Skinnee Labs - GTX480 Block Round-Up (working thread)

[Biffa]

I've tried EK and AG and they are much of a muchness performance wise so far in my rigs

[gmat]

The bottom line is this: in attempting to produce the all inclusive, massive, thorough review that this appears to be, Skinnee may have overshot the window of ideal usefulness for many of the enthusiasts who would utilize the data.

Based on what ? We could replace that "many of" with "a few of" and there, that would feel better wouldnt it. He does that for free, he was kind enough to post a worklog, but given the feedback i would understand if he doesnt repeat the experience. The amount of work he does for free is very impressive by itself, i found this worklog quite interesting in between the dozens of "are we there yet ?" posts.
And as for the review deadline vs buying decisions, if i want to buy smart i never rush to get items when they are just out on market. In any case if you go with your favorite brand you are seldom disappointed...

[zeroibis]

Yes, in my case the data is just too important to buy something before I read it all. I will just use air cooling for now and then add the HT and throw it in a loop once I know what to buy. I designed my setup so that adding a new grafix card is as simple as putting the ht on and then prefilling it with coolant (I have QDs) then putting it in the loop. All I got to do then is bleed what little air was left (should be virtually none) and top off the res. Literally takes about an hour and so I will just wait until I know what I am going to buy

[Coldon]

i think that most experienced watercoolers can roughly guess the performance of the blocks just by looking at the block design, I think we pretty much all know the top three blocks will probably be the EK, Koolance and DD, swiftech's block will probably have lower GPU temps but much higher VRM temps, and the aquacomputer block is the only one I wont hazard a guess about since I have no idea what the the effect of the inlet/outlet design will be be.

From what other reviews I can find the koolance 470 block beats the BP and the EK 470 blocks. The koolance 480 block came out on top in a german round up, but as other posters have said the GPU temps will be mostly the same across the range apart from that joke of a BP block, its the VRM temps that interest me and skinnee is so far the ONLY person that has bothered to actually record the vrm temps, i think this roundup will be the only one that can definitively compared the overall block performance!! And that is why I am nervous.

I am hoping to get at least 850 on the core for 24/7 use.

[zeroibis]

I am hoping to get at least 850 on the core for 24/7 use.

I hope that this is what we all are hoping for

[justin.kerr]

850 on the core is what my 470 does 24/7 and thats on air... now where are the results..

[Martinm210]

I think from a practical overclocking potential, you won't find any difference regardless of what their thermal performance is. But if its degrees you're measuring, I would guess VGA thermal performance would be close between the EVGA and Koolance blocks, but the Koolance block may do better at overall cooling the non VGA areas.

BUT...thinking in terms of overall usage in the most common systems..I like the Danger Den block because of the low pressure drop. I suspect if you surveyed the market, you would find that 90% or more users will run combined CPU/GPU loops. It may not have all that surface area, but it probably has a good balance of cooling vs pressure drop. I've always been along the lines of spoiling the CPU as much as possible because it tends to be a bit more critical to cooling. By nature of the constrained slot area to work in, most GPU stock cooling systems tend to be poor in comparison to the CPU, as a result they usually run hotter and are designed to do so.

At least that's the way it used to be. That's my guess anyhow...gonna have to retire this old 8800gtx someday and upgrade..my old system is getting pretty outdated by todays standards..

That's just my guess though, the DD block does appear to be king of low pressure drop. In separate dedicated loops, that's irrelevant(all that matters is thermals), but in a combined loop with the CPU, I personally like having the lower restriction

[UrbanSmooth]

Martin, when will you grace us with your next build, sir?

[zeroibis]

I just want to know when I can dl data directly to my brain so I can copy what Martin knows about all this stuff. Always so many things to lean and so little time.

[gabe]

BUT...thinking in terms of overall usage in the most common systems..I like the Danger Den block because of the low pressure drop. I suspect if you surveyed the market, you would find that 90% or more users will run combined CPU/GPU loops. It may not have all that surface area, but it probably has a good balance of cooling vs pressure drop. I've always been along the lines of spoiling the CPU as much as possible because it tends to be a bit more critical to cooling. By nature of the constrained slot area to work in, most GPU stock cooling systems tend to be poor in comparison to the CPU, as a result they usually run hotter and are designed to do so.

I tried to remove the subjective (underlined your text above) part of this reflection by publishing a paper here which attempts to define the issue in obective (as in "measured data") terms

[Martinm210]

Martin, when will you grace us with your next build, sir?

I've never been much of a builder, I just don't have the creative juice or patience to really build much of anything compared to most guys in the forums.

I also don't have the $, I just gave myself a fathers day, or maybe early christmas? present in the form of a Canon 550D, so Ill have to let the ole wallet recover for a while anyhow.

Maybe Ill try making some custom one off blocks again...still have plenty of copper and delrin stock gathering dust in the garage. Some of the recent block building threads has sparked some desire to cool with totally hand made parts. I kind of miss running my own block, even if it sucks in comparison..there is still an element I miss in that..

[gabe]

While the heatpipe will keep the VRM's stable, I recommend a fan over the mosfets.

that's exactly my concern as well, i am more worried about VRM temps than GPU, i don't particularly care whether the care is sitting at 40 or 45 as long as the VRMs aren't at 70+.

especially when overclocking, the VRMs and power circuitry is the weak link in the chain and not the GPU, so I'm hoping the koolance does a good enough job in that regard, my loop is also pretty bare so the added restriction from the koolance pin array wont be too much of an issue, anything to slow down my infernal pump (i swear i got sold a 3.25 instead of a 3.2).

i think that most experienced watercoolers can roughly guess the performance of the blocks just by looking at the block design, I think we pretty much all know the top three blocks will probably be the EK, Koolance and DD, swiftech's block will probably have lower GPU temps but much higher VRM temps, and the aquacomputer block is the only one I wont hazard a guess about since I have no idea what the the effect of the inlet/outlet design will be be.

From what other reviews I can find the koolance 470 block beats the BP and the EK 470 blocks. The koolance 480 block came out on top in a german round up, but as other posters have said the GPU temps will be mostly the same across the range apart from that joke of a BP block, its the VRM temps that interest me and skinnee is so far the ONLY person that has bothered to actually record the vrm temps, i think this roundup will be the only one that can definitively compared the overall block performance!! And that is why I am nervous.

I am hoping to get at least 850 on the core for 24/7 use.

There is one thing that everybody forgets to mention:

EVGA provides a factory warranty for the FTW cards. So there should be little if any concerns at all about VRM temps on this block.

[Martinm210]

I tried to remove the subjective (underlined your text above) part of this reflection by publishing a paper here which attempts to define the issue in obective (as in "measured data") terms

Thanks Gabe,
Ill have to catch up, have you also done any recent work or experience regarding trying to quantify the relationship of temperature to overclock potential?

That's always something I never really felt like I understood very well. I always wondered what margin of temperature difference is really significant in terms of overclock stablity if overclocking is the bottom line goal. Is it 1C, 5C, 10?

Is it different for CPU vs GPU? I always assumed it was more critical for the CPU based on my very limted hardware experience. Most of our testing is all temperature based, so I'm not sure its very well understood.?? Or at least I don't know..

I understand flow rate effects better now, but I still struggle with the overclock relationship.

Thanks as always for your testing, its much appreciated!

[DarthBeavis]

BFG provides . . .uh, never mind. Doh.

[gabe]

That's always something I never really felt like I understood very well. I always wondered what margin of temperature difference is really significant in terms of overclock stablity if overclocking is the bottom line goal. Is it 1C, 5C, 10?

Is it different for CPU vs GPU? I always assumed it was more critical for the CPU based on my very limted hardware experience. Most of our testing is all temperature based, so I'm not sure its very well understood.?? Or at least I don't know..

There is a very simple rule of thumb to quantify the overclocking potential of any silicon:

For every 10C drop in temperature, the operating frequency can gain approximately 1.5 to 2% (without changing the voltage). If the frequency is not increased, then the life of the silicon doubles. I have verified the frequency aspect of this rule in CPU's over and over again, and it is still true.

For example, if you CPU reaches 4Ghz on air at say 90C, and if you drop your temps down to say 50C, you should be able to OC approximately another 6 to 8% (4.2 to 4.3Ghz) without raising the voltage.

Note: when I say "reaches 4Ghz", I mean stable frequency..

[Hondacity]

There is a very simple rule of thumb to quantify the overclocking potential of any silicon:

For every 10C drop in temperature, the operating frequency can gain approximately 1.5 to 2% (without changing the voltage). If the frequency is not increased, then the life of the silicon doubles. I have verified the frequency aspect of this rule in CPU's over and over again, and it is still true.

For example, if you CPU reaches 4Ghz on air at say 90C, and if you drop your temps down to say 50C, you should be able to OC approximately another 6 to 8% (4.2 to 4.3Ghz) without raising the voltage.

Note: when I say "reaches 4Ghz", I mean stable frequency..

in addition to frequency and temperature..oc also involves leakage current of cpu and vcc....

[skinnee]

There is one thing that everybody forgets to mention:

EVGA provides a factory warranty for the FTW cards. So there should be little if any concerns at all about VRM temps on this block.

Noted and more than likely assumed due to EVGA's warranty policy in the US. However, keeping temps low (for the entire card) while having a solid OC may be the goal... 80mm or smaller fan and some zip-ties are well worth it. My recommendation is not do it or your card will die... but if you have the fan and some zipties, go for it. Hell... with tri-sli you could probably zip tie a GT-1850 to the top and bottom card and keep true single slot spacing.

[SuperchargedZ06]

There is one thing that everybody forgets to mention:

EVGA provides a factory warranty for the FTW cards. So there should be little if any concerns at all about VRM temps on this block.

Well, there's all the hassle of dealing with a dead card - getting the RMA, removal, packing it up, waiting for a replacement, and then re-installing it. Not as trivial as an air cooled card. Have to bleed the system, worry about spilling fluid, messing with fittings... And the system down time to contend with. Kind of hard to run a loop that has a component missing without some re-config. Add the hassle of waiting a week or two before you can get things running again.

Not knocking EVGA or their warranty service - they are among the best. But I'd rather not have to deal with a card that died in the first place due to it being pushed a bit. If one plans on keeping things stock, I think the EVGA FTW block is great... but for those of us that want to push the cards a bit harder on water... as "Xtreme" folk tend to do, I don't think the block's design is as optimal as some of the other options that are out there.

Just my 2 cents...

[Martinm210]

There is a very simple rule of thumb to quantify the overclocking potential of any silicon:

For every 10C drop in temperature, the operating frequency can gain approximately 1.5 to 2% (without changing the voltage). If the frequency is not increased, then the life of the silicon doubles. I have verified the frequency aspect of this rule in CPU's over and over again, and it is still true.

For example, if you CPU reaches 4Ghz on air at say 90C, and if you drop your temps down to say 50C, you should be able to OC approximately another 6 to 8% (4.2 to 4.3Ghz) without raising the voltage.

Note: when I say "reaches 4Ghz", I mean stable frequency..

That's interesting and great info to keep in mind..thanks!!

[Coldon]

There is one thing that everybody forgets to mention:

EVGA provides a factory warranty for the FTW cards. So there should be little if any concerns at all about VRM temps on this block.

I wasn't implying that the swiftech solution wasn't up to scratch, personally I think looks wise its at the tops. If we could get the EVGA cards locally I would have bought one simply not to have to deal with mounting it myself and voiding my warranty.

You do bring up a very important point, that out of all the solutions, it is the only one that doesn't void the warranty and so for beginners it should be the preferred choice.

VRMs in general have a pretty high temp tolerance and maybe its not such a critical issue as many make it out to be. I've pretty much air cooled my vram and vrms on all my water cooled GPUs and never once had a dead card.

I'm just still bummed you guys don't have a block available that we can all buy.

[gmat]

BFG provides . . .uh, never mind. Doh.

I'll miss them too, and their factory mounted DD blocks
EK should strike a deal with Zotac (or other enthusiast brand) !

[Eddy_EK]

I'll miss them too, and their factory mounted DD blocks
EK should strike a deal with Zotac (or other enthusiast brand) !

I dont think Zotac are enthusiastic enough about the water, last time we talked (about a year ago) we were too aggressive

[gmat]

That's a shame, and a lack of a vision for a brand that openly tries to enter the enthusiast market.
Maybe Gainward then as they seem to love silent products ?... Now that BFG is gone i wonder who DangerDen will (or did) partner with. Anyway i'd happily buy factory watercooled cards with EK blocks on them With your SLI bridge you have a very nice all-in-one solution...
Is there another factory watercooled 480 apart from the EVGA/Swiftech one ?

[Biffa]

Eddy do you know if EK are going to bring out a block or blocks for the GTX460?

I know there seems to be a few different PCB designs which makes it a PITA but that never stopped you before!

Cheers

[Church]

Biffa: check first EK site for news before asking