Milestones VI

MAGIC Quantum Mechanic
MAGIC Quantum M...
Joined: 18 Jan 05
Posts: 1888
Credit: 1412527824
RAC: 1192331

If we didn't have these

If we didn't have these Credits to look at there would probably only be about 10 members for a couple weeks and then they would move on to some other project.

And none of them would stay here since 2005

 

tullio
tullio
Joined: 22 Jan 05
Posts: 2118
Credit: 61407735
RAC: 0

We few, we happy few, we band

We few, we happy few, we band of brothers.

Tullio

Filipe
Filipe
Joined: 10 Mar 05
Posts: 186
Credit: 407207606
RAC: 354336

@IAN&STEVE C.I was

@IAN&STEVE C.

I was especifically refering to the FGRP1G search which is really faster with a higher Memory bandwitch

(My 2060 super is not faster than my old radeon R7 280X wich has a 384bit bus. 500.000 RAC for both cards.

Vega VIi can do a RAC of 1.000.000.

Is any one planning to run an AMD instict MI 200 card at einstein?

With a 3,2Gbps memory bandwith, is should be crazy fast at FGRP1G.

Later, the HBM3 memory will have a 8Gbps memory bandwith!!!

Ian&Steve C.
Ian&Steve C.
Joined: 19 Jan 20
Posts: 3960
Credit: 47060112642
RAC: 65297601

Filipe wrote: @IAN&STEVE

Filipe wrote:

@IAN&STEVE C.

I was especifically refering to the FGRP1G search which is really faster with a higher Memory bandwitch

(My 2060 super is not faster than my old radeon R7 280X wich has a 384bit bus. 500.000 RAC for both cards.

Vega VIi can do a RAC of 1.000.000.

Is any one planning to run an AMD instict MI 200 card at einstein?

With a 3,2Gbps memory bandwith, is should be crazy fast at FGRP1G.

Later, the HBM3 memory will have a 8Gbps memory bandwith!!!

 

its not just about memory bandwidth. 
 

my 3080Ti has less memory bandwidth and wholly outperforms a Radeon VII by a good margin. 
 

you should try the beta v1.28 app for your RTX 2060 Super, you’ll see a ~60% boost in speed. It will wreck your 280X. Einstein has long had a problem utilizing nvidia cards due to inefficiencies in the code. This was rectified in the new beta app. 

_________________________________________________________________________

Filipe
Filipe
Joined: 10 Mar 05
Posts: 186
Credit: 407207606
RAC: 354336

Ian&Steve C. wrote:   its

Ian&Steve C. wrote:

 

its not just about memory bandwidth. 
 

my 3080Ti has less memory bandwidth and wholly outperforms a Radeon VII by a good margin. 
 

you should try the beta v1.28 app for your RTX 2060 Super, you’ll see a ~60% boost in speed. It will wreck your 280X. Einstein has long had a problem utilizing nvidia cards due to inefficiencies in the code. This was rectified in the new beta app. 

 

Thanks for the tip.

I'm now running the windows 1.28 app. Let's see how the RAC behaves.

I've been away for long time, processing Covid-19 task from Rosetta + Folding

Ian&Steve C.
Ian&Steve C.
Joined: 19 Jan 20
Posts: 3960
Credit: 47060112642
RAC: 65297601

Filipe wrote: Ian&Steve C.

Filipe wrote:

Ian&Steve C. wrote:

 

its not just about memory bandwidth. 
 

my 3080Ti has less memory bandwidth and wholly outperforms a Radeon VII by a good margin. 
 

you should try the beta v1.28 app for your RTX 2060 Super, you’ll see a ~60% boost in speed. It will wreck your 280X. Einstein has long had a problem utilizing nvidia cards due to inefficiencies in the code. This was rectified in the new beta app. 

 

Thanks for the tip.

I'm now running the windows 1.28 app. Let's see how the RAC behaves.

I've been away for long time, processing Covid-19 task from Rosetta + Folding

 

if you’re running multiples on the nvidia card. Don’t do that. It runs Best with 1x task per GPU. Running multiples makes it run slower than 1x. This is the case with nvidia. AMD cards benefit from multiples though. 

_________________________________________________________________________

MAGIC Quantum Mechanic
MAGIC Quantum M...
Joined: 18 Jan 05
Posts: 1888
Credit: 1412527824
RAC: 1192331

Quote:if you’re running

Quote:
if you’re running multiples on the nvidia card. Don’t do that. It runs Best with 1x task per GPU. Running multiples makes it run slower than 1x. This is the case with nvidia. AMD cards benefit from multiples though. 

I agree

Over the years mine have all been Nvidia ( OC and SC ) and tried each one X2 and then ran singles and they all were faster running X1

I have a laptop these days that has one of the first Ryzen CPU's so I tried it running X2 and ended up running X1 since it gets more work done that way.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

I agree

Exard3k
Exard3k
Joined: 25 Jul 21
Posts: 66
Credit: 56155179
RAC: 0

Filipe wrote: I'm now

Filipe wrote:

 

I'm now running the windows 1.28 app. Let's see how the RAC behaves.

I've been away for long time, processing Covid-19 task from Rosetta + Folding

 

You'll be amazed by the leap in processing. My 2060 went from ~11:30 to ~6:50 per WU) with the new 1.28 app.

 

I tried Folding a bit, but after browsing and checking on some projects I saw more and more corporations issuing work. And I'm not supporting corporations with my computer(s), they can buy their HPC clusters themselves or they better pay their contributors. Max Planck Society with E@H is a totally different approach as a non-profit NGO.

 

Milestone....I'm shooting for my very first 5M credits. Joined only recently. Old CPU with a 2060 and not much RAM, but I can let it run most of the time. That 1.28 app for FGRP will help quite a bit. New system is configured, I only need AMD to release the new threadrippers and I can order it :)

Filipe
Filipe
Joined: 10 Mar 05
Posts: 186
Credit: 407207606
RAC: 354336

Quote:Exard3k wrote:   I'm

Quote:

Exard3k wrote:

 

I'm now running the windows 1.28 app. Let's see how the RAC behaves.

I've been away for long time, processing Covid-19 task from Rosetta + Folding

 

You'll be amazed by the leap in processing. My 2060 went from ~11:30 to ~6:50 per WU) with the new 1.28 app.

 

I tried Folding a bit, but after browsing and checking on some projects I saw more and more corporations issuing work. And I'm not supporting corporations with my computer(s), they can buy their HPC clusters themselves or they better pay their contributors. Max Planck Society with E@H is a totally different approach as a non-profit NGO.

 

Milestone....I'm shooting for my very first 5M credits. Joined only recently. Old CPU with a 2060 and not much RAM, but I can let it run most of the time. That 1.28 app for FGRP will help quite a bit. New system is configured, I only need AMD to release the new threadrippers and I can order it :)

 

Yes they are now running in 6:00 min

 

 

GWGeorge007
GWGeorge007
Joined: 8 Jan 18
Posts: 3065
Credit: 4971817686
RAC: 1430459

I just got to 500M on

I just got to 500M on Einstein alone with just two computers and three GPUs, an RTX 2060 & [2] RTX 2070 Super.

George

Proud member of the Old Farts Association

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.