Overclock.net › Forums › Overclockers Care › Overclock.net BOINC Team › Hardware Optimization › Ryzen 7 1700 computation time : any findings?
New Posts  All Forums:Forum Nav:

Ryzen 7 1700 computation time : any findings? - Page 6

post #51 of 54
Quote:
Originally Posted by mmonnin View Post

I just meant I wanted to spend money on that and this talk is temping me to get another system.

I couldn't sleep well last night, dreams of a 80 core/threaded demon trying to steal my credit card. eek.gif


@WhiteWulfe I ran some Denis typer.gif

Seems to be around 12-13k PPD? Not much difference between OS's, although I did have a small sample size.

If there's anymore projects that people are interested in just let me know, otherwise it will just go back on WCG.
post #52 of 54
I suspect an 80 thread demon of a rig would be nice to have in the long run ^_^
JägerWulfe
(19 items)
 
 
Gigadestroyer
(13 items)
 
CPUMotherboardGraphicsRAM
Intel i7-4770K Gigabyte Z87X-OC MSI GTX 980 Ti Gaming 6G Golden Edition G.Skill Triden-X 16GB DDR3-2400 
Hard DriveHard DriveOptical DriveCooling
Seagate 2TB SATA3 Maxtor 2TB Black LG DVD-RW Cryorig R1 ultimate 
OSMonitorMonitorKeyboard
Windows 7 Home Premium 64 bit BenQ VP2210 BenQ VP2210 Razer BlackWidow Ultimate 
PowerCaseMouseMouse Pad
EVGA SuperNova 1000G2 NZXT Phantom 820 Black Roccat Nyth Steel Series Worgen 
AudioAudioAudio
JDS Labs O2 + ODAC Rev.B AKG K712 Audio Technica ATH-M50X 
CPUMotherboardGraphicsRAM
i7 3770k Maximus V Formula EVGA GTX 780 Classified Corsair Dominator Platinum 1866MHz 
Hard DriveCoolingOSMonitor
Kingston 64GB SSDnow 200 CoolerMaster MasterLiquid Pro 240 Windows 7 64 Bit Dell E2213H 
KeyboardPowerCaseMouse
Logitech G15 v2 EVGA 1000w G2 Open Bench Table Corsair M95 
Mouse Pad
3' wide battle mat 
  hide details  
Reply
JägerWulfe
(19 items)
 
 
Gigadestroyer
(13 items)
 
CPUMotherboardGraphicsRAM
Intel i7-4770K Gigabyte Z87X-OC MSI GTX 980 Ti Gaming 6G Golden Edition G.Skill Triden-X 16GB DDR3-2400 
Hard DriveHard DriveOptical DriveCooling
Seagate 2TB SATA3 Maxtor 2TB Black LG DVD-RW Cryorig R1 ultimate 
OSMonitorMonitorKeyboard
Windows 7 Home Premium 64 bit BenQ VP2210 BenQ VP2210 Razer BlackWidow Ultimate 
PowerCaseMouseMouse Pad
EVGA SuperNova 1000G2 NZXT Phantom 820 Black Roccat Nyth Steel Series Worgen 
AudioAudioAudio
JDS Labs O2 + ODAC Rev.B AKG K712 Audio Technica ATH-M50X 
CPUMotherboardGraphicsRAM
i7 3770k Maximus V Formula EVGA GTX 780 Classified Corsair Dominator Platinum 1866MHz 
Hard DriveCoolingOSMonitor
Kingston 64GB SSDnow 200 CoolerMaster MasterLiquid Pro 240 Windows 7 64 Bit Dell E2213H 
KeyboardPowerCaseMouse
Logitech G15 v2 EVGA 1000w G2 Open Bench Table Corsair M95 
Mouse Pad
3' wide battle mat 
  hide details  
Reply
post #53 of 54
Quote:
Originally Posted by emoga View Post

I couldn't sleep well last night, dreams of a 80 core/threaded demon trying to steal my credit card. eek.gif

Now I think those chips stole my wallet..... {chuckle} (I'll settle for a 48core/threaded mini monster)
Quote:
Originally Posted by WhiteWulfe View Post

I suspect an 80 thread demon of a rig would be nice to have in the long run ^_^

Well, the MB & Memory is due to arrive by Thursday, the chips between the 1st and the 8th. I just might have 48 moar coars/threads to add to the second half of the pentathlon! biggrin.gif
Edited by Egilman - 4/25/17 at 1:21am
post #54 of 54
Thread Starter 
Asteroids@home
Asteroids@home is about 1 hour 10 min per WU on Stock Ryzen 7 1700X (3.5GHz XFR all core @ about 1.225V) and 3200MHz CL16 memory tightened to 16-16-16-36 AGESA 1.0.0.6, Linux Mint 18.2 (guest OS)
* Windows 7 SP1 64 bit VirtualBox 5.1.22 r115126 host
* SVM enabled in BIOS to allow for 64-bit guest OS
* on BOINC site listed as Linux 4.8.0-53-generic , BOINC Version listed as 7.6.31

Some AVX WUs seem to error out in Virtualbox ,or at least the ones I had in Asteroids@home. The sse2 and sse3 all are fine.



about 310 credits per hour CPU time
All WUs are 480 credits

Waiting on Manjaro result
Manjaro listed as Linux 4.11.7-1-rt3-MANJARO , BOINC client 7.6.33


about 280 credit per hour CPU time , but the sample size is much smaller

https://asteroidsathome.net/boinc/cpu_list.php lists Whetstone (floating point) so I would assume it is a primarily floating point application


Skynet POGS

First set of mint 18.2 results for Skynet POGS


... about 120 credits per hour of CPU time on average based on 220 WUs

Manjaro 17.0.2 results for POGs thus far

... about 120 credits per hour of CPU time on average based on 100WUs

Manjaro pending



Debian 9 xfce
* BOINC benchmark seems to report 4556 floating point whetstone but 69700 dhrystone integer which is very odd (higher integer by 3x vs Mint/Manjaro) , also no gui_rpc_auth.cfg password by default
* full reinstall resulted in 4804 floating point and 79183 integer
* identifies as Linux 4.9.0-3-amd64 and BoINC 7.6.33


= 130 credits per hour CPU time

Debian 9 lxde with SMT effects (i.e. multiple different VMs with 4 threads each) results in ~ 105 credits per hour of CPU time based on 50 WUs

Fedora 25 identifies as Linux 4.8.6-300.fc25.x86_64 with BOINC 7.6.22 , which updated to 4.11.8-200.fc25.x86_64
---> first few validations net ~ 110 credit per hour of CPU time in Fedora (~10% less)

Fedora 26 beta Linux 4.11.0-2.fc26.x86_64 with BOINC 7.6.22

= 125 credit per hour CPU Time , some cross CCX and SMT effects may be included

Fedora 26 Linux lxde with BOINC 7.6.22
=


Skynet POGS CPU List https://pogs.theskynet.org/pogs/cpu_list.php


LHC@Home ,sixtrack




Mint; only few units: ~42 credits per hour of CPU time
---> Thoughts/notes: faster than someone's Windows based i7-4770k and Intel Xeon E3-1270 v5 (Skylake 3.6GHz base)

Manjaro, : first few units are throwing odd credits per hour and one WU took over 24 hours
* appears to be ~ 17-18 credits per hour which is rather abysmal , I think it might have to do with the BOINC binaries being optimized for Debian



LHC CPU list https://lhcathome.cern.ch/lhcathome/cpu_list.php


Other projects that may be tested in the future:
* DENIS
* Cosmology (iffy due to camb_legacy being phased out for VBox units)
* Rosetta
* Einstein (mostly GPU so I don't see the value in CPU tests)
* Milkyway (mostly GPU)
* NumberFields
* NFS - not sure on this Number Field Sieve factorization

Also saw this thread from Ryzen launch time
Quote:
Originally Posted by https://www.reddit.com/r/BOINC/comments/5xog5v/boinc_performance_on_amd_ryzen/ 
Ryzen 7 1800X @stock - Built-in BOINC CPU Benchmarks: 4536 FP MIPS, 15295 Integer MIPS

Ryzen 7 1700X @ 3.7GHz - Built-in BOINC CPU Benchmarks - 4583 FP MIPS, 14623 Integer MIPS

Ryzen 7 1700X @3.5GHz - Built-in BOINC CPU Benchmarks: 4288 FP MIPS, 14529 Integer MIPS

Ryzen 7 1700 @3.9GHz - Built-in BOINC CPU Benchmarks: 4755 FP MIPS, 16497 integer MIPS

Ryzen 7 1700 @3.875GHz - Built-in BOINC CPU Benchmarks: 4764 FP MIPS, 16005 Integer MIPS

Ryzen 7 1700 @3.7GHz - Built-in BOINC CPU Benchmarks: 4532 FP MIPS, 16222 Integer MIPS

Ryzen 7 1700 @3.6GHz - Built-in BOINC CPU Benchmarks: 4351 FP MIPS, 13611 Integer MIPS

& this one recently
Quote:
Originally Posted by https://www.reddit.com/r/BOINC/comments/6k2k8y/ryzen_agesa_1006a_improved_my_boinc_speed/ 
Just an heads up for Ryzen owners, update your motherboard bios if you already haven't!

I went from AGESA 1.0.0.4 -> 1.0.0.6 -> 1.0.0.6a

FloatingPoint is up by 1.1%

Integer is up by 10.3%!

I can now also run my DDR4 ram at 2933mhz, before the latest bios update I could only run 2800mhz. This is free performance!

&
Quote:
Originally Posted by https://forums.evga.com/RYZEN-7-1800X-v-Intel-Core-i73930K-Processor-m2647189.aspx 
AMD Ryzen 7 1800X Outperforming the Intel(R) Core(TM) i7-5960X CPU on NumberFields @ home



Credits seem to vary mainly based on the CPU benchmark first and foremost.
I consulted the wiki and saw this:
http://boinc.berkeley.edu/wiki/Computation_credit
Quote:
Cobbelstone is 1/200 day of CPU time on a reference computer that does 1,000 MFLOPS based on the Whetstone benchmark

GigaFLOPs = RAC/200
TeraFLOPS = RAC/200,000

This explains the ~ 10% difference between distros. Some distros net 4800 - 5000 floating point benchmark while others have 4300 due to running as the second VM. It may be beneficial to overclock to the CPU turbo clock when running the CPU benchmark the first time, then return it back to stock.

Debian lxde gets me the highest CPU bench result due to the lightweight DE.

Regardless the results should be within 10% or so.
Edited by AlphaC - 7/18/17 at 8:19pm
Workstation stuff
(407 photos)
SpecViewperf 12.0.1
(136 photos)
 
Reply
Workstation stuff
(407 photos)
SpecViewperf 12.0.1
(136 photos)
 
Reply
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Hardware Optimization
Overclock.net › Forums › Overclockers Care › Overclock.net BOINC Team › Hardware Optimization › Ryzen 7 1700 computation time : any findings?