[Review] Skinneelabs CPU Block Roundup #3 Results Database

[Vapor]

Will be updating the with XSPC Rasa data shortly, review goes up in a few days. CPU-360 Rev1.2 testing hit a snag so I'm onto the Enzotech Stealth and hopefully can also get the Ybris Black Sun done before I go to GOOC and on vacation after that (though I don't think I can get the Stealth and the Black Sun 100% tested before departure).

Reintroducing the "Overall Performance" chart as well....this time it's a mix of 5 parts thermal and 2 parts flow. Will be an especially meaningful metric during midplate/nozzle testing (though I do flatly prefer thermal performance, a combined metric does have value).

Still tossing ideas around for mounting system grading.

EDIT: Rasa table:

[Martinm210]

You might also consider some sort of "Overclock Potential" bar chart. Gabe notes it takes 10C to see a 1.5-2% overclock gain.

If we are seeing at most 9 degrees from best to worst, that's only 1.5%

So if you ran the GT at 4000Mhz, the best block would afford you 4060Mhz??

I think that helps put things into a more bottom line Mhz perspective.

Or if someone currently has a HK and they are thinking of "Upgrading" to the AC Kros....the difference is 2.2 degrees or .38%. 4000MHz-->4015MHz

IMHO, while temperature is something we all THINK we understand because everyone has the ability to measure....I don't think people really understand what temperature means to overclock.

Some sort of simple bar chart showing OVERCLOCK in MHZ would be a good way of speaking a common language to a larger audience.

[justin.kerr]

I may be way out of line here, but from my own, very poorly done tests. some blocks seem to scale better than others as CPU output rises. I know this would be a very exhaustive test to perform and Vapor needs a new CPU that will overclock higher to do it. lol
It would be a very interesting test.. unless I have too many variables, and poor testing methods, and that is what is causing the scaling to be different?

[MpG]

IMHO, while temperature is something we all THINK we understand because everyone has the ability to measure....I don't think people really understand what temperature means to overclock.

Some sort of simple bar chart showing OVERCLOCK in MHZ would be a good way of speaking a common language to a larger audience.

It would also be horribly depressing. Nothing like realizing that your $400 WC system got you less than 100Mhz over an air cooler.

Although seriously, regarding the 10*C=2% improvement, quite often, I think you run into temperature limitations (i.e. thermal throttling) before you technically hit that limit. What the water cooling also does does is give you the headroom to increase voltage/frequency, to reach the chip's limits without running into temperature walls (i.e. GTX480). And that's something that's a little harder to put into chart form.

[Martinm210]

It would also be horribly depressing. Nothing like realizing that your $400 WC system got you less than 100Mhz over an air cooler.

Although seriously, regarding the 10*C=2% improvement, quite often, I think you run into temperature limitations (i.e. thermal throttling) before you technically hit that limit. What the water cooling also does does is give you the headroom to increase voltage/frequency, to reach the chip's limits without running into temperature walls (i.e. GTX480). And that's something that's a little harder to put into chart form.

Ahh I see, I just figure it a better unit to compare with. When I started test it was just on flow rate and found that really tends to give some false understanding because flow rate changes just didn't amount to much regarding temperature. From the little I do understand it seems a similar relation is there between temperature and overclock.

I just figure if we are all cooling for max overclock, it would be awesome to compare overclock directly.

Perhaps that's too messy, just an idea. It would at least be a good tool to understand when deciding to upgrade existing harware. Is $80 really worth 15MHZ etc.

For example, I currently have an EK supreme v1 and I thought about upgrading to an HF or other newer block. Personally, I leave a good 500Mhz cushion in my overclock, so from a performance standpoint it doesn't pencil out. Now I have to make my decision on other factors than performance...etc.

[Vapor]

You might also consider some sort of "Overclock Potential" bar chart. Gabe notes it takes 10C to see a 1.5-2% overclock gain.

If we are seeing at most 9 degrees from best to worst, that's only 1.5%

So if you ran the GT at 4000Mhz, the best block would afford you 4060Mhz??

I think that helps put things into a more bottom line Mhz perspective.

Or if someone currently has a HK and they are thinking of "Upgrading" to the AC Kros....the difference is 2.2 degrees or .38%. 4000MHz-->4015MHz

IMHO, while temperature is something we all THINK we understand because everyone has the ability to measure....I don't think people really understand what temperature means to overclock.

Some sort of simple bar chart showing OVERCLOCK in MHZ would be a good way of speaking a common language to a larger audience.

That would be a big mess

And I just don't have applicable data for how this CPU, let alone any CPU, scales with temperature.

I may be way out of line here, but from my own, very poorly done tests. some blocks seem to scale better than others as CPU output rises. I know this would be a very exhaustive test to perform and Vapor needs a new CPU that will overclock higher to do it. lol
It would be a very interesting test.. unless I have too many variables, and poor testing methods, and that is what is causing the scaling to be different?

If I were to have a second test setting, it would be at lower settings. 4GHz at 1.5v (measured via DMM....CPUz says 1.55v) on a D0 (this chip was also binned from a selection of 5 as the hottest of the group) is about as hot as I can reliably do. Surely I'm already causing degradation at this kind of voltage and this why clocks are only at 4GHz--I want the chip to have a moderately long life so that I don't have to keep breaking test continuity. 4.4GHz at 1.4v isn't as hot or as stable as 4GHz at 1.5v.

That said, I have no intention of doubling my workload for data that doesn't add much value.

[Martinm210]

I just don't have applicable data for how this CPU, let alone any CPU, scales with temperature.

I have no idea either, that's why I'm wondering and why I'm interested. The first I heard anything was from Gabe after I asked a few weeks back. He said it takes about 10C lowering of temperature to see a 1.5-2% overclock gain or doubling the life of the silicon.

I know heat and temperature are two different things and it's the increase in voltage that creates the extra heat more than anything and that it's pretty darn tricky to define stability as well.

What I would like to understand better is the overclock significance between blocks. Obviously we can measure the resulting temperature differences between blocks...but is that temperature difference of any significant overclock value.

Basically, does one block allow you to overclock more than another?

If not, then we can say there's no statistical significance in overclock potential. There is a difference in temperature, but it's not large enough to allow an increase in processor performance...etc. The only real value would be a very small increase silicon life.

I gave up my a/c unit testing room to my kids, so I don't even have the ability to control ambient temps anymore or I'd try it myself. Just too hard to do any stability testing when my house varies in ambients from 50F to 75F throughout the day.

I have seen that my once stable Q6600 @ 3.6Ghz has degraded in about 1 year's time and no longer stable there despite the same setup.

This just seems to be an area that's not very well explored...would be awesome to learn more about it even if it was just one test of worst vs best as a supplemental or completely independent test.

[zeropluszero]

where does the Swiftech Apogee XT come into the graph?
seems to be missing from the high end blocks.

[Vapor]

where does the Swiftech Apogee XT come into the graph?
seems to be missing from the high end blocks.

Apogee XT is incompatible with my testbed's motherboard, unfortunately.

[zeropluszero]

Apogee XT is incompatible with my testbed's motherboard, unfortunately.

you didnt get a Gigabyte baseplate?

[Vapor]

you didnt get a Gigabyte baseplate?

I did....but it's still incompatible with the EVGA E758.

[Vinas]

Apogee XT is incompatible with my testbed's motherboard, unfortunately.

boo

[Vapor]

Enzotech Stealth numbers go up tomorrow!

Rasa review is delayed due to travel. Will say this about the mounting: really liked it the first mount but was really, really tired of it by the sixth mount (it's way better than AC/EK/HK but way behind Koolance/Swiftech/D-Tek). Overall, it's probably the best value out there right now--flagship performance at $55 is really attractive. That same price ceiling limits you to an Apogee GTZ, a Supreme LT, an HK3.0LC, or an Enzo Sapphire Rev.A, none of which have top-of-the-class performance.

[Vapor]

Updated with Enzotech Stealth!

[SlackerXL]

really interesting stuff
thanks for sharing
maybe on the next round you could add some new and cheap contenders like the phobya and the Zern PQ which seem to do quite well in regard to their msrp

[Vapor]

I have a Phobya here and have done prelim testing on it and it's just plain awful. I have more blocks to add to this round too....I don't see this round ending until 2011, at least.

[SlackerXL]

I have a Phobya here and have done prelim testing on it and it's just plain awful. I have more blocks to add to this round too....I don't see this round ending until 2011, at least.

i'd really love to see that in your testbed
cause so far i've seen quite well results in other reviews..
but since i haven't watched these reviewers
i cannot bet on their objectiveness

reviews like these two don't seem so bad from my pov

[Vapor]

Just got back from vacation and got started on testing the DimasTech Nethuns when the board and PSU committed murder-suicide

EPS 12V connector burned out completely--few inches of the 12V wires are totally stripped and corroded too. So much for the into-2011 prediction I made a few posts up

Back to the drawing board and to Newegg for the next testbed, sigh.

[Biffa]

Did you guys ever get any feedback from Heatkiller about the "silicon hack" that improved temps?

[Church]

Vapor: i suggest new Corsair AX PSUs for replacement. Good ripple supression/gold efficiency/fully modular.