Ryzen 2700x OC air cooling and a scratched heatsync base.

[TRIPTEX_CAN]

I think if I aim for the depth of the machined grooves in the pipes I can end up with a pretty good finish because they're consistent at their depth.

 

[ArbitraryAffection]

when you take off the cooler, what does the paste look like? is it evenly distributed? I've had issues with some coolers mounting on AM4 where they are not putting enough pressure on the CPU or the contact isn't great because of the mounting bracket. Also Prime95 is like furmark for CPUs, it gets HOT.

In my 100% honest opinion I would say you are best leaving the 2700X at stock and letting PBO and XFR2 do their job. With that cooler it will run 4.2 all core in most workloads, only dropping below in power hog scenarios like P95.

 

[TRIPTEX_CAN]

@ArbitraryAffection yes, the contact is quite good I alsochecked the flatness of the IHS and the block and they're both satisfactory.

The more I play around with the chip the more I lean towards that suggestion. I don't have my heart set on a 24/7 overclock I think I'm just trying to determine whether I'm getting the cooling I feel like I paid for.

 

[lexluthermiester]

I think I'm just trying to determine whether I'm getting the cooling I feel like I paid for.

In that case, you very likely are.

 

[phanbuey]

ur fine... the paste will fill in the gaps... ive had those style coolers along with machined coolers, and they perform very similarly.

 

[TRIPTEX_CAN]

I dont know how much I like the XFR clocks and voltages though. This HWinfo sreenclip is from running 3dmark Time Spy.

While loading the tests the core boost almost to 4.5Ghz but while the CPU test is running they clock down to 3.9Ghz. I know its a benchmark and not a real game but I havent seen the CPU offer more than 4.1 in any game I've tested so far and that 4.1 is short lived.

 

[phanbuey]

I dont know how much I like the XFR clocks and voltages though. This HWinfo sreenclip is from running 3dmark Time Spy.

While loading the tests the core boost almost to 4.5Ghz but while the CPU test is running they clock down to 3.9Ghz. I know its a benchmark and not a real game but I havent seen the CPU offer more than 4.1 in any game I've tested so far and that 4.1 is short lived.

View attachment 111027

can you up the power limits?

At 68C you're prolly just running out of boost.

You can try raising pbo, and then LOWERING the voltage offset (makes it boost longer) as long as the system is stable.

but stock boost is like 4Ghz for all cores, so it looks good... 4.5ghz is an insane boost.

 

[TRIPTEX_CAN]

I thouht the 4.5 did look promising.

pbo? I know where the powerlimit setting is in my bios. I can push it to 120% (its 100% now).

For voltage offset... something like "-.012" to start?

 

[eidairaman1]

Be careful lapping HDT coolers. The pipes are thin and lapping them only makes them thinner... potentially not good things could happen.

Leaks

 

[phanbuey]

I thouht the 4.5 did look promising.

pbo? I know where the powerlimit setting is in my bios. I can push it to 120% (its 100% now).

For voltage offset... something like "-.012" to start?

Yeah give it more juice by going to 120% and slightly less volts... you should see a higher consistent boost. And then just keep lowering the volts bit .02v by bit until you get unstable then bump back up 2 notches.

But just giving it more power will give you better scores.

And ram... if you can OC the ram you will see your FPS really climb.

 

[Final_Fighter]

the best way to check the cooler is with a straight edge. if you dont have a straight edge, pull the proc and put it against the base of the cooler and see if it rocks back and forth or if there is any massive gaps that light passes threw. you may be surprised at what you find.

 

[TRIPTEX_CAN]

Yeah give it more juice by going to 120% and slightly less volts... you should see a higher consistent boost. And then just keep lowering the volts bit .02v by bit until you get unstable then bump back up 2 notches.

But just giving it more power will give you better scores.

And ram... if you can OC the ram you will see your FPS really climb.

Yeah, I put it to 120% and then 130%. The sustained boost didnt improve but the actual peak boost decreased. Still drops to 3.9Ghz in the CPU test. Even running Unigine Valley the it wont hold more than 4075Mhz consistently. Boosted to 4.35 to load the program but that didnt hold.

My ram is running 3466mhz at a sloppy cas16.

 

[phanbuey]

Yeah, I put it to 120% and then 130%. The sustained boost didnt improve but the actual peak boost decreased. Still drops to 3.9Ghz in the CPU test. Even running Unigine Valley the it wont hold more than 4075Mhz consistently. Boosted to 4.35 to load the program but that didnt hold.

My ram is running 3466mhz at a sloppy cas16.

But what about your scores? did they change???

That's really odd... try lowering pbo then? might have the opposite effect? Did you undervolt as well?

 

[TRIPTEX_CAN]

Time Spy total score is about 100pts higher with XFR than with a manual OC to 4.1Ghz 1.31v. Pretty small margin really.

 

[phanbuey]

try superpi or cinebench single core - that will give you a decent single core sampling; timespy is fully threaded so the fact that XFR is faster than a 4.1Ghz straight OC is pretty damn good.

 

[Athlonite]

Remember AS5 takes time to cure and will produce it's best results after about 4~5 days of use and thermal cycling as for the scratch it's probably no worse for perf than the shocking factory machining marks already there

 

[TRIPTEX_CAN]

I set the clocks and voltage to automatic but left the power at 110%.

Singlecore cinebench runs are akin to watching paint dry.
Xfr boosts to 4.34Ghzish but bounces the load from core to core as they are boosted between 2700 and whatever they fell like. I havent seen 4.5 again though. Im not sure what conditions that took to make possible.

Cinebench single core
OC 4.1 = 170cb
Xfr 4.34ish = 175cb

Interesting though, fully threaded cinebench is another story where at straight OC 4.2 it will break 1900cb but xfr only boosts to 3.9 and scores about 1760cb.

 

[GlacierNine]

Remember AS5 takes time to cure and will produce it's best results after about 4~5 days of use and thermal cycling as for the scratch it's probably no worse for perf than the shocking factory machining marks already there

I think everyone is overdoing it on the surface finish, frankly. It's not the greatest ever, sure, but we've known for a very long time that mirror finishes don't necessarily produce the best thermal results. Look at noctua - If you grab any of their coolers the baseplates have a fine texture to them.

In fact, Noctua even has a page on their website explaining why - https://noctua.at/en/why-doesnt-the-base-of-noctua-coolers-have-a-polished-mirror-like-finish

Noctua does this on purpose and produces, bar none, the highest performing air coolers on the market. A mirror shine is not required. In fact, I daresay the surface finish on a noctua is a little *rougher* than on those heatpipes - it's enough you can feel it with a fingernail if you drag it over the surface, like on those old lenticular pictures you used to get on cereal box toys, just a bit finer.

The difference is that Noctua achieve it through machining a flat base whereas these heatpipes are clearly flatted with a sandpaper-like abrasive, so the noctua version looks more even/consistent.

 

[Athlonite]

Yeah Na my Thermolab Baram would like to disagree with that poorly written statement from Noctua... Two really smooth surfaces where very little thermal paste is required is much better than one good surface and one shit one that requires alot more thermal compound

 

[MrGenius]

Why doesn't the base of Noctua coolers have a polished, mirror like finish?

As most of today's PC enthusiasts use high-viscosity thermal compounds, the contact surface of Noctua coolers is optimised for use with this type of pastes. The micro-grooves at the contact surface ensure that high-viscosity thermal compounds are dispersed to a uniform thin layer across the whole contact area and that no air pockets remain between the cooler and the CPU. With a polished, mirror like surface, the risk of uneven dispersion is much higher. As too thick layers of thermal paste and air pockets drastically deteriorate heat transmission, the micro-grooves are is of vital importance to the overall cooling performance of Noctua coolers when used with today's high-viscosity thermal pastes.

Bullshit. All of it. Just a bunch pseudoscientific drivel meant to fool the naïve.

 

[lexluthermiester]

Bullshit. All of it. Just a bunch pseudoscientific drivel meant to fool the naïve.

Not really. It has to do with the microscopic particles in the TIM. None of them are perfectly square or have flat edges. They are of various shapes and align together in semi-regular formations. Perfectly flat surfaces on an IHS and heatsink can lead to less effective cooling. Irregular surfaces are actually more effective. Physics for the win.

 

[GlacierNine]

Bullshit. All of it. Just a bunch pseudoscientific drivel meant to fool the naïve.

I think the performance of Noctua's products speaks for itself here, yes? If mirror finishes were simply better, then it would be trivial for the competition to produce heatsinks that outperform Noctua's offerings.

And yet, at every size, 120mm, 140mm, downdraft, and low profile, Noctua produces product that is either within spitting distance of the top performers, or *literally is* the best performing product available.

That doesn't happen because Noctua doesn't know what they're doing. It doesn't happen because their products are based on "psuedoscience".

The fact their products work so well is the reason their products continue to sell well, despite their lack of aesthetic flair.

Yeah Na my Thermolab Baram would like to disagree with that poorly written statement from Noctua... Two really smooth surfaces where very little thermal paste is required is much better than one good surface and one shit one that requires alot more thermal compound

That would of course be why your 9 year old heatsink with an 8 microinch surface roughness according to Frostytech, did not cool more effectively than a Tuniq Tower 120 which had a surface roughness of 32 microinches, also when tested by Frostytech: http://www.frostytech.com/articleview.cfm?articleid=2198&page=2

You are also welcome to go compare the surface roughness of all the other heatsinks that outperform it: http://www.frostytech.com/articleview.cfm?articleid=2736&page=5

Finally, and though this is incredibly childish of me, I'm not really sure it's wise of you to criticise Noctua's writing style given your own writing. "Yeah Na" isn't valid grammatically, and even if it were, the "na" shouldn't be capitalised. In addition, "really" should be replaced with "very" in the next sentence, you should consider using an oxford comma after the word "required", "a lot" is two words not one, and you failed to end your post with a period or other appropriate punctuation.

Glass houses, and all that.

 

[Athlonite]

you might want to check again where they actually test the Thermlab Baram and the Baram 2010 vs that Tuniq Tower 120 ( which I also have and was using before buying the Baram)
http://www.frostytech.com/articleview.cfm?articleid=2544&page=4

Tuniq Tower 120 high 16.9c 51.6db vs Thermolab BARAM 2010 (80CFM) high 14.0c 51.3db

Baram (2010) wins by -2.9 degrees C

and In my case the Baram does even better in a two fan push/pull configuration

As for the Yeah Na it's a kiwi saying so suck it up butter cup and nobody likes a Grammar Nazi so you can now consider yourself blocked and ignored from now on

 

[GlacierNine]

you might want to check again where they actually test the Thermlab Baram and the Baram 2010 vs that Tuniq Tower 120 ( which I also have and was using before buying the Baram)
http://www.frostytech.com/articleview.cfm?articleid=2544&page=4

Tuniq Tower 120 high 16.9c 51.6db vs Thermolab BARAM 2010 (80CFM) high 14.0c 51.3db

Baram (2010) wins by -2.9 degrees C

and In my case the Baram does even better in a two fan push/pull configuration

You're quite right, I linked the wrong review. The exposed heatpipe version beats the Baram - http://www.frostytech.com/articleview.cfm?articleid=2480&page=4

As does the Spire Thermax II, which had a surface roughness of ~48 microinches. http://www.frostytech.com/articleview.cfm?articleID=2348

 

[John Naylor]

Hi all. I've been busting my ass installing and reinstalling my CPU cooler hoping to get better temps than I have been. Maybe my expectations are too high.

I replaced the stock cooler a) because it's really loud when it's effective and b) because I figured I could get more from a massive heatsync. Enter the Cooler Master MA620P. It's huge, It has good fans, and 6 heat pipes vs the 4 pipes on the stock cooler. Not to mention its almost 4x the surface area of the stock cooler.

In the end the temps Im getting arent as good as I expected...

As I remounted the cooler for the 4th time I decided to look closer. I did notice scratch/deformity in the base of the MA620p and I wondering if it's been part of the problem.

I should also mention that I've tried a few different methods of applying the TIM and Ive tried the included CM Master Gel and AS5.

What temps should I expect when running the CPUz stress test or Prime95 Small FFTs?

1. And what temps are those ?

2. Did you read any comparative reviews before buying ?

3. What TIM ?

4. Are you using old P95 w/o AVX or new P95 w/ AVX (the later is scary). Using P95 to test PCs is like setting up your home security for demons and zombies.... it does not represent anything close to real wporld conditions. On top of that, taking P95 stable OC running single tasks on each core doesn't prove real world stability ... I have had 2 4 hour stable P95 OCs fail under multitasking test like RoG Real Bench which is a suite of 4 of the most demanding Real World applications, and runs them individually and then at the same time.

http://dlcdnmkt.asus.com/rog/RealBench_v2.56.zip

Noctua does this on purpose and produces, bar none, the highest performing air coolers on the market. A mirror shine is not required.

the best way to check the cooler is with a straight edge. if you dont have a straight edge, pull the proc and put it against the base of the cooler and see if it rocks back and forth or if there is any massive gaps that light passes threw. you may be surprised at what you find.

1. Back in the day Noctua was the dominant air cooler... Phanteks topped the NH-D14 and "took the title". Then Noc grabbed it back with the D15 ... Scythe took it away again with the Mugen Max and now the Fuma... and for half the price. Its has lower temps and lower noise.

https://tpucdn.com/reviews/Scythe/Fuma/images/temp_oc_aida64.png
https://tpucdn.com/reviews/Scythe/Fuma/images/fan_noise_100.png

2. The IHS is not usually flat but concave and many heat sink manufacturers who produce a matching convex curbed heat sink to math void their warranty if you grind it flat. Does it make a difference,... seen some test that say yes, some that say no.... Id say about 65-35. If you are not worrying about what TIM ya chose, because its only 2-3C difference, you certainly have nothing to worry about with regard to lapping, concave or convex.[/QUOTE]