator-ryzena-1-5-1-overclocking-dram-am4-membench-0-7-dram-bench.html
It really does make it a lot easier to achieve decent memory timings on Ryzen. He works with AMD and is privvy to lots of technical information that only developers are given.

The current version is not setup for Zen 2 yet but if you use the Zen+ selection, it will be close. He says a new version able to handle Zen 2 is forthcoming shortly.
I usually check HWinfo before running Testmem to verify the memory clocks are correct. I have had it revert back to the slower speed on occasion.

I am using G. Skill Flare X 3200, 14 CAS. 2 x 8 GB dimms (also plugged into the correct slots) Its single rank and B-die, so it should be optimal for overclocking. Yes, I am using the new calculator and I have been following 1usmus' guide. I was able to get 3266 with little pain, but at 3333 I keep getting a few errors in testmem. To be honest, I probably have not been as meticulous as I need to be, so I feel I need to start over and keep better records of what settings I have in bios.

Also, its a 2200G, so that should be using the Zen 1 settings. I'm not aware of anything else that needs to be considered since it is an APU, although I'm curious if I should be doing something different since the memory is shared for the Vega 8 GPU.
Seti@home classic: 1,456 results, 1.613 years CPU time
ID: 2003029 · Report as offensive     Reply Quote
Cosmic_Ocean
Avatar

Send message
Joined: 23 Dec 00
Posts: 3027
Credit: 13,516,867
RAC: 13
United States
Message 2003031 - Posted: 17 Jul 2019, 17:28:44 UTC - in response to Message 2003024.  
Last modified: 17 Jul 2019, 17:29:51 UTC

If you don't get the three memory training cycles and a final POST beep, the memory settings are bad and will fall back to 2133. It can also lock up the machine requiring a BIOS CMOS reset to get back to square one. Ryzen has always done best with single rank memory sticks in the 2 X 8GB configuration. It also has preferred memory made with Samsung B-dies. That memory is easiest to achieve rated XMP settings of the memory sticks.


The first time I booted my board up, it went through those three cycles and settled on 2133.

Then I changed it to XMP#1 and it cycled three more times and accepted XMP#1

I'm using two pairs of these: G.Skill TridentZ RGB F4-3200C14D-16GTZR

They weren't on the approved list for my board (Gigabyte X470 Aorus Gaming 7), but the first pair worked and I was happy with it. And then there was a good deal for Black Friday on the RAM again, so I got a second pair. Didn't even have to change any BIOS settings or anything. Shut down, install second pair, press power button, took about 10 seconds to run its memory tests, and booted right up in XMP#1 for 3200 @ 14-14-14-34.

Researched way back when I bought all of this in the first place back in May of 2018 and all of the G.Skill CL14 memory is exclusively Samsung B-Die, which is why the price for CL14 modules were 40-50% higher than the CL16 counterparts. Worth it though, because it worked at 3200 right out of the box, unlike gen1.
Linux laptop:
record uptime: 1511d 20h 19m (ended due to the power brick giving-up)
ID: 2003031 · Report as offensive     Reply Quote
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13161
Credit: 1,160,866,277
RAC: 1,873
United States
Message 2003038 - Posted: 17 Jul 2019, 18:16:42 UTC - in response to Message 2003029.  

Also, its a 2200G, so that should be using the Zen 1 settings. I'm not aware of anything else that needs to be considered since it is an APU, although I'm curious if I should be doing something different since the memory is shared for the Vega 8 GPU.

Yes, the APU's complicate things. Have you tried any LLC on the SoC if it is available in your BIOS. Some little bit of LLC3 can help. I also have read that going up a notch in ProcODT can help with the APU's. Your FlareX 3200CL14 kit is a great kit. I would first try the 53.0 ohms from the calculator, but if you still get errors, jump to 60 ohms. Oh, BTW, don't be afraid to set Geardown Mode ON. It really does not impact the speed all the much compared to the usual Calculator recommendation to set GD Mode OFF. But it can really help stabilize the memory for no errors.

I just tried mine out in a contest machine I'm prepping. I picked it up just a while ago for a good $120 price. Simply used the Stilt memory presets for single rank 1.4V, 3466Mhz B-die to boot and then used the 3466 Fast timings from the calculator to tighten up a bit more. On a 2700X though, not an APU.
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 2003038 · Report as offensive     Reply Quote
Profile Bill Special Project $75 donor
Volunteer tester
Avatar

Send message
Joined: 30 Nov 05
Posts: 282
Credit: 6,916,194
RAC: 60
United States
Message 2003052 - Posted: 17 Jul 2019, 19:27:00 UTC - in response to Message 2003038.  

I'm not sure what LLC(3) is. I'd search for it online, but that acronym is conveniently the same as Limited Liability Company. Do you have a spot you can point me to off the top of your head?

Can I jump up to 60 ohms? The calculator lists 53, 48, and 40 for the recommended and alternate procODTs, if I recall correctly. 53 is definitely the greatest. Oh, and I am pretty sure my Bios lists 53.3, not 53.0. I assumed that is not a big enough deal.

Thanks for the pointers. I haven't combed through the overclocking forms yet, so I have a bit of research to do still.
Seti@home classic: 1,456 results, 1.613 years CPU time
ID: 2003052 · Report as offensive     Reply Quote
Ian&Steve C.
Avatar

Send message
Joined: 28 Sep 99
Posts: 4267
Credit: 1,282,604,591
RAC: 6,640
United States
Message 2003055 - Posted: 17 Jul 2019, 19:47:10 UTC - in response to Message 2003052.  

LLC = load line calibration. It’s Vdroop control.
Seti@Home classic workunits: 29,492 CPU time: 134,419 hours

ID: 2003055 · Report as offensive     Reply Quote
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13161
Credit: 1,160,866,277
RAC: 1,873
United States
Message 2003057 - Posted: 17 Jul 2019, 19:53:23 UTC - in response to Message 2003052.  

I'm not sure what LLC(3) is. I'd search for it online, but that acronym is conveniently the same as Limited Liability Company. Do you have a spot you can point me to off the top of your head?

Can I jump up to 60 ohms? The calculator lists 53, 48, and 40 for the recommended and alternate procODTs, if I recall correctly. 53 is definitely the greatest. Oh, and I am pretty sure my Bios lists 53.3, not 53.0. I assumed that is not a big enough deal.

Thanks for the pointers. I haven't combed through the overclocking forms yet, so I have a bit of research to do still.

Just a typo on my part. 53.3 ohms is the normal recommendation for 2 X 8 GB sticks. I have not actually tried to set the Calculator up for an APU before. I have always had every possible ProcODT valuable in the pull down menu in my BIOS. Your BIOS is probably very simplified compared to mine. Load Line Calibration prevents voltage droop under current loading. Can help stabilize the Vcore and Vsoc voltages under load so the cpu doesn't barf under Seti loading.
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 2003057 · Report as offensive     Reply Quote
Profile Tom M
Volunteer tester

Send message
Joined: 28 Nov 02
Posts: 5124
Credit: 276,046,078
RAC: 462
Message 2003061 - Posted: 17 Jul 2019, 20:03:30 UTC
Last modified: 17 Jul 2019, 20:03:55 UTC

I wonder.

I have what maybe a Zen 1 motherboard (Biostar bt350-BTC) running with an AMD 2700 (8c/16t) cpu.

I have a Asrock B450 Pro4 (presumably Zen+) MB running with an AMD 2600 (6/12t).

Right now both are running near 3.7GHz.

But the 2600 appeared to be crunching tasks faster than the 2700 when both were running 50% cpu to gpu without the "-obs" on the gpu command line.

So it makes me wonder if my supposition is right. A Zen 1 design MB is significantly slower than a Zen 1+ MB?

Tom
A proud member of the OFA (Old Farts Association).
ID: 2003061 · Report as offensive     Reply Quote
Profile Bill Special Project $75 donor
Volunteer tester
Avatar

Send message
Joined: 30 Nov 05
Posts: 282
Credit: 6,916,194
RAC: 60
United States
Message 2003063 - Posted: 17 Jul 2019, 20:35:07 UTC - in response to Message 2003057.  

Just a typo on my part. 53.3 ohms is the normal recommendation for 2 X 8 GB sticks. I have not actually tried to set the Calculator up for an APU before. I have always had every possible ProcODT valuable in the pull down menu in my BIOS. Your BIOS is probably very simplified compared to mine. Load Line Calibration prevents voltage droop under current loading. Can help stabilize the Vcore and Vsoc voltages under load so the cpu doesn't barf under Seti loading.
Ok, I have heard Load Line Calibration before, thanks Keith & Ian. I did set my SOC pretty low at first (1.05 V, as recommended at one speed), but found I had to bump it up to 1.0875 V. Even though the memory test ran fine at the lower SOC voltage, when I ran Timespy it would not complete. I should have enough to be dangerous, now I just need some time to play with it all.
Seti@home classic: 1,456 results, 1.613 years CPU time
ID: 2003063 · Report as offensive     Reply Quote
Profile Bill Special Project $75 donor
Volunteer tester
Avatar

Send message
Joined: 30 Nov 05
Posts: 282
Credit: 6,916,194
RAC: 60
United States
Message 2003064 - Posted: 17 Jul 2019, 20:40:04 UTC - in response to Message 2003061.  

I wonder.

I have what maybe a Zen 1 motherboard (Biostar bt350-BTC) running with an AMD 2700 (8c/16t) cpu.

I have a Asrock B450 Pro4 (presumably Zen+) MB running with an AMD 2600 (6/12t).

Right now both are running near 3.7GHz.

But the 2600 appeared to be crunching tasks faster than the 2700 when both were running 50% cpu to gpu without the "-obs" on the gpu command line.

So it makes me wonder if my supposition is right. A Zen 1 design MB is significantly slower than a Zen 1+ MB?

Tom
I'm using that same Pro4 board for my 2200G. My guess is based on what I paid for it it isn't a great overclocking board. That's fine, I am only OC'ing the memory, so whatever I can get out of it is fine.

Remind me never to come to your house, Tom. I don't know what magic you are doing over there but that isn't the first CPU you've had that is slower than the next processor lower. ;) To answer your wonder, I don't think there is much of a performance difference between the different AM4 motherboards.
Seti@home classic: 1,456 results, 1.613 years CPU time
ID: 2003064 · Report as offensive     Reply Quote
Profile Jord
Volunteer tester
Avatar

Send message
Joined: 9 Jun 99
Posts: 15184
Credit: 4,362,181
RAC: 3
Netherlands
Message 2003066 - Posted: 17 Jul 2019, 21:58:19 UTC

The first part of the new system is in, the memory. I'll get the rest together over the next weeks, but will wait with the actual motherboard and CPU until October. I also see that the new CPUs are very hard to get, most stores say 7 days and more.

Not in a hurry. At least that'll give the motherboard manufacturers time to iron out some of the kinks.
ID: 2003066 · Report as offensive     Reply Quote
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13161
Credit: 1,160,866,277
RAC: 1,873
United States
Message 2003070 - Posted: 17 Jul 2019, 22:38:00 UTC - in response to Message 2003063.  

Just a typo on my part. 53.3 ohms is the normal recommendation for 2 X 8 GB sticks. I have not actually tried to set the Calculator up for an APU before. I have always had every possible ProcODT valuable in the pull down menu in my BIOS. Your BIOS is probably very simplified compared to mine. Load Line Calibration prevents voltage droop under current loading. Can help stabilize the Vcore and Vsoc voltages under load so the cpu doesn't barf under Seti loading.
Ok, I have heard Load Line Calibration before, thanks Keith & Ian. I did set my SOC pretty low at first (1.05 V, as recommended at one speed), but found I had to bump it up to 1.0875 V. Even though the memory test ran fine at the lower SOC voltage, when I ran Timespy it would not complete. I should have enough to be dangerous, now I just need some time to play with it all.

SoC voltage can have a large effect on attainable overclocks and stability. You might try around 1.13 - 1.15V on the SoC. If the mobo wasn't a APU host, I would say go the other way and drop SoC down to 0.95V. A lot of people have found that cranking ever more voltage into the SoC is counterproductive for achieving higher memory overclocks. But again, the APU is whole different animal. Tuning memory on Ryzen is much harder than on Intel since the platform doesn't have ten years of prior experience like Intel. The Calculator really helps get you to a starting point faster.

Also Samsung B-die memory likes voltage. Up to 1.5V is acceptable. You have those in your Flare-X kit. I would raise the Vdimm voltage to 1.4V right off the bat and see if you can't get 3333 or 3400 stable.
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 2003070 · Report as offensive     Reply Quote
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13161
Credit: 1,160,866,277
RAC: 1,873
United States
Message 2003207 - Posted: 19 Jul 2019, 2:40:15 UTC
Last modified: 19 Jul 2019, 2:40:47 UTC

Get a Threadripper 2950X for $670 today at Amazon. https://www.amazon.com/AMD-Ryzen-Threadripper-Processor-YD295XA8AFWOF/dp/B07GFN6CVF
$230 off the regular price.
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 2003207 · Report as offensive     Reply Quote
Profile Tom M
Volunteer tester

Send message
Joined: 28 Nov 02
Posts: 5124
Credit: 276,046,078
RAC: 462
Message 2003216 - Posted: 19 Jul 2019, 3:44:34 UTC - in response to Message 2003207.  

Get a Threadripper 2950X for $670 today at Amazon. https://www.amazon.com/AMD-Ryzen-Threadripper-Processor-YD295XA8AFWOF/dp/B07GFN6CVF
$230 off the regular price.


Isn't that less than list price for the Ryzen 3950x (which I think is also a 16c/32t)?

Tom
A proud member of the OFA (Old Farts Association).
ID: 2003216 · Report as offensive     Reply Quote
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13161
Credit: 1,160,866,277
RAC: 1,873
United States
Message 2003219 - Posted: 19 Jul 2019, 4:32:05 UTC - in response to Message 2003216.  

Get a Threadripper 2950X for $670 today at Amazon. https://www.amazon.com/AMD-Ryzen-Threadripper-Processor-YD295XA8AFWOF/dp/B07GFN6CVF
$230 off the regular price.


Isn't that less than list price for the Ryzen 3950x (which I think is also a 16c/32t)?

Tom

Yes, that is correct. The SRP of the Ryzen 9 3950X is supposed to be $750.
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 2003219 · Report as offensive     Reply Quote
jsm

Send message
Joined: 1 Oct 16
Posts: 124
Credit: 51,135,572
RAC: 298
Isle of Man
Message 2003226 - Posted: 19 Jul 2019, 5:43:46 UTC - in response to Message 2003219.  

I wonder if it is significant that the 2920 and 2950 have been significantly price cut but the 2990 hasn't?
jsm
ID: 2003226 · Report as offensive     Reply Quote
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13161
Credit: 1,160,866,277
RAC: 1,873
United States
Message 2003231 - Posted: 19 Jul 2019, 7:27:21 UTC - in response to Message 2003226.  

I wonder if it is significant that the 2920 and 2950 have been significantly price cut but the 2990 hasn't?
jsm

I wouldn't expect a price cut on the 2990WX yet until there is competition. That won't happen until the Threadripper 3's arrive. The 2920 and 2950 have direct competition from the Ryzen 9 product line now.
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 2003231 · Report as offensive     Reply Quote
Profile Tom M
Volunteer tester

Send message
Joined: 28 Nov 02
Posts: 5124
Credit: 276,046,078
RAC: 462
Message 2003247 - Posted: 19 Jul 2019, 13:00:01 UTC - in response to Message 2003219.  
Last modified: 19 Jul 2019, 13:02:48 UTC

Get a Threadripper 2950X for $670 today at Amazon. https://www.amazon.com/AMD-Ryzen-Threadripper-Processor-YD295XA8AFWOF/dp/B07GFN6CVF
$230 off the regular price.


Isn't that less than list price for the Ryzen 3950x (which I think is also a 16c/32t)?

Tom

Yes, that is correct. The SRP of the Ryzen 9 3950X is supposed to be $750.


So is that going to be the "new" trending price? I guess I need to wait till mid-fall after the 3950x is shipping to see.
-edit--
I used the link and the price is down a tiny bit more. Plus there are a bunch of Amazon merchants that are offering it for a tiny bit lower. So I would say, probably a solid trend.

Tom
A proud member of the OFA (Old Farts Association).
ID: 2003247 · Report as offensive     Reply Quote
Profile Jord
Volunteer tester
Avatar

Send message
Joined: 9 Jun 99
Posts: 15184
Credit: 4,362,181
RAC: 3
Netherlands
Message 2003319 - Posted: 19 Jul 2019, 21:45:45 UTC

Despite earlier rumors, it seems that even motherboards without X570 chipset can offer PCIe 4.0 on the Ryzen 3000 CPUs. AMD denied this during computex bit, Asus has now published a list of various X470 and B450 motherboards that should at least offer partially PCIe 4.0.

The PCIe 4.0 lines of the CPUs should be at least partially usable with a new BIOS on various Asus motherboards with X470 and B450 chipset.

https://www.guru3d.com/news-story/ryzen-3000-asus-opens-up-pcie-4-support-for-selected-x470-and-b450-boards.html
ID: 2003319 · Report as offensive     Reply Quote
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13161
Credit: 1,160,866,277
RAC: 1,873
United States
Message 2003321 - Posted: 19 Jul 2019, 21:52:39 UTC - in response to Message 2003319.  

Except that ASUS pulled the BIOS that allows that at the last minute yesterday because they discovered another major bug. Still on AGESA 1.0.0.2 and the interim 2501 BIOS for C7H. Still on BIOS 0702 for the X570 platforms.
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 2003321 · Report as offensive     Reply Quote
Ian&Steve C.
Avatar

Send message
Joined: 28 Sep 99
Posts: 4267
Credit: 1,282,604,591
RAC: 6,640
United States
Message 2003324 - Posted: 19 Jul 2019, 22:23:35 UTC - in response to Message 2003321.  

I heard that ASUS pulled it because they got a stern "talking to" from AMD about it. from what I hear, PCIe wont be coming to the older boards.
Seti@Home classic workunits: 29,492 CPU time: 134,419 hours

ID: 2003324 · Report as offensive     Reply Quote
Previous · 1 . . . 20 · 21 · 22 · 23 · 24 · 25 · 26 . . . 69 · Next

Message boards : Number crunching : Ryzen and Threadripper


 
©2024 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.