Swiftech Storm -Temps Up, but OC's higher too??

[Kunaak]

got a swiftech storm here, and temps are wierd.

they jump quickly from 31C to 36C randomly.
it's extremely hot today here in alaska (about 90F) , so the temps aren't a suprise, but the flucaution is.

its a new block for me, since I just been making brackets for a old AMD Xp DTek Spiral for the last 2 years to fit it on anything from A64's to P4's.

but so far, this block seems unstable almost.
unpredictable, is how I would describe it so far.

but its strange, cause my OC is up by 70 mhz, now topping out at 3050 mhz from my SD 3700 benchmarkable on water.
nice, but strange.
cause the OC is up, but the temps at my daily speed (2.9 ghz daily) , are also up...

any idea whats going on here?

[Absolute_0]

I think that the G4 is performing well, as is attested by the increased overclock. As to the weird temps... well don't trust what the temp programs say :)

[Kunaak]

in this case, I am inclined to trust the temp program.
it's been reliable and predictable for a few months now.
until this block atleast.

[Absolute_0]

The only way a waterblock can affect overclocking is by lowering the temps; thus if you are overclocking further the temps must be lowered. It's a swiftech G4; chances are its an upgrade from whatever you were running before, so lower temps and higher OCs are anticipated. Myself, i do not put much faith in temp programs.

[MaxxxRacer]

Dont be inclined to trust the program.. EVER.

a few others have noticed higher temps with the G4, measured by software that is, but realworld results proved to show the opposite.

in any case, just make sure u crank down the spring as far as it will let u (with the spacer in there) and you will be fine. this will pretty much assure that you are not getting a bad mount, short of doing something catostrophically wrong.


On a side note. i have a feeling that the way the storm cools the cpu is different from most blocks and this is causing a possible rise in temperature from the readings. might be a combo of the sensor placement (which is on the edge of the core [amd morons]) and the way the storm works..

[Absolute_0]

Oh, its definitely true to tighten the G4 down all the way. I had a bad mount and i was idling in the high 40s... at 1.2 volts. Tightened her down REAL tight... went down to low 30s :).

[cartmanea]

Kunaak, how did you apply the thermal compound? BB-sized dot in the middle or spread evenly with credit card or razor? With heatspreaders you should do the BB-sized dot in the mittle to eliminate air pockets getting trapped.

Perhaps if you have an air pocket directly over the core, the block contacts the rest of the heatspreader and gets ok temps, but when there is a spike in cpu activity it takes a minute to catch up?

[Cathar]

This phenomenon is something that I've seen myself time to time with the Cascade ad Storm blocks in testing on IHS capped CPU's, and MaxxxRacer is correct, it appears to be with how the Storm does its job and the assumptions that CPU makers make with the user-readable thermal die sensors. CPU makers seem to assume that a device that is providing a uniform cooling effect will be applied to the entire IHS surface with respect to thermal measurements, while my block designs focus on cooling the hottest spots.

In general I've struggled to get consistent temperature results with IHS capped CPU's when using my focused cooling block designs, but almost always I've seen consistently better overclocks, and in the absence of consistent temperature results I often used overclocking as a means to guage if I was improving things with a design change.

Now I say "almost always" because there is one thing that can happen that can mess things up, and that is a concave heat-spreader. In this scenario the cooling device will mostly be cooling the outer edges of the heat-spreader, and for the focused cooling designs this is where they apply the least cooling effect. I believe that this was the issue that affected eva2000 with his CPU in comparison to an MCW6002. When he mounted the Storm block on a different CPU, one that I suspect did not have a concave IHS, the Storm block did its job and worked better than the MCW6002.

Still, for the OP, it doesn't sound to me like his IHS is concave. If that were the case then he would not be seeing a better overclock. This just sounds like one of those inconsistent thermal probe scenarios relating to the sometimes weird interactions that happen between focused cooling designs on IHS capped CPU's.

I seriously don't like IHS's from a consistency point of view with respect to applying a consistent cooling effect to the CPU die. Some are convex, some are concave, and over time the thermal interface between the CPU and the IHS will irrevocably degrade after many mountings.

IHS's make sense for air-cooling, and for protecting the CPU dies from damage due to ham-fisted installers, but for consistent cooling of a CPU die, I believe they are a serious step backwards.

[Kamel]

I've noticed this thing just two years ago testing some waterblocks made from a friend of mine.

And i think i have found a reason.
We should know what really happens inside the CPU core, we don't know if the temperature readings can measure any peaks ( i don't think so), maybe in a very little time ( less than a second) we can have a 5°..10°..15° ( during a bench ora a game, depending on the operations requested to the cpu)temperature increasing not registered from the temp controller, and this can result in a system crash even if the temperature seems to be "normal".
(In overclock condition naturally)

So some waterblock can "contain" theese peaks, other wb cannot...
just for the reason they can't "absorb" temperature peaks quickly!

This is the reason (IMHO) why ,evenif we can read higher temps, we can reach as well better overclock levels...just because some blocks can keep a better linearity in cooling performances.

Byez


Kam

p.s. i hope my english is good enough.... :) just to explain what i mean..

i'd like to know u're opinion about my idea.....

[moonlightcheese]

well this brings up another question:

i've seen a lot of people removing the IHS altogether. is this recommended? if i remove the IHS should i apply less pressure on the block?

[mad mikee]

A good way to mount wb (or TTBT in this case) (http://www.xtremesystems.org/forums/showthread.php?p=938347#post938347)
How I adjust for best contact (http://www.i4memory.com/showthread.php?p=14162#post14162)
:toast:

[eva2000]

yeah i had prob with original G4 Storm with my 3700+ SD, 3200+ winchester and now my FX-57 but not my 3000+ venice http://i4memory.com/showthread.php?t=954

Now with FX-57/DFI NF4 SLI-D yesterday I swapped from original G4 Storm to my MCW6002 i got boost in max oc from 3014mhz 1.392v and 3091mhz at 1.504v to 3105mhz at 1.456v - no setup change just straight waterblock swap.

My Swiftech G4 Storm arrived and did test mount on my 2nd DFI NF4 SLI-DR board and it seems the way the mounting is setup i have a feeling i'll get more consistent mounts this time round (crosses fingers). Gonna swap from MCW6002 to swiftech G4 Storm later this week :)

FYI, both original G4 Storm and MCW6002 exhibit the same immediate jump in load temps usually between 8-13C on DFI NF4 SLI-D according to smartguardian and MBM5 custom data profile.

so beautiful

http://www.fileshosts.com/watercooling/DIY/G4Storm_Swiftech/1/web/G4_Storm_008.jpg

http://www.fileshosts.com/watercooling/DIY/G4Storm_Swiftech/1/web/G4_Storm_016.jpg

[jinu117]

I just got off phone with someone in AMD talking about thermal diode inside CPU. Quite funny that the engineer would be recommending to turn off MOBO thermal shut off :P For 90nm core especially, he mentioned the thermal probing method changed slightly and majority of boards and bioses out there didn't bother reflecting this modification. One company he mentions that implemented was ABIT. Also, his wording of top 3 vendor on mobo (Abit, MSI, Gigabyte) gives me bit strange feeling about WTH is going on with Asus...