Ryzen and Threadripper

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

To post messages, you must log in.

Previous · 1 . . . 34 · 35 · 36 · 37 · 38 · 39 · 40 . . . 69 · Next

AuthorMessage
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 2018013 - Posted: 6 Nov 2019, 22:56:42 UTC

The new GTX 1660 Super cards may be the new bang for the buck leader. No difference in core count between the old 1660 and the new Super, but the Super gets the same amount of memory but GDDR6 instead of GDDR5 memory. That should speed up the card for both gaming and crunching.
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 2018013 · 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 2018016 - Posted: 6 Nov 2019, 23:28:36 UTC - in response to Message 2018011.  

Whichever card it's going to be it's not for doing calculations here or at any other project. I have now got four Android devices doing that, the new system is going to be my gaming rig.
ID: 2018016 · 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 2018018 - Posted: 6 Nov 2019, 23:42:29 UTC - in response to Message 2018016.  

Oh. Then spend as much as you can for the highest FPS rating in your preferred games.
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 2018018 · Report as offensive     Reply Quote
Profile johnsmith

Send message
Joined: 13 Mar 19
Posts: 18
Credit: 2,712,800
RAC: 9
Netherlands
Message 2018073 - Posted: 7 Nov 2019, 16:43:40 UTC

https://videocardz.com/82654/amd-announces-ryzen-threadripper-3960x-3970x-and-ryzen-9-3950x

CPU Crunching times :O
The new threadrippers look awesome i am curious how there performance will be
ID: 2018073 · Report as offensive     Reply Quote
rob smith Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer moderator
Volunteer tester

Send message
Joined: 7 Mar 03
Posts: 22160
Credit: 416,307,556
RAC: 380
United Kingdom
Message 2018075 - Posted: 7 Nov 2019, 16:47:33 UTC

Link clickised:
https://videocardz.com/82654/amd-announces-ryzen-threadripper-3960x-3970x-and-ryzen-9-3950x
Bob Smith
Member of Seti PIPPS (Pluto is a Planet Protest Society)
Somewhere in the (un)known Universe?
ID: 2018075 · 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 2018090 - Posted: 7 Nov 2019, 20:05:31 UTC - in response to Message 2018073.  

https://videocardz.com/82654/amd-announces-ryzen-threadripper-3960x-3970x-and-ryzen-9-3950x

CPU Crunching times :O
The new threadrippers look awesome i am curious how there performance will be

I'm sure once the review sites get their hands on motherboards and cpus, the crunching performance will be phenomenal. Look for reviews at AnandTech and Phoronix for scientific application benchmark performance.

Still have a while for that to happen though. Maybe in December?
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 2018090 · 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 2018113 - Posted: 8 Nov 2019, 0:03:04 UTC - in response to Message 2018090.  
Last modified: 8 Nov 2019, 0:06:48 UTC

3rd party Bench marking:
Still have a while for that to happen though. Maybe in December?

Maybe. I think I read STH say their benchmarks being as through as they are, take 3 or more days.

https://www.pcgamesn.com/amd/ryzen-9-3950x-release-date
I am seeing posts that say that 2 Threadrippers(24 core and 32 core), the 3950x and one low end APU (2 cores, 4 threads, unlocked) will go on sale November 25, 2019.

Tom
A proud member of the OFA (Old Farts Association).
ID: 2018113 · 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 2018116 - Posted: 8 Nov 2019, 0:19:43 UTC - in response to Message 2018113.  

I am seeing posts that say that 2 Threadrippers(24 core and 32 core), the 3950x and one low end APU (2 cores, 4 threads, unlocked) will go on sale November 25, 2019.


From AnandTech today:

The short version of today’s announcement revolves around several parts, in completely different markets.

The Ryzen 9 3950X 16-core CPU for desktops
Ryzen Threadripper 3rd Gen Family: Starting with the 24-core 3960X and 32-core 3970X
New TRX40 motherboards for the new Threadripper Processors
A new $49 Athlon 3000G for the entry level market
All of this hardware is set to come to market through the month of November at the following dates:

Ryzen 9 3950X: Retail on November 25
TR 3960X and TR 3970X on November 25th
TRX40 motherboards on November 25th
Athlon 3000G on November 19th

Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 2018116 · 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 2018120 - Posted: 8 Nov 2019, 0:37:39 UTC - in response to Message 2017753.  
Last modified: 8 Nov 2019, 0:42:08 UTC

If you disable core performance boost, your CPU will operate only on its base clock and won't boost. You will loose performance, but its package power consumption will be close to the advertised TDP value.

You can also just set a manual all-core overclock you are comfortable with. And set the necessary voltage to run that. The plus side of doing that is once you set a manual multiplier the Vcore starts out at a default low value of 1.017V. Which runs very cool. You just have to add in some positive offset to get your set clock to run stable at the minimum required voltage. Which generates less heat and uses less power.


Thank you for the reminder. My AMD 2700 has been running without turbo-boost but I also hadn't examined the CPU voltage either. And for the last two weeks it has been "quitting" after maybe a week. I would notice it powered up but completely unresponsive with cold gpus.

I just looked in the bios and it was reporting 0.9xxxx volts. So I said, lots of head room and added a +0.5 volt offset. But when I look at the overclocking articles it seems like 1.4 volts is around the cpu maximum? So what voltage would you shoot for?
tom@MoonShot3:~$ sudo modprobe asus-wmi-sensors
tom@MoonShot3:~$ sensors
asuswmisensors-isa-0000
Adapter: ISA adapter
CPU Core Voltage:         +1.31 V  
CPU SOC Voltage:          +1.13 V  
DRAM Voltage:             +1.35 V  
VDDP Voltage:             +0.74 V  
1.8V PLL Voltage:         +1.85 V  
+12V Voltage:            +11.94 V  
+5V Voltage:              +4.99 V  
3VSB Voltage:             +3.36 V  
VBAT Voltage:             +3.23 V  
AVCC3 Voltage:            +3.36 V  
SB 1.05V Voltage:         +1.07 V  
CPU Core Voltage:         +1.33 V  
CPU SOC Voltage:          +1.13 V  
DRAM Voltage:             +1.36 V  
CPU Fan:                  866 RPM
Chassis Fan 1:           3708 RPM
Chassis Fan 2:              0 RPM
Chassis Fan 3:              0 RPM
HAMP Fan:                   0 RPM
Water Pump:                 0 RPM
CPU OPT:                    0 RPM
Water Flow:                 0 RPM
AIO Pump:                1408 RPM
CPU Temperature:          +52.0°C  
CPU Socket Temperature:   +38.0°C  
Motherboard Temperature:  +29.0°C  
Chipset Temperature:      +42.0°C  
Tsensor 1 Temperature:   +216.0°C  
CPU VRM Temperature:      +44.0°C  
Water In:                +216.0°C  
Water Out:               +216.0°C  
CPU VRM Output Current:  +55.00 A  

k10temp-pci-00c3
Adapter: PCI adapter
temp1:        +52.8°C  (high = +70.0°C)

asus-isa-0000
Adapter: ISA adapter
cpu_fan:        0 RPM

tom@MoonShot3:~$ ^C
tom@MoonShot3:~$ 


Tom
A proud member of the OFA (Old Farts Association).
ID: 2018120 · 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 2018125 - Posted: 8 Nov 2019, 1:23:22 UTC - in response to Message 2018120.  

Adding 500mV is total overkill and not necessary unless you want to make a space heater. I can just use the default BIOS setting for 4.025Ghz on the 40.25 multiplier for my 2700X hosts. The VID lookup table in the cpu and the BIOS sets the Vcore automatically to 1.35V more or less in the BIOS. With a LLC Auto setting in the BIOS, that droops down to 1.32-1.33V under BOINC load.

The 1.017V I mentioned in my previous post was with regard to my 3900X not the 2700X. You don't mention what your target cpu clock is going to be. If you want to run at the stock non-turbo base clock speed of 3.7Ghz for the 2700X, then I would either set the Vcore manually to around 1.25V or use the Offset function with a negative sign and put in 0.10V for the offset to knock 100mV off the base VID for that clock frequency.

You will just have to experiment. All silicon is different and you may have won the 'silicon lottery' . . . . . or not. Just pick a voltage that boots and then load the cpu with Prime95 and see if it holds up for a half hour. If it starts dropping threads, then reboot and add some more Vcore in the BIOS or drop the clock frequency for better stability.
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 2018125 · 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 2018126 - Posted: 8 Nov 2019, 1:33:02 UTC - in response to Message 2018125.  

I just looked in the bios and it was reporting 0.9xxxx volts.

Was that at 3.7Ghz base clock but unloaded? The cpu will drop to that voltage if you have Cool 'n Quiet enabled or called PSS Support in the C7H BIOS and with C-states enabled.

It won't be that voltage when under BOINC load. If your sensors report is for the 3.7Ghz base clock and BOINC loaded, then the 1.31V you show is what I would expect for that clock frequency under load. If you are still "black screening" that indicates not enough Vcore for the sustained cpu load and/or you are letting the voltage droop too far from the 1.31V and need to add in some LLC to keep it closer to the 1.31V.

I think I am running either LLC3 or maybe LLC4 load line calibration in the BIOS for the 2700X hosts. I also use the higher LLC settings for the memory and bump the current delivery for both cpu and memory to 120%.

It should be relatively easy to arrive at a stable system for just base clock. I have my systems run for over 2 weeks straight before rebooting them for an update.
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 2018126 · 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 2018153 - Posted: 8 Nov 2019, 8:21:40 UTC - in response to Message 2018126.  

I just looked in the bios and it was reporting 0.9xxxx volts.

Was that at 3.7Ghz base clock but unloaded? The cpu will drop to that voltage if you have Cool 'n Quiet enabled or called PSS Support in the C7H BIOS and with C-states enabled.


That voltage was in the bios during boot. I posted the loaded voltage later. I am going to have to go take a look at my bios settings. I am pretty sure "everything" is set to stock except ram voltage at 1.35 with the 3200 memory. I disabled the turbo so the 2700 runs right up there to just below 3.2Ghz which I believe is the stock maximum.

It is entirely possible that I can run "auto" with LLC3 and be done with it. Since my goal is stable, non-turbo processing at the lowest cost I can manage I will be happy if that works.

I guess I would have to download a copy of Prime95.... its to late/early to do that right now.

Thank you for the advice and reminders Keith.

Tom
A proud member of the OFA (Old Farts Association).
ID: 2018153 · 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 2018154 - Posted: 8 Nov 2019, 8:37:50 UTC - in response to Message 2018153.  

Just booted with 0.01 voltage offset and LLC3 (that was already applied). So apparently LLC3 + auto on the voltage wasn't enough to keep it from crashing.
The watch command is reporting 1.3 to 1.4Ghz.
tom@MoonShot3:~$  sudo modprobe asus-wmi-sensors
[sudo] password for tom: 
tom@MoonShot3:~$ sensors
asuswmisensors-isa-0000
Adapter: ISA adapter
CPU Core Voltage:         +0.86 V  
CPU SOC Voltage:          +1.13 V  
DRAM Voltage:             +1.35 V  
VDDP Voltage:             +0.71 V  
1.8V PLL Voltage:         +1.85 V  
+12V Voltage:            +12.15 V  
+5V Voltage:              +5.01 V  
3VSB Voltage:             +3.36 V  
VBAT Voltage:             +3.23 V  
AVCC3 Voltage:            +3.38 V  
SB 1.05V Voltage:         +1.07 V  
CPU Core Voltage:         +0.87 V  
CPU SOC Voltage:          +1.14 V  
DRAM Voltage:             +1.35 V  
CPU Fan:                  490 RPM
Chassis Fan 1:           2700 RPM
Chassis Fan 2:              0 RPM
Chassis Fan 3:              0 RPM
HAMP Fan:                   0 RPM
Water Pump:                 0 RPM
CPU OPT:                    0 RPM
Water Flow:                 0 RPM
AIO Pump:                1405 RPM
CPU Temperature:          +29.0°C  
CPU Socket Temperature:   +29.0°C  
Motherboard Temperature:  +29.0°C  
Chipset Temperature:      +42.0°C  
Tsensor 1 Temperature:   +216.0°C  
CPU VRM Temperature:      +36.0°C  
Water In:                +216.0°C  
Water Out:               +216.0°C  
CPU VRM Output Current:   +0.00 A  

k10temp-pci-00c3
Adapter: PCI adapter
temp1:        +29.6°C  (high = +70.0°C)

asus-isa-0000
Adapter: ISA adapter
cpu_fan:        0 RPM

tom@MoonShot3:~$ 

After starting up the Boinc Manager/Seti@Home + WCG, the cpu runs upto just under 3.2Ghz which I believe is the non-turbo top speed.
tom@MoonShot3:~$ sensors
asuswmisensors-isa-0000
Adapter: ISA adapter
CPU Core Voltage:         +1.06 V  
CPU SOC Voltage:          +1.13 V  
DRAM Voltage:             +1.35 V  
VDDP Voltage:             +0.72 V  
1.8V PLL Voltage:         +1.85 V  
+12V Voltage:            +11.99 V  
+5V Voltage:              +4.99 V  
3VSB Voltage:             +3.36 V  
VBAT Voltage:             +3.23 V  
AVCC3 Voltage:            +3.33 V  
SB 1.05V Voltage:         +1.07 V  
CPU Core Voltage:         +1.07 V  
CPU SOC Voltage:          +1.13 V  
DRAM Voltage:             +1.36 V  
CPU Fan:                  729 RPM
Chassis Fan 1:           3276 RPM
Chassis Fan 2:              0 RPM
Chassis Fan 3:              0 RPM
HAMP Fan:                   0 RPM
Water Pump:                 0 RPM
CPU OPT:                    0 RPM
Water Flow:                 0 RPM
AIO Pump:                1407 RPM
CPU Temperature:          +44.0°C  
CPU Socket Temperature:   +34.0°C  
Motherboard Temperature:  +27.0°C  
Chipset Temperature:      +40.0°C  
Tsensor 1 Temperature:   +216.0°C  
CPU VRM Temperature:      +38.0°C  
Water In:                +216.0°C  
Water Out:               +216.0°C  
CPU VRM Output Current:  +47.00 A  

k10temp-pci-00c3
Adapter: PCI adapter
temp1:        +44.2°C  (high = +70.0°C)

asus-isa-0000
Adapter: ISA adapter
cpu_fan:        0 RPM

tom@MoonShot3:~$ 

If this doesn't crash or survives the Prime95 test later. I will stop tinkering since I have met my minimum goal. (reliable and cheap cpu production).

Tom
A proud member of the OFA (Old Farts Association).
ID: 2018154 · 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 2018163 - Posted: 8 Nov 2019, 13:09:13 UTC - in response to Message 2018154.  

It is now running maybe 4C cooler than it was with the higher offset.
A proud member of the OFA (Old Farts Association).
ID: 2018163 · 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 2018187 - Posted: 8 Nov 2019, 17:41:27 UTC - in response to Message 2018163.  
Last modified: 8 Nov 2019, 17:42:41 UTC

It is now running maybe 4C cooler than it was with the higher offset.

OK, I mistook you had a 2700X and not a 2700. Yes, the base clock for that cpu is 3.2Ghz. That is an awfully low Vcore. You only put in a 10mV offset and it appears from your previous sensor output and this recent one, you applied it with a negative offset. I would have put in a 100mV offset or whatever it takes to arrive at 1.2V under load.

If it runs P95 without dumping threads, then all is good I guess. I would have expected to need about 1.2V for 3.2Ghz. If it is stable at that 1.1 voltage, then you get lower temps and lower power consumption.
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 2018187 · 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 2018189 - Posted: 8 Nov 2019, 17:45:40 UTC

Just booted with 0.01 voltage offset and LLC3 (that was already applied). So apparently LLC3 + auto on the voltage wasn't enough to keep it from crashing.
The watch command is reporting 1.3 to 1.4Ghz.

That low core clock indicates you are not giving the cpu enough voltage to run at its 3.2Ghz base clock. The cpu is downclocking to the 1.3Ghz range which probably aligns with the 1.09V sensor output in the VID table for that clock.
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 2018189 · 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 2018214 - Posted: 8 Nov 2019, 23:12:20 UTC - in response to Message 2018189.  

Just booted with 0.01 voltage offset and LLC3 (that was already applied). So apparently LLC3 + auto on the voltage wasn't enough to keep it from crashing.
The watch command is reporting 1.3 to 1.4Ghz.

That low core clock indicates you are not giving the cpu enough voltage to run at its 3.2Ghz base clock. The cpu is downclocking to the 1.3Ghz range which probably aligns with the 1.09V sensor output in the VID table for that clock.


That speed is the cpu unloaded. So only the loaded numbers seem to fit....

Tom
A proud member of the OFA (Old Farts Association).
ID: 2018214 · 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 2018215 - Posted: 8 Nov 2019, 23:16:31 UTC - in response to Message 2018187.  

[quote] I would have put in a 100mV offset or whatever it takes to arrive at 1.2V under load.{/quote]

Thank you for the target. I will try the cpu test to see but... now I have an idea for what I am trying to stay under if I need to up the voltage.
Here test.... Here test.... :)
A proud member of the OFA (Old Farts Association).
ID: 2018215 · 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 2018217 - Posted: 8 Nov 2019, 23:35:53 UTC - in response to Message 2018215.  

[quote] I would have put in a 100mV offset or whatever it takes to arrive at 1.2V under load.{/quote]

Thank you for the target. I will try the cpu test to see but... now I have an idea for what I am trying to stay under if I need to up the voltage.
Here test.... Here test.... :)


Found a Prime95 for linux (apparently) but while I can click it to execute it doesn't....
A proud member of the OFA (Old Farts Association).
ID: 2018217 · 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 2018244 - Posted: 9 Nov 2019, 6:52:04 UTC - in response to Message 2018217.  

[quote] I would have put in a 100mV offset or whatever it takes to arrive at 1.2V under load.{/quote]

Thank you for the target. I will try the cpu test to see but... now I have an idea for what I am trying to stay under if I need to up the voltage.
Here test.... Here test.... :)


Found a Prime95 for linux (apparently) but while I can click it to execute it doesn't....

It needs to have its execute permission set as usual for any downloaded linux executable. Just navigate to the Prime95 folder and open a Terminal session, then run
./mprime

and choose 15 for the stress test menu. Then either accept the default blend Torture Test option or choose alternate stress levels. I think that Options 15 >> 10 threads out of 16 for the 2700 are similar to the normal BOINC workload, then >> Option 11 small FFT's with a bound of 8K and 24K best approximates what the Seti tasks look like to the cpu. I also add in 2000MB of memory per thread to load the memory subsystem. Then finally I run each prime for 2 minutes and then rotate to the next prime. Run for at least 30 minutes or 1 hour and if you still have all of the thread count you specified running without dropping out, you are golden at that clock speed and voltage for cpu.
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 2018244 · Report as offensive     Reply Quote
Previous · 1 . . . 34 · 35 · 36 · 37 · 38 · 39 · 40 . . . 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.