VRM cooling solutions

[KeltonDSMer]

Hey LTT, I was screwing around today with the new set up and it imediately became clear I need to cool this socket a bit better somehow.

 

-8320

-M5A99fx Pro R 2.0

-CM 212

-NZXT source 210 with only 1 exhust fan

-1.225vcore, medium llc (drops to 1.164v in P95)

-1.2525v NB

-4.0ghz from 20xmult 200NB

 

With the case closed, core temps max at 50-52C but the socket is way up there, hits 68C-70C and stays there after about an hour of P95 (seems to take longer to get that hot in blend for some reason). The board does not seem to throttle power delivery, but that is at the upper limit of the socket, correct?

 

Two suprising things to me:

1-The fact that 4ghz is P95 stable (1.5hr small ftt and blend, will test further) at only 1.164Vcore reported in cpuz and HWmonitor

2-The fact that 1.164Vcore makes so much goddam heat on this board with the case closed

 

Should I just get some more case fans? What is the max "socket" temp for the ASUS 970/990 boards? If the board can take 80C or something I'm golden, but does anyone know for sure? Will the board throttle bofre it hurts itself with heat?

 

Another side question for fan control: I am using fan expert program that came with the Asus board, but the fan control is dependent on the socket temp and that pisses me off because the CPU fan doesn't even point at the socket. There is no way to profile a nice curve for the CPU temp if is looking at the socket temp. Can I change this somehow, or is there a free fan program that allows me to change the variables for the fan curves?

 

Oh side note: I was a little worried for a couple days about updating the bios even though I knew I should. I decided not to OC untill I did and then realized ASUS disk has a program that pretty much does everything for you from the OS. Literally takes 5 minutes from Windows, super quick and easy.

 

Double side note: core #6 can go to hell. He is the little bastard that fails first everytime when I was zeroing in on minimizing vcore and llc.

Actually I really like him I'm glad he is there... don't die on me :wub:  (typing while running prime)

 

Thanks.

[LogicalDrm]

Only one fan in whole case? How do you expect components to survive without air? Add two fan as front intakes. Maybe one to top intake or exhaust.

[deathjester]

Only one fan in whole case? How do you expect components to survive without air? Add two fan as front intakes. Maybe one to top intake or exhaust.

 

^ Do that. 

 

If you are looking for the cheapest option this will work. 4 120mm CM fans for 12 bucks. No they are not Noctua, and silent, but they will give you a lot nicer temps in that case. You simply need more cool air being sucked in. 

 

http://www.newegg.com/Product/Product.aspx?Item=N82E16835103052

[Bagmup]

@KeltonDSMer, 72 degrees celsius us is the maximum socket temperature for that board, I had the same one and ended up using the fan from the stock cooler, ghetto mounted on the back of the socket to control the temps.

8350 @ 4.8Ghz with 1.4v

[KeltonDSMer]

@LoGiCalDrm , I wanted to get a baseline for temps and such before going out and buying things I might not need. I did not want to be the guy that has fans worth more than the CPU or something (not like that is bad), just keeping costs down is all. I figured some case fans were going to be needed, thanks for the suggestion.

 

@deathjester , thanks for the link, that is a sweet deal! I think I'll have to hold off though, I leave next Thursday for the weekend and don't want to miss a delivery. Plus, I think I am burned out from filling out MIR bull crap. I really dislike having to cut out the UPC of the original boxes, what if I run into a problem with something and need to RMA? I guess for fans though, MIR is not an issue. I would never RMA a $4 fan and could care less if I need to cut up the boxes.

 

@Bagmup , thanks for confirming 72C max temp for the board! I totally forgot about the dinky little AMD cooler fan, I'm definitely going to mount that sucker behind the tray! That is a perfect option, its free, and directly addresses the only issue I have run into with lack of airflow in the case.

 

Thanks guys, much appreciated!

[KeltonDSMer]

@Bagmup , thanks so much for mentioning the stock cooler fan! I tie wired that little fcker behind the NB and am "blown away" (silly pun fun) how much it actually helps. I was idling yesterday at 40-45C on the socket depending on how hot it was in the room, and hot dayum it is now 30-35C depending on the new fan's speed. The chip is also much cooler too, it is now reading lower than ambient (lol impossible) at idle which did not happen yesterday without the new tray fan.

 

 

I fired up P95 and in the time it took to hit 65C yesterday, this new setup barely cracked 50C on the socket! That little fan is super loud above 80% or so, but I could not be more happy with the solution you reccomended. Now my bitching about the fan control software only reading the socket temp is moot, because I now actually have a fan that directly affects those temps...super helpful to get a nice profile to balance cooling/noise!

 

I was a little bummed about how little voltage I was able to run with the case closed and just the one exhaust fan, but this gives me so much headroom!

 

Thanks again, what a great solution that cost me 5 minutes and $0. Hell yeah!

 

Edit: New temps are of course with the case closed again, I could care less how cool I can get things without the side panels on. I ever so slightly tilted the new fan off the plane of the side panel in order to allow more air to get in, it is a pretty tight fit behind the panel and the thing needs to breathe!

 

Double edit: Oh and Bagmup, I prefer the term "thoughtful custom install" over "ghetto mounted"...who am I kidding I turn on my coummuter car with a marine switch hanging out the front bumper

[LogicalDrm]

@KeltonDSMer, I guess thats smallest mistake to make in terms of cooling. I'm really amazed that exhaust for chipset behind mobo works as it doesn't have anywhere to push hot ait.

[Bagmup]

@KeltonDSMer, I guess thats smallest mistake to make in terms of cooling. I'm really amazed that exhaust for chipset behind mobo works as it doesn't have anywhere to push hot ait.

Cases aren't air tight...

[LogicalDrm]

Cases aren't air tight...

 

No, some even have holes and mesh for air. That one hasn't any on left side. Thats why I'm so impressed. I wonder if there's nice, hot spot on sidepanel over that fan.

[KeltonDSMer]

No, some even have holes and mesh for air. That one hasn't any on left side. Thats why I'm so impressed. I wonder if there's nice, hot spot on sidepanel over that fan.

 

I was shocked how effective the little fan is behind the tray for that exact reason, airflow is pitiful back there.

 

I did, howerver, angle the fan as much as possible off the plane of the panel to give as big of a "gap" as possible for intake. The right panel where the fan is does get hot lol. In this situation, I imagine the black case to be slightly beneficial considering how well is radiates the heat off the panel. The "hot spot" is not much bigger than the fan so the panel must be expelling the heat pretty efficiently (steel conducts heat quite well so not having the whole panel be uniform temp is a good sign).

 

Fortunately, with the negative case pressure, there is natural flow up around the top of the mobo where the 8pin goes through. The gaps for the intake draw cool air behind the tray and it is expelled over the top of the tray and out the back.

 

With all panels closed and in a 30+C room, here are my temps after just under 4 hours of P95 blend, 4.0ghz at 1.2v under load:

 

 

That is IMO fine and dandy considering my very much lacking airflow arrangement. I have been able to run up to 4.3ghz stable for over an hour with the case closed, but anything higher has needed the panel off to keep temps in check.

 

I am currently testing 4.5ghz at 1.32Vcore and almost 1.4v NB; it has been stable for 30 minutes in P95 blend so far but temps are as high as I'm comfortable with, even with the side panel off. I said I don't care what it will do with the panel off, but I figure I might as well get a stable voltage profile at as high of a clockspeed as possible, because only P95 heats up the system this much. If I get it stable enough at whatever speed/voltage with the side panel off, I'm sure I can bench with that profile with the case closed without coming close to having a temp issue.

 

Thanks for the help guys, this is a lot of fun!

[Priller]

 

 

 

I am currently testing 4.5ghz at 1.32Vcore and almost 1.4v NB; it has been stable for 30 minutes in P95 blend so far but temps are as high as I'm comfortable with, even with the side panel off. I said I don't care what it will do with the panel off, but I figure I might as well get a stable voltage profile at as high of a clockspeed as possible, because only P95 heats up the system this much. If I get it stable enough at whatever speed/voltage with the side panel off, I'm sure I can bench with that profile with the case closed without coming close to having a temp issue.

 

Thanks for the help guys, this is a lot of fun!

 

Use 1.35 for Vcore with Very High LLC and back down the CPU NB to 1.33 or under. You will find the sweet spot to be 1.25-1.30 volts with ram in the 1866 range.

 

52C is fine as well. You have until 62C until you need to start watching temps heavily.

[KeltonDSMer]

Use 1.35 for Vcore with Very High LLC and back down the CPU NB to 1.33 or under. You will find the sweet spot to be 1.25-1.30 volts with ram in the 1866 range.

 

52C is fine as well. You have until 62C until you need to start watching temps heavily.

 

Yeah llc has been very useful to me, I am also at very high on the cpu which seems to just about cancel out vdroop under load (stays very close to voltage set in bios). This seems handy in keeping idle and low load temps down by not having to set such a high Vcore. Thanks for the tip, I'll see if I can back off the NBv a bit and still maintain stability. I was running out of thermal headroom on the chip but had quite a bit left on the socket, so I thought bumping NBv over Vcore for the 4.5ghz push might allow for stability without heating the chip too much directly (could be way off in how those temps/voltages are dependent on each other but I was just playing around).

 

My 4.5ghz profile had no errors when I stopped Blend after 1 hour. The base NB voltage was bumped up that high because I used the bus speed partially to hit that frequency (21.5x210) and the RAM is running 1680mhz at 9-9-9-24 timings at that point. Are you changing the bus to hit those crazy speeds or are you using just the multiplier?

 

I think 4.5ghz is all I am going to see with this cooler, at least for being able to run p95 for any significant period of time. I might be able to see 4.6/4.7, but only for those short CB runs and not anything sustainable.

 

I just built my first PC exactly a week ago and am having a blast! This is the same fun I experience when goofing around with spark/fueling on an engine's tune, but way less expensive and time consuming if I screw something up It is funny to me how much optimization there was to be had compared to the stock 8320 profile on this board. 4Ghz 24/7 at 1.2Vcore runs way cooler at idle and under load than the factory setup at 3.5Ghz and auto voltages. Just for kicks I let the board do its automatic OC and it stopped at 4.16Ghz (all from bus increase) and had the RAM running at 15xxmhz 11-11-11-28 timings and the voltage was so high I couldn't get through the first 6-7minute blend test without temps being an issue!

[Priller]

Yeah llc has been very useful to me, I am also at very high on the cpu which seems to just about cancel out vdroop under load (stays very close to voltage set in bios). This seems handy in keeping idle and low load temps down by not having to set such a high Vcore. Thanks for the tip, I'll see if I can back off the NBv a bit and still maintain stability. I was running out of thermal headroom on the chip but had quite a bit left on the socket, so I thought bumping NBv over Vcore for the 4.5ghz push might allow for stability without heating the chip too much directly (could be way off in how those temps/voltages are dependent on each other but I was just playing around).

 

My 4.5ghz profile had no errors when I stopped Blend after 1 hour. The base NB voltage was bumped up that high because I used the bus speed partially to hit that frequency (21.5x210) and the RAM is running 1680mhz at 9-9-9-24 timings at that point. Are you changing the bus to hit those crazy speeds or are you using just the multiplier?

 

I think 4.5ghz is all I am going to see with this cooler, at least for being able to run p95 for any significant period of time. I might be able to see 4.6/4.7, but only for those short CB runs and not anything sustainable.

 

I just built my first PC exactly a week ago and am having a blast! This is the same fun I experience when goofing around with spark/fueling on an engine's tune, but way less expensive and time consuming if I screw something up It is funny to me how much optimization there was to be had compared to the stock 8320 profile on this board. 4Ghz 24/7 at 1.2Vcore runs way cooler at idle and under load than the factory setup at 3.5Ghz and auto voltages. Just for kicks I let the board do its automatic OC and it stopped at 4.16Ghz (all from bus increase) and had the RAM running at 15xxmhz 11-11-11-28 timings and the voltage was so high I couldn't get through the first 6-7minute blend test without temps being an issue!

Both and it depends on the CPU and ram. You have to feel out the CPU to see if it likes multi or FSB better. And if you hit a FSB wall are you actually hitting a CPU issue or ram with voltage, speeds, and latency. It all varies.

 

I would start with ram at 1600 (CPU NB at 1.25V) and work the multi alone all the way up until you cant go any further. Then start adding FSB to it . You will get further then just using one or the other.