Heatsink disappointment

[Assimilator1]

aigomorla
Yea fair point ,& my results (data) seem to agree with that too ,though I am already running a quad.But perhaps because it is 'only' at 3GHz & 1.3v that it is not showing its (hopefully superior) cooling abilities over my Scythe Katana 2?.I'm hoping that with a higher clock speed (& thus heat load) it will be a significant improvement ,otherwise it will of been a waste of money! .Though atm I seem to be having some sort of contact/mounting problem with this cooler (see my previous post to Daverino).

My question in my 1st post to you was just to understand why you thought the Ult 90 was so bad ,though I was suprised by your comment I wasn't necasarily saying you were completely wrong
(Btw I just realised I missed your post on the 29/12 ,my bad ,sorry)

I think I see what your saying now ,that the Ult 90 isn't 'fantastic' because under moderate load (C2D) it doesn't really show its abilities ,but under higher loads (C2Q) it performs well enough ,have I got ya right now? (btw I still think saying its 'crap' is somewhat harsh)
Odd behaviour from a HS though ,usually regardless of load a 'better' HS will perform better under all loads regardless.I assume this is a factor of having so many heat pipes relative to the HS size?

JEDIYoda

You are as arrogant as you are dense yourself! :roll:,you seemed to of totally failed to understand my points.
You claim that different sites post different results (rankings, not actual CPU temp figures I'm talking about) yet you still haven't posted any links to them to backup your claims! ,which is what I was asking for in the 1st place!.

So for the most part (excluding small variances from thermal paste application & for the few heatsinks that do not have spring loaded screws) ,a HS which is best under high thermal load will still be the best in any case regardless of how many fans & or +/- pressure, with the same thermal load. You have not proved me wrong on that yet.

I proved my point!!
Total BS you've proved next nothing! (aside from 2 minor fitting points) ,where's your data? your word alone isn't enough!

Instead of asking me to do your work why don`t you do your own and try to understand that there is NO such thing as the definitve best heat sink for all situations!!

I'm not asking you to do my work ,I'm asking you to prove your point which you've almost totally failed to do so far! ,oh holy JEDIYodas word itself is not enough & so far is nothing more than heresay.
You make claims (which may or may not be right) & then either can't or can't be bothered to back them up ,utterly pointless & as it proves nothing utterly useless.

Oh & I take back the apology ,my 1st reaction about you was right afterall ,(you're gobby & arrogant).I give up on you & I have nothing further to say to you.

 

[Assimilator1]

Getting back to the Ultima 90, I'll copy a bit I said about mine to Daverino above incase it was missed in the arguement.

I did notice one problem though ,I had a huge difference in temps between 2 cores ,upto 12C! ,where as with my Katana 2 it was about 3-4C at most. (under load running F@H SMP ,measured with Core temp).
Now I had 2 thoughts about this ,1. I screwed up! 2. I noticed that the clamp didn't clamp down very hard (yes the screws were tight) & the HS seemed far too easy to twist!.
Anyway ,I took the HSF off (& saw that 2 opposing corners of the CPU hadn't touched the HS!!),took the springs off the screws & put a washer on each screw ,put it all back together (without re-doing the paste! ,yea I know bad me!).That alone cut the difference down to 7C (still not good but better).I'm going to try putting new paste on now & then afterwards, turning the washers around (their doomed so it'll applie slightly more pressure).
I'll let you know how it goes!.

Turns out I had the washers in the position to give the mosting spacing anyway.I'm wondering whether to add 2 more (thinner) washers to the screws to increase clamping pressure to see if that further evens out & drops temps ,however I'm worried about overdoing it.
I am convinced though ,on my setup at least, that the standard clamping pressure wasn't enough ,actually it was pathetic ,it took very little effort with 2 fingers & thumb to twist the HS!:Q ,I'm not bothered by it being able to twist but I am by how little force it took!.
How stiff is it in other peoples PCs?

Maybe the other part of the problem is the base not being flat?, Vandraad2 in the Ultra-120 extreme lapped thread measured the distortion in the base with feeler gauges ,which I have several sets too (being a mechanic) & I could check for distortion.
I have a feeling I may need to lap this HS ,which is a PITA I could do without especially as I've never done it! (know any good lapping guides?)

Daverino & other Ultima 90 owners

What are your CPU temps under load? (with Prime95 ,measured running Core temp) what's the difference between coolest & warmest core?

 

[aigomorla]

Sinks work better as the delta from ambient to sink temp is greater.

The hotter your sink gets at a colder ambient, the better it will cool.

Now if your sink is only getting slightly warm to the touch, and your ambients arent allowing it to radiate that heat properly, then yeah, performance will suck.

So theres 2 ways to get better performance on the sink.

1. increase the heat load, or reduce ambients significantly to see a benifit
2. change medium. [IE water or a different sink]

Otherwise your sitting in the gray level where all sinks will perform the same.

 

[DerwenArtos12]

I haven't seen the same convex issues with the ultima 90's as I have tieh the ultra120x's but, it is quite possible. I agree that the mounting option is less than idea but, that is why I own a machine shop. Not a single CNC machine in sight and yet I've been able to re-create all the missing parts from my grandfathers hand-made 1/10th scale locomotive. I happen to think a lot of the sence of pride of making something is removed when it's all done by a computer program. Thats just the old-school in me leaking out though.

 

[Old Hippie]

yet I've been able to re-create all the missing parts from my grandfathers hand-made 1/10th scale locomotive.

Oh Wow! You wouldn't happen to have any pics of that, would ya? That sounds very interesting! :thumbsup:

 

[Assimilator1]

DerwenArtos12

Sounds good ,I wanta see loco pics too
Though I take it I don't have to tell you that CNCs are good for keeping tolernaces down & elimating the variable 'human' factor which is better for mass production

Re Ultima 90s ,have you had some & checked their bases?


Would like some input from other Ultima 90 owners too:thumbsup:

 

[Elcs]

Originally posted by: Assimilator1

Re Ultima 90s ,have you had some & checked their bases?


Would like some input from other Ultima 90 owners too:thumbsup:

Unfortunately, i hopped straight into lapping my Ultima 90 so I didnt really check.

During lapping, it seemed to go down at roughly the same pace do Im assuming there wasnt much unevenness on the surface.

 

[RussianSensation]

As has been mentioned different reviews produce different results. No one should base their product purchase on 1 single review. IMO the reasons to spend $ for an aftermarket heatsink include:

1) to lower noise levels
2) to increase overclocking through lower temperatures
3) if you are seeking aesthetics for a window case
4) to lower overall system case temperatures if such cause instability of other components

To start with, OP's temperatures were great and there was no need to replace the SI-120 heatsink in the first place. For some reason people start to freak out if their cpu runs at say 60*C. How many of us keep our systems for 5-10 years? Intel certifies their CPUs to run at certain temps (i.e. 69-71*C, etc). So maybe it's important to reevaluate why is it that we try to lower our temperatures prior to making the upgrade because it's not like mid-40s cpu temps are unacceptable...

In my view the problem stems not necessarily from an inferior heatsink purchase but that there was hardly any improvement to be made with that setup. It's not as if lowering the temps from 43*C to 35*C would have changed 1 thing whatsoever. Obviously I am no way implying how OP should spend his $ but it's something to consider to avoid disappointment (recall everyone chasing DDR1 2-2-2-5 memory for barely any performance increase just because Joe, James and John have it).

I would have approached your purchase decision process like this:

Outcome 1: CPU Temp falls from 43*C to 35*C (Happy? Yes, since lowered your temperatures as desired. Performance/stability improved? No. Result: Decision to purchase justified? Not really since no effect on performance <> Yes if lower temperatures themselves make you sleep better at night)

Outcome 2: CPU Temps rise (Result: Disappointed, waste of $)

OP?

 

[Insidious]

Kind of wierd thread here...

My observations, based on the two Ultima-90s I purchased from (I didn't know anyone else used these guys... ) FrozenCPU and installed on Q6600s:

These heat sinks were a VAST improvement over the intel stock heat sinks. I participate in distributed computing and have processors (yes, all 4 of them) at 100% load 24/7.

I see the highest core temperature (measured with CoreTemp.... please don't waste my time with inaccurate speed fan readings) running at ~62C.

To the poster (I really didn't pay attention to... or care about specific names) that states (with, IMO unqualified authority) that the Ultima-90 is "crap"... I say you need to learn how to install a heat sink properly or get a better case... that assesment is totally bogus (to be nice about it)

I don't like the mounting scheme for the ultima-90 though. I had to install a #8 flat washer on each spring screw (between spring base and mounting piece) to have the pressure that I feel (ymmv) is necessary for a good mount that doesn't swivel and displays minimized temperature spreads between cores (~3 to 4C in my case, loaded)

-Sid

I think some of you need to spend a lot less time chest thumping and more time verifying your claims. If you can't do that.... leave the science to some who can!

 

[Assimilator1]

Hi Sid
The guy who said it was 'crap' did clarify & somewhat 'adjust' his position in a latter post in this long(ish) thread.
Btw Speedfan can be calibrated to show correct temps using coretemp as the base (though it seems to only use 1 sensor AFAI can tell).

RussianSensation
Interesting points but not what I was asking about ,maybe I should of started a new thread?....

Anyone else got data on the spread of loaded CPU temps with the Ultima 90?

Elcs
Well interesting to know that at least the base on your HSF was flat.
Do you find the HS easy to twist?
What's the spread of CPU temps on your rig? (measured with coretemp & all cores @ 100%)

 

[Insidious]

Good Point AS1

Sorry for the rant.... exaggerations that stray away from fact get me going sometimes

I did forget to mention that I went with pretty wimpy fan on mine (92mm panaflo M1)

These results with such a mild fan is what makes me feel like this HSF is up to the task of effective air cooling that I have yet to see surpassed in any but the most extreme situations. (I bought a Q6600... it's not going to see 3.9GHz no matter what I use)

other pertinent info:

my clock is 3.5GHz and my Vcore is 1.475 for the temps I mentioned in my first post. This is in an Antec Sonatta III case which is not the best ventilated and made me even more impressed with the Ultima90 results I have in two 'identical' PC buiilds.

-Sid

 

[Elcs]

Originally posted by: Assimilator1

Elcs
Well interesting to know that at least the base on your HSF was flat.
Do you find the HS easy to twist?
What's the spread of CPU temps on your rig? (measured with coretemp & all cores @ 100%)

Not easy, it requires some force.... more torsion force than would ever be placed on the heatsink in any scenario other than knocking it when doing internal PC surgery in my opinion.

AMD X2 6000+ AM2 socket @ stock speed, 1.35V (undervolted from 1.4V) I get a 0-4C spread, guesstimated around 2C average spread. Load was placed on my system using Large FFT's in Prime using the auto-sensing core edition. Load temps were approx 48-53 over the timeframe, ambients 23-25C.

I am using a 120mm Scythe Slipstream on a Sunbeam Fan Controller, dialled to the lowest voltage the fan will start at. I am aiming for silence rather than overclocking power.

Please view my topic should you require any more information. It is currently floating around on the front page.

 

[DrMrLordX]

Sorry I went MIA in this thread Assimilator1. Most of the posts in this thread about insufficient pressure from the retail mounting system reflect complaints I've heard made about the Ultima-90 (mostly in the thread I linked above). Mods involving washers have done the trick in most instances.

 

[Assimilator1]

Hi Sid
Np

I'm currently getting a loaded temp spread of upto 8C ,I'm probably going to try to put more pressure on using more washers to see if that helps.I'd be happy with a spread of 4C.If I can get that then maybe I can finally see what this Q6600 can really do

Elcs
Interesting that you don't seem to have the mounting problem ,but lucky for you
I originally could fairly easily twist the HS with just 2 fingers & thumb!
Good undervolt btw:thumbsup:

DrMrLordX
Thanks & np
I wonder what the variances are that mean some people have the problem & others don't? , mbrd thickness maybe?? (though if that were the case I would think it'd effect all HSFs).
Maybe manufacturing differences in the springs? or clamping screws? brkts? lol could be anything!

 

[Insidious]

The motherboard brackets are going to have sjome vertical freedom no matter what brand you are using. (TR says this is to help distribute stresses)

I am guessing the lack of problem for some and presence of a problem for others is really more a definition of what "problem" means.

I know that while I complained about a 'loose' mount, it is important to note that while yes, I could twist the HS "easily" (<-- again, we all probably define "easy" differently) I still had a very effective heat sink before I added the washers. without CoreTemp to show me individual core temperatures, I would have no idea that my surface contact was uneven.

IMPORTANT NOTE: while I did force my individual core temperatures to be more in line with each other, and I did see a small overall temperature decrease after tightening my mount with a washer on each side I did not get additional OC headroom. Tried to go 45MHz higher after mod..... no joy. So what did I really gain with my modification... nothing but a little bit more of a 'warm n fuzzy'.

So did I really fix a 'problem'? Did I even have a 'problem' to start with?

Very subjective questions.

-Sid

(I did lower my average core temp. enough (by tightening the mount) that I may try raising Vcore another step or two... so the jury is still out on if it truly bought me anything or not. I suspect any additional speed gain will be very small and probably will only demonstrate the point of diminishing returns. )