Posts by _hiVe*
log in
1) Message boards : Number crunching : help! customized app_info? [running specific GPU independently of project] (Message 11728)
Posted 2509 days ago by _hiVe*
Thanks a lot, all of you.

I will try playing with the two clients...see how it works, how much of a CPU strain will it add, etc.

I plan to run my ATI card here on CC, and nVidia card over at PG.

Hopefully i won't be getting cuda workunits at the wrong client and vice versa :)
2) Message boards : Number crunching : help! customized app_info? [running specific GPU independently of project] (Message 11713)
Posted 2511 days ago by _hiVe*
Hello everyone,

I have a little bit of a puzzle here to solve.

A while ago, I was wondering if its possible to run a multi GPU system [different vendors] while binding each GPU to a specific project, independently of each other.

Goal? To achieve better efficiency overall [power, time, investment, you name it...]

As you all know, its a well known fact, that its possible to mix different vendor GPU's [amd-nvidia] and use them across the board of various projects, as long as both the hardware+drivers and the project itself supports CAL/CUDA/openCL/, but what isn't a general knowledge though is, whether the more complex scenario [above] can be made possible.

Do you think, that by customizing the app_info file there is a way to make it all work?
If so, please shed some light on HOW :)

Thanks in advance!
3) Message boards : Number crunching : New MacBook Pro (Message 11615)
Posted 2521 days ago by _hiVe*
You could've gone with a previous generation MBP, they had nVidia GPUs--and you would've saved some money. Unless you ABSOLUTELY have to have the newest & best available.


I think the last gen. had the problem with those nVidia GPU's, lots and lots of RMA.
Perhaps he'd soon end up with a paperweight...not a pleasant thought.

IIRC that was also one of the reasons Apple switched camps...

cheers
4) Message boards : Number crunching : Dnetc shutting down (Message 11181)
Posted 2552 days ago by _hiVe*
Thats fked up about DnetC...the only project that payd well enough to consider running, other than Collatz, [but still better]. Primegrid is out of the question, I for sure wont be wasting my "Watts" where I get looked down upon...PG is a nvidia project, running Ati/amd there is just plain stupidity...no offense
5) Message boards : Number crunching : mixed vendor GPU computing with Lucid hydra... (Message 11129)
Posted 2554 days ago by _hiVe*
Thanks guys, I believe this will be helpful!

For the record I'm getting a MSI 870A Fuzion..it'll be running a 8800GTS 512, will probably try to add a HD5870 later on, or smth. My target projects are Collatz, PG, and DNETc.

I'll be continuously monitoring this thread for further info :)

cheers
6) Message boards : Number crunching : mixed vendor GPU computing with Lucid hydra... (Message 11125)
Posted 2555 days ago by _hiVe*
Thanks a lot Mikey!
I've been looking through some of the profiles and it appears, that there really aren't any special requirements such as a Lucid containing mobo.

cheers
7) Message boards : Number crunching : mixed vendor GPU computing with Lucid hydra... (Message 11122)
Posted 2555 days ago by _hiVe*
Hello everyone,

I was thinking of building a rig with a board utilizing a LucidĀ® LT22102 Chip. As I have a couple of GPU's from different camps [amd+nvidia] this would be a perfect solution to get rid of them w/o building a rig for each one of them.

The question is, would such a combo work at Collatz?
Individually, both AMD and Nvidia GPU's are supported, but would they work in tandem, say a 8800GTS 512 + HD5870 ?

My only interest is full utilization of the GPU's and purely scientific use of them at various BOINC projects, no gaming or multimedia involved.

I'v been looking for the answer to my question for a while now, sadly no answers were found. Please help.

Thanks a lot in advance.
8) Message boards : Number crunching : 6850 performance? (Message 10583)
Posted 2603 days ago by _hiVe*
Funny^^
Took you guys so much speculation, while a glimpse at the shader count would have made it clear to all, that it is in fact a HD6870 =)
9) Message boards : Number crunching : Google Ads (Message 10158)
Posted 2637 days ago by _hiVe*
So has any cash came in by now?
I wonder how much has been generated ~
10) Message boards : Number crunching : ATI RADEON 5670 average crunch time (Message 9983)
Posted 2658 days ago by _hiVe*
Beautiful indeed, and i'm not much of a fan of the GT220 and GT240 either, although I've considered them when I was making these low power "experimental" machines -> was a good choice to go with the other camp.

have fun~
11) Message boards : Number crunching : ATI RADEON 5670 average crunch time (Message 9979)
Posted 2658 days ago by _hiVe*
Hi Pepi,

Look up my hosts hooked up to Collatz. You'll see the 2 of them with the following GPU markings -> "ATI ATI Radeon HD5x00 series (Redwood)" both of which have a HD5670 installed.

They're both overclocked to 950mhz [core] I think, average run times are 1000sec's - in other words 16,6 min's per WU.

Awesome cards btw, especially for the price.

regards
Jan
12) Message boards : Number crunching : offline Benchmark for Collantz (Message 9950)
Posted 2662 days ago by _hiVe*
Great! hope you'll enjoy your stay.
13) Message boards : Number crunching : offline Benchmark for Collantz (Message 9940)
Posted 2663 days ago by _hiVe*
Hi Pepi,

There is no such thing as an offline bench for Collatz, join up and test them here, won't do any harm.
14) Message boards : Number crunching : Google Ads (Message 9939)
Posted 2663 days ago by _hiVe*
lul~
Elphidieus, guess its not that bad :D
In the end it'll get down to getting used to it anyway heh, at least this is a constant source of cash, unlike the fluctuating donations system.
15) Message boards : Number crunching : Google Ads (Message 9912)
Posted 2665 days ago by _hiVe*
Slicker, adds like that doesn't really make much, or am I wrong?
Why not add a "donate" button instead?
16) Message boards : Number crunching : DUAL 5970 ONLY RUNNING ON 3 GPU'S (Message 9727)
Posted 2680 days ago by _hiVe*
As far as i know the purpose for a dummy plug is to give the video card which is not connected to a real monitor a so called active status in the os.
Crossfire is not needed if your not gonna game on it anyway, the only thing you have todo is extend your desktop to the other cards.

So there is never a need for more then 1 dummy plug on 1 videocard single or double gpu does not matter, again if a monitor is attached the dummy plug is not needed.

The dummy plug just prevents the os from switching the card to inactive or sleep mode.

Well lol if you switch of the monitor the card prolly gets the status inactive again, hence not being used by the gpu projects :D

Every setup has its own problems indeed, and every os its own version of driver.
It even becomes complexer since displayport is added in the ways to connect a gpu.
I have had my fair share of problems with a dvi connector also, and a friend of mine is fighting with this displayport connector.


yea, tell me about it...
I was just wrestling with that problem this morning >.>
I think the OP could try the latest iteration of Win, 7x64
17) Message boards : Number crunching : 5870 - overclocking / RAC expectancy - etc. (Message 8100)
Posted 2798 days ago by _hiVe*
[/quote]

Can you please do me a small favor...?

I notice you've installed your 5870 on your C2D... If you have a kill-a-watt meter, can you please measure your rig power consumption at system idle, and when the GPU is at full load crunching Collatz...?

Thank you...[/quote]

I'd do that gladly, but the problem is I cannot find a device such as kill-a-watt on the market. I did order something similar a while back, but it doesn't show the consumption in real time, only a 24h average, which is far from accurate.
My estimate is that I peak at 350W-380W at full load, but its hard to prove w/o a power meter ^^
I will look around once again, and try to order something that actually works as advertised and provide you the data.
*note - one problem with my idle consumption may be that due to my extensive CPU o/c it doesn't switch back to lower power state, which may affect the overall results a lot...
18) Message boards : Number crunching : 5870 - overclocking / RAC expectancy - etc. (Message 8098)
Posted 2798 days ago by _hiVe*
After sessions of extensive testing, and sleepless nights heh, I'm ready to report my findings, that although the 5870 is probably worth it, and there are very few things that can match its performance at Collatz out there, overclocking the core may look far less rewarding [run-time wise] compared to lower end models.

Default run times were around normal 350ish [at 850 core/1200 mem] clocking to 1000 core/1250 mem brought them down to 300~310 sec/WU. That is a 12~14 % improvement over stock.

Ain't bad considering it was almost effort less, except 1 disassembly/cooler re-seating and TIM replacement due to severe issues with heat.

I heard ATI ships the SKU's with Arctic's MX3 thermal compound applied on the GPU, now I do not if that is true or what ASUS had on it, but the heat transfer just sucked! Applying AS5 gained me 25~30 degrees!! Perhaps just a bad sample, but keep an eye out for ASUS HD5870 V2, its marketed as more silent and with better cooling capacity, which is true considering its full cooper HSF and larger fan, but there may be other cards like mine out there that may appear to have thermal issues though a quick HSF re-seating and TIM replacement fixes the issue perfectly. Embarrassing and weird for a product of that category, they should check their manufacturing processes...

Anyway, next target is 1000+ on core, should be achievable since I have not touched the voltages one bit to reach 1000mhz, but the stock VCC was definitely higher than normal - 3D being at 1.212v. Any idea what the normal vCore is? Think somewhere in the 1.168 range?


19) Message boards : Number crunching : 5870 - overclocking / RAC expectancy - etc. (Message 8070)
Posted 2800 days ago by _hiVe*
The 5870 I've ordered should be arriving today, so I'll be updating/adding some #'s to this thread later on, hopefully that sample will clock very well :)

I find it "important" to cover this aspect of # crunching as well, hope there are others that appreciate it the same way I do.

This should prove to be a good day =)
20) Message boards : Number crunching : PCI-e speed... (Message 7939)
Posted 2808 days ago by _hiVe*
Hi,

This has been asked once. The Collatz computational speed is not affected by the PCI-e interface bandwidth.

gl


Next 20

Main page · Your account · Message boards


Copyright © 2018 Jon Sonntag; All rights reserved.