My water cooled GTX 280 story.

[flesheatinvirus]

Yes this card is awesome on water. Here is a run I did this morning on a P5Q with a EVGA GTX280 with EK block on its own loop with a GTX240 rad. http://service.futuremark.com/compare?3dm06=9955987

[zeroibis]

WOW, very nice!

[T_Flight]

Ah, what is keeping it from me is vista... how is your score on default settings in 06 though I am interested in comparing a similar setup.

In fact it would be really interesting if you set your card to 702 1512 1269 because we have the same cpu at the same clock and the result should thus show how big of a difference ram/m/b makes!

Will do those runs later tonight. I did a run earlier at those high clocks and got near 22K... it was 21,97x I think 21,973 but cannot remember the last number.

It's just shy of 22K though. I'm using high BCLK and low multi. 21x Multi on the CPU for a 4.2GHz OC.

I'll run at reduced clocks later tonight. UFC is coming on in a few minutes, and I watch it religiously.

That 23,000 score above is a good one!

[T_Flight]

Ah, what is keeping it from me is vista... how is your score on default settings in 06 though I am interested in comparing a similar setup.

In fact it would be really interesting if you set your card to 702 1512 1269 because we have the same cpu at the same clock and the result should thus show how big of a difference ram/m/b makes!

Just for you I did the 3DM06 run at 702/1512/1269 and got a score of 21,130. My OS is not tweaked at all. No tricks, no tweaks. Just a straight up run at those clocks and the CPU clocks in my sig. I could get alot more out of it on XP tweaked, and using some other tweaks, but haven'y got to that yet.

I nioticed a distinct difference in the runs from the higher clocks on the card in my sig. The max FPS in both the HDR tests (Canyon Flight and Firefly Forest) were lower, and in all tests I didn't see the FPS recovery and rebound I did with the higher clocks...it just didn't have that "ummmph" like it does with the higher clocks.

[mit78]

flesheatinvirus,man but you modd de voltage,for this frequencie.with standard voltage its not possible to increase that much

[Talonman]

You do have a good 280 T_Flight...

750/1625/1250 !!

go 756/1620/1296... I bet it will work fine for you!

756 is still my max core, but runs error free all the time...

1620 on shaders is outstanding! 1566 is my absolute max, but normally keep them on 1512. Mine wont even tun at 1620, so you have better shaders than me for sure!!

Memore will run fine for you 1296, but can get dicey at 1323, 1332, or 1350.

Keep us posted, and glad you have a runner!!

[T_Flight]

I need an EVGA FTUW (FOr The Uber Win) BIOS with loosened mem timings, and need a slightly increased mem voltage if that's possible with these cards.

Heck I might just decide to seal up the boards and hoses, pour in some antifreeze, and drop the rad in dry Ice and pull a holeshot run. It's been a thought.

[flesheatinvirus]

flesheatinvirus,man but you modd de voltage,for this frequencie.with standard voltage its not possible to increase that much

No voltage mods where done to this card for that clock, it will actually do 820 core max.

[mit78]

i understand,but you have xp.try in vista 64,your score its not better than 20000 points

[zeroibis]

Just for you I did the 3DM06 run at 702/1512/1269 and got a score of 21,130. My OS is not tweaked at all. No tricks, no tweaks. Just a straight up run at those clocks and the CPU clocks in my sig. I could get alot more out of it on XP tweaked, and using some other tweaks, but haven'y got to that yet.

I nioticed a distinct difference in the runs from the higher clocks on the card in my sig. The max FPS in both the HDR tests (Canyon Flight and Firefly Forest) were lower, and in all tests I didn't see the FPS recovery and rebound I did with the higher clocks...it just didn't have that "ummmph" like it does with the higher clocks.

Thanks for the data! Yea I am running with no special tweaks done to anything. Although I would like to hear about them lol. It is interesting to see how close virtually identical systems preform. I noticed that although we are running the same cpu at the same clock we both used different methods. (I used multiplier)

As for the difference in power at higher clocks, I am recording my scores as I increase speed so I can see what the change is per 10mhz. I have excel setup to show the point difference as well as the % change in performance. Later I will be able to go back and compile the data and we should be able to see some interesting trends. I will post up my data once I have maxed the card out. What I also hope to do is go back and look at the temps at different speeds to also get data on heat output changes. I plan to do that after the fact though.

[flesheatinvirus]

Here is another run I just did this time with my Yorky at 4400mhz at 1.45v. Card is on stock voltage. The GTX280 still scales nicely with a faster CPU and is not maxed yet. My 9650 has more left as well. http://service.futuremark.com/compare?3dm06=10004906

[raidflex]

This thread has been very helpful as the OP has a very close setup to what have and great testing+ feedback.

My main problem is that with a Q6600 at 3.6 (1.45V in BIOS) and an EVGA GTX280 at 756/1566/2600 my CPU temps can get to 70C under prime95 small ffts load as measured by real temp.

For the GPU 58C seems to about the max it will get to while running furmark stability test, all other games/benchmarks don't really go above 52C.

Compared to the OP these do seem quite higher even taking into account a slightly higher ambient temperature.

What I am currently contemplating is the current swiftech MCR320 rad I have may not be enough for my system. I am wondering by upgrading to a PA120.3 would this net me at least about 4-5C cooler temperatures or more to justify the cost of the rad. I understand two loops do provide better cooling but I would like to stick to one loop also.

Current WC system:

Swiftech MCR320 Raditor
Swiftech MCP-655 Pump
EK Supreme Universal GPU block
Swiftech GTZ CPU Block
Swiftech MCRES Micro Reservoir
1/2" tubing
feser one Cooling liquid
BP Compression Fittings

I also am in the process of upgrading to a EVGA X58 Classified MOBO when it is released and 920 i7 processor, and from what I hear the i7's put out a lot of heat, although I do not know if it would be more or less heat then what my current Q6600, any info on this would be helpful.

[raidflex]

I also forgot to mention I have 3 Scythe S-FLEX SFF21F 120mm fans on the radiator pushing air through the radiator through the top of the case. The radiator is mounted horizontally on the top inside of the case.

[alpha1ma]

man i din believe that running core freq of over 800mhz was possible..

i tried to get mine higher..but seems like the max stable for my core is 720..

i have it under wc..three cards..but even with the adjusting of voltage to 1.3 yielded no higher clocks..

any other way to get round this? need higher voltage or sth?

note: will use precision tool to get an accurate reading on my frequencies...

[warriorpoet]

man i din believe that running core freq of over 800mhz was possible..

i tried to get mine higher..but seems like the max stable for my core is 720..

i have it under wc..three cards..but even with the adjusting of voltage to 1.3 yielded no higher clocks..

any other way to get round this? need higher voltage or sth?

note: will use precision tool to get an accurate reading on my frequencies...

Time to start working on those shaders

[zeroibis]

Have you checked your temps?

[alpha1ma]

Time to start working on those shaders

hmmm..righto..will get to it to see if it helps..

Have you checked your temps?

yes zero..i din notice much change with the voltage change..

heres one at 720/1250/1500 er..do pardon mi for using this not so accurate frequencies readings..will use precision tool to set next time..

Intel Core 2 Extreme QX 9650 @ 4.4 Ghz (400 x 11) 1.55V
3 x Asus GTX 280 1GB in 3 Way SLI @ 720/1250/1500

Software Used: 3DMark Vantage Professional Edition (As LOAD)
Software Used: CPU-Z 1.49
Software Used: GPU-Z 0.3.0
Software Used: Hardware Monitor 1.13
Ambient Temperature: 24 Degrees Celcius

CPU IDLE Temperature (in degrees celcius): 37 37 34 39
CPU LOAD Temperature (in degrees celcius): 65 60 61 67

GPU IDLE Temperature (in degrees celcius): 38 39 40
GPU LOAD Temperature (in degrees celcius): 48 55 62

[raidflex]

bump...

Anyone have any input or recommendations on my post above? Thanks.

[warriorpoet]

bump...

Anyone have any input or recommendations on my post above? Thanks.

Try 1566 then 1620 on the shaders. My mem doesn't clock very high, and for me, you already have a pretty decent memory overclock.

[raidflex]

Try 1566 then 1620 on the shaders. My mem doesn't clock very high, and for me, you already have a pretty decent memory overclock.

Thanks, but i'm looking more for any advice on maximizing my current Water cooling system and improving my temperatures.

Also if a core i7 CPU would dump much more heat then my current Q6600 or about the same. I do know Vcore voltage will have an effect on heat dump also but im just looking to confirm that the Core i7's will be in the same ballpark.

[zeroibis]

Personally I would try to do separate loops, this way there is no worry and best performance, especially for the cpu. If that is not an option, always consider having a bigger rad when ever heat dump is an issue. At the end of the day all the heat dump in the world should not matter if you have enough radiator capacity. Actually with those temps (in the burn) I would recommend more and bigger rads for your loop (assuming your pump is good). What is you coolant temp, I am a fan of water probes so you know how well the heatsink vs the loop is working to cool your system. Not to say those temps are bad, I am just a bit crazy about trying to keep things <60 but more importantly in the low 100's max.

[raidflex]

Personally I would try to do separate loops, this way there is no worry and best performance, especially for the cpu. If that is not an option, always consider having a bigger rad when ever heat dump is an issue. At the end of the day all the heat dump in the world should not matter if you have enough radiator capacity. Actually with those temps (in the burn) I would recommend more and bigger rads for your loop (assuming your pump is good). What is you coolant temp, I am a fan of water probes so you know how well the heatsink vs the loop is working to cool your system. Not to say those temps are bad, I am just a bit crazy about trying to keep things <60 but more importantly in the low 100's max.

I have the Swiftech MCP-655 pump which should be plenty.

I am considering either adding a second radiator or I would really prefer to just purchase a better radiator like a PA 120.3.

I do not have a temperature probe currently but that is a good idea and I will invest in one.

I agree with the lower temps even if what I have is within hardware spec I always feel more comfortable with lower temps.

I was also considering using a flow rate indicator but I am afraid that this will have too much of an affect on my flow rate to justify adding the device to the system. But if you have any idea's on this it would be much appreciated.

[flesheatinvirus]

I am running my GTX280 on its own loop with a EK Nickel block and GTX240 rad running with a Petras top modded Swifty MCP355 pump. Rad is remoted outside case and it works awesome. CPU has its own loop with a PA120.3 rad, Fuzion block and MCP655 pump.

[Talonman]

I wanted to make an update boy's...

I found the program LinX which runs the IBM LINPACK Test I believe...
It is a CPU heat generating monster for sure, and a stability checker all in one.

It didn't like my 3.8GHz setting, even though Prime and OCCT passed fine.

I set my FSB back to 414 to give me 3.7265GHz, as reported from cpu-z.

I lowered my BIOS CPU voltage to see the lowest that 3.7GHz could run LinX successfully.

BIOS voltage 1.48125v failed in Prime given 25 minutes.
BIOS voltage 1.48750v failed in Prime given 43 minutes.
BIOS voltage 1.49375v passed Prime and OCCT, but would not run a LINPACK Test error free.
BIOS voltage 1.50000v, or 1.48v in Everest runs everything I throw at it including LinX.

With me running at 3.7GHz, and lowering my core voltage down from 1.53750v in the BIOS to 1.50000v, my GPU idle temp reads 37C.
I consider this good given the CPU load that is on the system.

I think the entire temp of my loop likes the lower CPU voltage?

63C was the highest temp any CPU core hit during this test...

[raidflex]

I wanted to make an update boy's...

I found the program LinX which runs the IBM LINPACK Test I believe...
It is a CPU heat generating monster for sure, and a stability checker all in one.

It didn't like my 3.8GHz setting, even though Prime and OCCT passed fine.

I set my FSB back to 414 to give me 3.7265GHz, as reported from cpu-z.

I lowered my BIOS CPU voltage to see the lowest that 3.7GHz could run LinX successfully.

BIOS voltage 1.48125v failed in Prime given 25 minutes.
BIOS voltage 1.48750v failed in Prime given 43 minutes.
BIOS voltage 1.49375v passed Prime and OCCT, but would not run a LINPACK Test error free.
BIOS voltage 1.50000v, or 1.48v in Everest runs everything I throw at it including LinX.

With me running at 3.7GHz, and lowering my core voltage down from 1.53750v in the BIOS to 1.50000v, my GPU idle temp reads 37C.
I consider this good given the CPU load that is on the system.

I think the entire temp of my loop likes the lower CPU voltage?

63C was the highest temp any CPU core hit during this test...

Those temps to me at least seem very good. I have almost the same configuration as you but cannot seem to get my temps down. And with only 1.45V in BIOS on CPU I can hit 70C with prime95 small fft. And the GPU can hit 57C.

Any ideas on what I can do to help lower the temps?

Also what do you use for your cooling fluid?