FurMark - How to use?

[g0dM@n]

So this is really the first time I'm using it. I'm not the extreme OCer I used to be and as I've been using higher-end single GPU cards for my games, I haven't felt the need for OC... esp b/c I only play Source games.

Anyway, my 4890 runs cool at 31C idle and low 40s under Xtreme burn mode on FurMark when OCed a bit, @ 1920x1200 8xMSAA.

I want to get the 24/7 OC for my card. It's got an MCW-4870 waterblock and man the temps rock in my setup.

1. What should I do as far as FurMark setup:
-What resolution
-What MSAA setting
-Any other settings

2. Should I run any 3dMark loops, if so which version and which tests?

3. When running FurMark, do I have to visually test for artifacts or does FurMark do that for me?

Thanks in advance. :)

[mxthunder]

I only use it short term <10 min at a time, to scan for artifacts. Afaik there is no auto way to scan for artifacting. I just look at the donut and the background and if I see any small black squares, I know to back down on clocks a little bit. VRM temps will climb very high during this test. I have seen them on my 4890's reach 100C. Thats when I know to back down. Your's may be a different story though, if you are under water, depending on if you have the stock red plate on or not.
I run it at the smallest resolution with no AA, in windowed mode. I think this creates a little more practical of a load. Using AA will actually create lower GPU temps than having it off, because it does not use the GPU as efficiently.
As for the 3dmark testing, yes the more different tests you do, the better. I would run a few loops of 05, 06, and vantage just to be sure.

[thideras]

1) Set it to your native resolution.
2) Put it in stability mode
3) I normally set 2x or 4x; enough that the core isn't starved for vmem bandwidth (play with it)
4) Hit go
5) Stare at the furry donut in amazement; mouth agape attempting to understand what it is

[g0dM@n]

Okay I ran native resolution 1920x1080 on my main monitor with 8x MSAA and man my load temp was 37C with a max of only 39C.

How/where do I check the VRM temps? I'm using Rivatuner and see these options:
Core clock
Memory clock
Core Temp
Core Temp \ display IO
Core Temp \ memory IO
Core Temp \ shader core
GPU usage %

[Neuromancer]

Furmark is terrible for stability testing :( at default settings (I have not tried "tweaking" it though. Will try to follow this thread :) )

I can run furmark with no errors at 1100/1100 on my 4890

Anything past 1000/1050 fails the newer futuremark test though :(

[thideras]

I don't see how it would be bad unless you are running it way below your native resolution...

[Neuromancer]

Scroll down in Riva, it should list your VDDC temps. (Have you applied the 4890 config file? And VT1165 plugin?)

[g0dM@n]

Scroll down in Riva, it should list your VDDC temps. (Have you applied the 4890 config file? And VT1165 plugin?)

not sure... any links?

Btw, no clue why but I think I can only run 955mhz GPU without artifacts. Is that typical? My load max temp was 40C only...

Also, how can I alter voltages without a hardmod (if that's possible)?

*Edit*
Just saw this
http://www.xtremesystems.org/forums/showthread.php?t=217002

So now that I've got VRM monitoring, voltage control... what are max gpu voltages for water for benching and for 24/7? If there's a thread, just link me to it so I stop wasting everyone's time here. :)

And I assume there's no vdimm control?

[g0dM@n]

So the card is going to level out around 950mhz GPU is my guess with no voltage adjustment... not too happy with the yield, but perhaps it's expected.

I want to know what gpu voltage is safe for 24/7 usage on water.
Default seems to be at 1.3125v
The card I have is the MSI R4890-T2D1G OC Radeon HD 4890 (http://www.newegg.com/Product/Product.aspx?Item=N82E16814127427).
Stock clocks are 880/1000.

Gonna try 1.4v for now...

[Neuromancer]

I have seen 1.45 max recommend for air, personally I run lower than that as the VRMs start hitting 80C on my stock cooler at far less than those volts. (GPU core stays relatively cool though with the stock fan spun up to 60%)

955 MHz stock volts sounds about right also (in my limited experience)

Overclocking in Riva, or using Afterburner or smartdoctor causes "screen jump" in 2D mode for me only when I OC the memory. I instead used riva for monitoring and voltage control and CCC (BIOS modded of course) for OCing. If you are keeping it up to 1000 GPU, then I would not even bother modding it.


Since you are not using a full cover block I would keep a careful eye on my VDDC temps and find my "max safe voltage" from that. For me on air That translates to about 1.375v Every card is different though and depending on how you are cooling them (stock cooler with GPU sink pulled out for the MCW or RAM sinks and a big ole fan) with stock heatsink completely removed...) should be able to figure it out pretty easy :)

Oh and adjusting MVDDC voltage did not help my memory overclocks at all :( (then again I only gave it a bump or two :) )

[g0dM@n]

I don't see adjustments for MVDDC voltage, nor for VDDC... which temp setting shows me VDDC?

Also, it's not a REAL full cover waterblock, but it is somewhat:
Swiftech MCW-4870 (http://www.swiftnets.com/products/mcw60-4870.asp)

It covers the ram and VRMs. Unfortunately, my phase 3 VRM temp seems to hit high 80s when running 1.40v on the GPU, and it seemed fine @ 1000mhz gpu...

Catch this... while stressing with Furmark at 1000mhz gpu (my first OC attempt with 1.40 volts), there was a surge, my backup UPS kicked in, my Logitech G15 keyboard went NUTS (the lights flickering), and then my system went into hibernation... I'm guessing I forgot to configure my APC powerchute software (UPS software) when I installed Win7 RC.

When I went to turn the machine on, my keyboard stroke wouldn't respond. The only thing I could do was get into BIOS, but then that was it... no other keystroke would work. I couldn't even push UP or DOWN in BIOS menu.

Long story short, I got everything working.. still don't know WTF happened, but it was my Bday tonight so I had to leave the system to rest several hours while I had dinner with family and my bros and watched the game.

Anyway, I'll have to get back to OCing, but I have to figure out a way to keep my VRM temps lower. What are the safe voltages for them?

[g0dM@n]

Did we figure out safe voltages?
Also, how do I cool the vregs... or does the Swiftech MCW-4870 do the job...

[g0dM@n]

Did we figure out safe voltages?
Also, how do I cool the vregs... or does the Swiftech MCW-4870 do the job...

Come on... someone help me out. :)

I'm trying 1.350vgpu on Furmark Xtreme burn with the results you see below (maximum temps) with this block here (http://www.swiftnets.com/products/mcw60-4870.asp) (temps measured with Rivatuner):
GPU: 985mhz, Mem: 1100mhz (was artifacting a bit though)
Core Temp: 44C
Core Temp \ Display IO: 45.50C
Core Temp memory IO: 49.50C
Core Temp \ shader core: 47.50C
VRM Phase 1 Temp: 70C
VRM Phase 2 Temp: 98C
VRM Phase 3 Temp: 111c

How can I lower these temps? (I have tons of RAMsinks and whatnot)... I don't really know what each temp represents. Perhaps I need a better mount on my waterblock? Again, check out the link to the block here (http://www.swiftnets.com/products/mcw60-4870.asp), so you can better understand how it's mounted.
I thought the MCW60-4870 was somewhat of a full waterblock... I guess it's not as "full" as I wanted... my core temps seem pretty good, though I still don't understand why there's 4 core temps!!