Heat dissipation of nForce3/4 chipsets - any clues?

Cooling Processors quietly

Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee

Post Reply
Tzupy
*Lifetime Patron*
Posts: 1561
Joined: Wed Jan 12, 2005 10:47 am
Location: Bucharest, Romania

Heat dissipation of nForce3/4 chipsets - any clues?

Post by Tzupy » Mon Feb 07, 2005 7:45 am

I plan to get an Asus A8N-SLI motherboard and change the wicked chipset HSF with a NB47J. However, opinions are mixed on how effective would this solution be in cooling the chipset. So if anyone knows something precise about the heat dissipation of the nForce4 chip, please let me and other people know. I'm interested in overclocking the FSB from 200 to 250, and I would lower the HT to 3x. I believe the frequency at which the chipset runs depends on the HT, not directly on the FSB, so I should lower the heat dissipation - is this reasoning correct? Or should I go for a Thermalright NB1 chipset and (lower speed) fan?

GlassMan
Posts: 168
Joined: Thu Jul 01, 2004 4:55 am

Post by GlassMan » Mon Feb 07, 2005 9:40 am

Background Multiplier is FDT, what you are calling FSB is HTT(clock) . FDTxHTT(clock) =HTT bus speed. Technically FSB is cpu speed, the effective speed is HTT(clock)x CPU multiplier/mem divider which = memspeed.
Stock HTT bus speed is 200x5, so reducing to 3 may cool chip enough to live. But read the link to see what can happen with a large passive heat sink.
http://www.xbitlabs.com/articles/mainbo ... orce4.html
The new version VnF4 will have a fan.
The nF3 runs much cooler, the addition of pci-x has increased heat production
No doubt a quieter solution than the stock fan can be found, but this area needs your careful attention.

SometimesWarrior
Patron of SPCR
Posts: 700
Joined: Thu Mar 13, 2003 2:38 pm
Location: California, US
Contact:

Post by SometimesWarrior » Mon Feb 07, 2005 9:51 am

Re: GlassMan's Xbit link: check page 7 for the relevant comments.
At x5 HyperTransport multiplier the system remained stable till the clock generator frequency reached 237MHz. With 4x HyperTransport, we raised the frequency to 292MHz. When we dropped the HyperTransport multiplier to x3, we managed to increase the clock-generator frequency to 305MHz. After that we encountered a new problem, this time a fatal one. As I have mentioned above, the chipset on Chaintech VNF4 Ultra VE mainboard is cooled by a passive needle-shaped heatsink of a medium size, and our supposition that this heatsink wouldn’t handle the chipset during overclocking came true at 305MHz clock generator frequency. When the mainboard was tested for stability in this mode, its nForce4 Ultra chip got overheated and burned out beyond repair. So, basing on our tragic experience, we have to claim now that nForce4 series chipsets require active cooling, at least the two top models of the series do. You should keep this fact in mind when choosing a mainboard for overclocking. For the owners of Chaintech VNF4 Ultra VE we recommend to mount a fan on the chipset heatsink. Just in case.
I don't know if they tested the motherboard inside a case or not, but if they didn't, then a case with a front-mounted fan might have better luck. You can try other standard tricks, like sanding the base of the heatsink an re-mounting with a thin layer of TIM, but those are only incremental solutions. Xbit-labs is probably right: the passive heatsink is designed for stable operation only at stock or slightly-overclocked speeds.

Tzupy
*Lifetime Patron*
Posts: 1561
Joined: Wed Jan 12, 2005 10:47 am
Location: Bucharest, Romania

Post by Tzupy » Mon Feb 07, 2005 10:11 am

I am using the term FSB for historical reasons, I know it's not a real FSB. Maybe from now I'll call it CGF (for clock generator frequency), how many people are going to be confused? Anyway thank you for the link, it's quite interesting. It seems that this CGF matters, not the Hypertransport speed (which is CGF times HT multiplier) - because they ran it at 237x5 stable and it burned out at 305x3. I only want to get to 250x10 with a 2 GHz Winnie, XP-90 and DDR500 memory. The Zalman NB47J is a bit better than the heatsink on the Chaintech chipset, but I still am unsure if I should go for Thermalright NB1 instead...

Blappo
Posts: 85
Joined: Thu May 27, 2004 10:21 am
Location: Waterloo, ON

Post by Blappo » Mon Feb 07, 2005 10:23 am

I just recieved my DFI LanParty NF4 Ultra-D. I'm planning on swapping the stock HSF for the chipset with the NB47J. I'll let you know how it goes. Unfortunately I will have to wait for the rest of my components. (The damn Seasonic PSU is on backorder).

Josd
Posts: 7
Joined: Mon Feb 07, 2005 9:58 am

Post by Josd » Mon Feb 07, 2005 11:49 am

(I posted the same message on another thread as well)

I had the opportunity to play around with a few components , and do some comparison. I used different fans, different PSUs and different HDs.
These are my experiences:

The only noise I hear from the systems is the airflow. The Samsung SP1614S drives are indaudible. The Maxtor Diamondmax 10 is comparable with the Hitachi 250GB SATA drives. The only sound I hear is the head seek, but it is very acceptable (for me).

I did not notice any differnce in noise level between between a Seasonic Super Silencer 460W and the Enermax Noisetaker 470W. I dedided to use the Enermax because it has a a 24 pin mainboard power connector and because Seasonic is hard to get in The Netherlands.

The main reason I used the Asus A8N-SLI board is because it was the only NF4 board available when I started experimenting in the middle of December. Although I will not use SLI, this motherboard has everything I need and I think it's price is fair.

I have build two systems now with the following config:

Case: Antec SLK3700BQE with 2x Nexus 120mm case fans
Mainboard: Asus A8N-SLI (nForce4)
CPU: AMD64 3200+/3500+
CPU cooler: Thermalright XP-120 (XP-90)
Cooler fan: Nexus 120mm
Harddisk: 1-4 Hitachi 250GB SATA
PSU: Enermax Noisetaker 470W
Video: Asus X300 extreme (passive cooling) / Powercolor X800XL + Zalman VF700-AlCu

On both systems I replaced the noisy NB fan with a Zalman NB47J.

The first system with the 3200+, XP-120 and Asus X300 runs for 3 weeks now without problems. My room temperature is 21C. The CPU idle temperature is 33/34C and under load no more than 46C. The drives are 32C for the lower drive and 34 for the upper one.
The nF4 chip runs hot (estimate 50C) but does not overheat due to the airflow.
This system was overclocked to 2.6 Ghz without problems.

The other system has a 3500+, XP-90 and a PowerColor X800XL. This system has 4 Hitachi drives internally and 3 in a SATA backplane, which fits into 2 5.25" bays. The internal drives are 32C for the lower drive and 34 for the upper one. The CPU idle temp is 34C. The removable drives in the backplane run a little bit warmer (45-47C), but you can switch individual drives off, if they are not in use.
I replaced the stock VGA cooler with a Zalman VF700-AlCu. The heatsink touches the NB cooler, but the flexible VGA cooler fins bend enough to make it fit without any problems. A side effect is that the VGA cooler generates additional airflow for the NB cooler.

Both systems run without problems problems for some time now.

The CPU temperatures are readouts of the ASUS probe monitor. The disk temperatures are readouts of HDtune 2.1 SMART monitor.

Next week I will assemble 3 more systems with a similar config, one being my new rig. I am really looking forward to it.

GlassMan
Posts: 168
Joined: Thu Jul 01, 2004 4:55 am

Post by GlassMan » Mon Feb 07, 2005 11:50 am

No insult intended with the background info. I didn't think it would hurt any to clarify.
Don't assume that it was FDT that caused the failure.
From pg4 of the article,
The temperature of the chipset got as high as 60-70°C throughout the benchmarking process, which is hardly a normal operational mode for it.
Benchmarking was at stock speeds (200x5), Chaintech chipset hs(f)'s are known to be concave with a large dollop of tim to compensate. My Chaintech VnF3-250 with the same problems ran for months at 250x4 at a temp of 34C,although I did have some air blowing over it to keep it from approaching 40C. While re TIMing I damaged one of the pins and ran 240x3 with basically the same temps until I got some new pins, lapped for a couple of hours, and placed a fan on the passive HS (very low voltage).
My point is that the nF4 chipset makes a lot more heat. The nF4 chipset is usually placed clear of the vid card, so a quiet(er) solution can be found, as SW noted ensuring cool air over a passive heatsink may be all it needs.
The nF3 can be passively cooled and ignored, nF4 needs some thought and attention.

Edit: I see now that the chipset is not clear on the Asus, and josd has a workable solution already.

Tzupy
*Lifetime Patron*
Posts: 1561
Joined: Wed Jan 12, 2005 10:47 am
Location: Bucharest, Romania

Post by Tzupy » Mon Feb 07, 2005 12:59 pm

Thank you for the answers!
I will probably go for the NB47J and try to add airflow, since I also plan to use a passive 6600GT. Maybe I can still find a Zalman bracket fan...
Josd, I do envy you! It was expectable that the VF-700 helps cooling the NB. For the one with the passive graphic card, I think the temp would be higher than 50C. I am only going to use 2 HDD and a DVD, I suppose the heat also increases with the number of ports effectively in use.
After a lot of thinking on the nForce4 chipset cooling, I have a couple of concepts of my own, but I have no idea how to post pictures on this site. I also can't find anyone willing to manufacture such devices...

Post Reply