News:

Members can see all forum boards and posts. Non members can only see a few boards.  If you have forgotten your password use this link to change it.
https://forum.boinc-australia.net/index.php?action=reminder

Main Menu

Upgrade ?

Started by TouchuvGrey, July 29, 2020, 05:51:58 PM

Previous topic - Next topic

TouchuvGrey

i have a homebuilt rig with Intel(R) Core(TM) i5-8600K CPU @ 3.60GHz [Family 6 Model 158 Stepping 10]
(6 processors) with 16GB DDR4 RAM  and a NVIDIA GeForce GTX 680 (2048MB) driver: 451.67 OpenCL: 1.2

    Would adding more RAM boost production or would i be better
off upgrading my video card to a GTX 1660 Super ?

Any idea how much improvement i would see from those options ?


                  Mike

tazzduke

Greetings TouchuvGrey

The CPU and RAM are definitely fine for crunching.

But yes your next step would be to upgrade the GPU.

A GTX 1660 Super or even a 1660 ti would be a good step

These cards runs cool and more energy efficient than the GTX 680, even quieter.

Just some food for thought, am sure other member will throw their 2 cents in as well.

I have a GTX 1660 Super and our team Captain (Dingo) has the 1660ti model.

Cheers



 AA 24 - 53 participant

NudgeyNR

Quote from: tazzduke on July 29, 2020, 06:55:03 PM

The CPU and RAM are definitely fine for crunching.

But yes your next step would be to upgrade the GPU.

Totally agree with tazzduke.
AMD Ryzen 9 5950X 96GB ram RTX 4070ti Windows 11 Ent 64bit
AMD Ryzen 9 3900X 64GB ram RTX 2080 Windows 10 Ent 64bit
AMD Ryzen 9 3950X 32GB ram RTX 2070S Windows 10 Ent 64bit
Intel XEON E5-2697-V2 32GB ram GT 1030 Windows 10 Ent 64bit
Intel i7 4790 16GB ram GTX 750 Windows 10 Ent 64bit
Intel i7 5930K 32GB ram RTX 2060 Super Windows 10 Ent 64bit
Intel i7 4770K 28GB ram GTX 1060 6GB Windows 10 Ent 64bit
Intel i5 3470 12GB ram Windows 10 Pro 64bit
Intel i5 3470 16GB ram Windows 10 Pro 64bit
Intel i5 4460 16GB ram GT 710 Windows 10 Pro 64bit
Intel i5 4200M 16GB ram Intel HD 4600 Kubuntu 22.10 64bit (NCI only)
Intel Atom C3538 6GB Kubuntu 20.04 (pi-hole) (VM Synology DS1819+ NAS 16Gb ram) 2cores
RPI4 8GB Kubuntu 22.10 64bit
RPI5 8GB Raspberry Pi OS 64bit
Crunching Projects - NumberFields, Asteroids, OLDK, Rakesearch, Universe, Einstein, Milkyway, iThena, WCG & Wuprop.



turbo0

#3
I agree too. I got a GTX 1650 for the low profile card and efficient but lacks the cuda cores. 1660 super and ti is a sweet spot. Find one at the right price and be a decent low power rig to crunch away on.

If your feeling lucky you could play the EVGA GeForce RTX 2060 KO lottery, I was tempted when i saw the cards cheap. The KO range is ment to be rebinned 2070 Super and 2080 GPUS and sometimes perform alot better. https://youtu.be/mUFRBnJdx3Y -Gamer nexus review

I'm just not lucky or a gambler so didn't like my odds and there is pros and cons to getting the defective crippled 2080 chip
Desktop 1 - Ryzen 9 7950x, RTX 4070ti, 64GB DDR5 6000,
Desktop 2 - Ryzen 7 5700x, RTX 3060ti, 16gb DDR4-3200,
Laptop - Ryzen 7 4800H, GTX 1660ti, 16gb DDR4-3200
SFF Dell - i7 8700, RX 6400, 16GB DDR4,
SFF Dell - i7 7700, Tesla P4, 16GB DDR4,
SFF Dell - i5 7500, GT 1030, 16GB DDR4
SFF HP - i7 3770, GTX 1650, 16GB RAM
Micro PC - N100, 12GB 4800mhz Ram


Folding@Home Stats

chooka03

I have no idea how the 1650's, 1660's, etc perform with Milkyway@Home. Radeon cards are king of Milkyway.
RX480's, RX580's, Radeon VII, or old R9 280X (but they are heavy on electricity)

The 16xx cards are very efficient and great for other projects than M@H like Primegrid.

Oh..this list will help.

https://milkyway.cs.rpi.edu/milkyway/gpu_list.php


TouchuvGrey

Thank you everyone. Chooka, i had not even considered
a Radeon card as i have been using NVIDIA cards since i
started modifying and building computers  15 years ago.

:banghead

              Mike

turbo0

Chooka is dead right. Milkway uses double precision which is 1/32 speed on modern NVIDIA and 1/16 of single float on modern Radeon. Depends on what projects your  running. A second hand Rx 480-rx580 would run amazing still for milkway. A Radeon VII is the dream card but never seen them for sale.
Desktop 1 - Ryzen 9 7950x, RTX 4070ti, 64GB DDR5 6000,
Desktop 2 - Ryzen 7 5700x, RTX 3060ti, 16gb DDR4-3200,
Laptop - Ryzen 7 4800H, GTX 1660ti, 16gb DDR4-3200
SFF Dell - i7 8700, RX 6400, 16GB DDR4,
SFF Dell - i7 7700, Tesla P4, 16GB DDR4,
SFF Dell - i5 7500, GT 1030, 16GB DDR4
SFF HP - i7 3770, GTX 1650, 16GB RAM
Micro PC - N100, 12GB 4800mhz Ram


Folding@Home Stats

chooka03

My Radeon cards are great from Einstein and Milkyway but I get flogged at Numberfields, Primegrid. A guess a mix of cards is good sometimes but I like space projects the most. I have thought about building a PC specifically for Primegrid. Intel CPU and I'd probably get a 1660 SUPER or a Ti...something like that because they sip on the power. A 2070 SUPER would be ideal price vs performance for me but the power usage goes up quite a bit compared to the 16xx cards.
The other option is the 5700XT I suppose. They are nearly as good as a Radeon VII at Einstein.... and yes Radeon VII's rarely appear for sale and when they do a second hand one sells for about the same price as they were new $900.00.

Traditionally, NVIDIA cards use less power than the Radeon cards if that's a factor. The 5700 may be an exception.

TouchuvGrey

i came over here from SETI@Home after crunching for them
for 16 years, so i am a space nut too. RX 580 is less  expensive
that a GTX 1660ti so i like that. Hmmmm, will have to think on this.


                              Mike

Daniel


TouchuvGrey

That is good.   If i'm reading this correctly the Radeon RX580 is about twice
as fasr as the GTX 1660ti on double precision ?

you can do direct comparisons om many different cards here
https://www.videocardbenchmark.net/


                                   Mike

turbo0

#11
Yeah AMD should be at least twice as fast as same tier Nvidia. I can run tasks on a 1660ti to compare. I will say the rx570/580 AMD Polaris GPUs run more optimal at Milkyway if that project is your focus. I have a 5600xt which if fine at Einsteins single precision but not optimal at Milkway. Not sure if its drivers or the way the Navi processes double precision or memory, not sure. It's not terrible but should be theoretically nearly x4 faster than my rx560 polaris card but only just over twice as fast. It should dominate a rx590 on paper but its maybe the same speed or even slower than the rx570 or rx580.

Milkyway@home Separation times - running single WU per GPU *credit much the same 227-240 points each
RX560 'polaris' -208 seconds  Benchmarked 164.7 GFLOP Double
5600xt 'Navi' - 135 seconds  Benchmarked 508.7 GFLOP Double
GTX 1660ti @80w TDP - 233 seconds  Benchmarked 178.6 GFLOP Double
Vega 7 integrated GPU - 361 seconds (just because)  Benchmarked 89.24 GFLOP Double

Times show raw power is not the best indicator of speed across cards and application


*** I found these card times off another site**  https://forums.anandtech.com/threads/milkyway-h-benchmark-thread-winter-2016-on-updated-1-2019-gpu-cpu-times-wanted-for-new-wus.2495905/
RX 580, GPU 1350 MHz  97.3s
RX 570, GPU stock 121s
RX 480, GPU ?? 110s
Desktop 1 - Ryzen 9 7950x, RTX 4070ti, 64GB DDR5 6000,
Desktop 2 - Ryzen 7 5700x, RTX 3060ti, 16gb DDR4-3200,
Laptop - Ryzen 7 4800H, GTX 1660ti, 16gb DDR4-3200
SFF Dell - i7 8700, RX 6400, 16GB DDR4,
SFF Dell - i7 7700, Tesla P4, 16GB DDR4,
SFF Dell - i5 7500, GT 1030, 16GB DDR4
SFF HP - i7 3770, GTX 1650, 16GB RAM
Micro PC - N100, 12GB 4800mhz Ram


Folding@Home Stats

TouchuvGrey

When i get my RX580, can i put it in the box along side of the GTX680 i have now ?

i've run dual ( Nvidia ) card setups before but i've never
tried mixing cards.

chooka03

Ooh I'm pretty sure that can be done but I've never tired it so don't have the answer sorry.

turbo0

I run AMD and NVIDIA cards side by side in all my machines. Never had issues. Sometimes the raw power of the AMD is good sometimes I need something CUDA compatibility. Works well it the Power supply can feed both.
Desktop 1 - Ryzen 9 7950x, RTX 4070ti, 64GB DDR5 6000,
Desktop 2 - Ryzen 7 5700x, RTX 3060ti, 16gb DDR4-3200,
Laptop - Ryzen 7 4800H, GTX 1660ti, 16gb DDR4-3200
SFF Dell - i7 8700, RX 6400, 16GB DDR4,
SFF Dell - i7 7700, Tesla P4, 16GB DDR4,
SFF Dell - i5 7500, GT 1030, 16GB DDR4
SFF HP - i7 3770, GTX 1650, 16GB RAM
Micro PC - N100, 12GB 4800mhz Ram


Folding@Home Stats