My Ninja has been defeated...

Sizzle · Post by **Sizzle** » Tue Apr 04, 2006 4:35 am

Been using a Ninja on my A64 3500+ for a while now. It certainly put good temps down for the CPU, but my SYS temp and PWM temp on my passive ABIT AN8 Ultra were high. So I dug out my old XP-120, had it lapped and swapped it in for the Ninja last night. Very happy with the results.

Before: Ninja with Papst 4412F/2GLLL @6v (all temps at idle)
CPU: 22 C
SYS: 36 C
PWM: 47C

After: XP120 with Scythe 21D @6.5v (all temps at idle)
CPU: 23 C
SYS: 28 C
PWM: 38 C

Plus I also replaced the PSU in the system from a Antec SP 500 2.0 with my Phantom 350 I got back from Antec on RMA.

diver · Post by **diver** » Tue Apr 04, 2006 4:47 am

What is your room temperature? Is it something like 20C? I won't see anything like that until Thanksgiving.

Sizzle · Post by **Sizzle** » Tue Apr 04, 2006 5:04 am

Ambient temp was 20 C. I see that from Mid-Sept to May.

Aris · Post by **Aris** » Tue Apr 04, 2006 5:56 am

why'd you switch from the papst fan to the Scythe fan?

were you havnt any problems with the papst?

EDIT: also, this is a common known issue with all tower heatsinks. since they blow across the motherboard and not down on it, alot of the motherboard components tend to run alot warmer than normal.

diver · Post by **diver** » Tue Apr 04, 2006 6:22 am

I am not so sure about the blow across vs blow down issue. In a recent SPCR article about an overclocked dual core Intel system the author found lower PWM temperatures with a Ninja than a SP-120.

mbetea · Post by **mbetea** » Tue Apr 04, 2006 6:28 am

What other case fans do you have? I have the same setup as far as mobo/heatsink goes, even under load sys temp gets to 35C. PWM idles around 36C-38C goes up to about 47-48C under load. The only time PWM will heat up a bit is if I run the Ninja passive which I did for a while. But found the Nexus running at 7v doesn't increase any noise so and helps lower PWM temps. This is in a p180 case with the top vent covered.

Sizzle · Post by **Sizzle** » Tue Apr 04, 2006 12:41 pm

I have a 21D at 6v for exhaust, and one for intake.

I replaced my Papst on my cpu as it seems to make more motor noise than the Scythe does.

I think in your case mbetea, you don't have the added heat from a passive PSU in that region as I do.

BillyBuerger · Post by **BillyBuerger** » Tue Apr 04, 2006 12:46 pm

diver wrote:I am not so sure about the blow across vs blow down issue. In a recent SPCR article about an overclocked dual core Intel system the author found lower PWM temperatures with a Ninja than a SP-120.

Check out the discussion about this issue. The OC'd Intel generated enough heat that the XP blew hot air down at the PWM causing them to run hotter. I don't expect you would see this problem with a cooler running setup like Sizzle is talking about.

diver · Post by **diver** » Tue Apr 04, 2006 1:24 pm

As far as the effect of the hot Intel CPU goes on blow down, that is a real possibility.

Anyway, I have a Ninja Plus with the stock fan running 700 rpm. Exhaust is via an Antec fan rated for 1600 rpm @ 12v, but running @ 5v. The PSU is a Smart Power 2.0 350 with the push-pull 80 mm fans. No intake fans at all. 7600GT on a ZM80D, with the draft going into the case, plus two hard drives all on an Abit KN8-SLI.

It is a toasty warm 27C in the room. At idle:
CPU: 30
System: 32
PWM: 37

My CPU is a 3800 X2 running 2400 mhz @ 1.35 v, more a or less the same thing as a stock 4600 X2, heat wise. So, how come I don't have the PWM problem in my hotter than H-E-double-hockey-sticks study?

I just wonder if there was some problem other than the type of cooler used.

halcyon · Post by **halcyon** » Tue Apr 11, 2006 12:47 am

I also got seriously raised PWM / chipset temps with a switch from XP-120 to Ninja plus.

I must say that I'm running fairly tight on space and making a good airflow from front to back (or vice versa) in my case is really really difficult.

For my situation, 120 is just better, because I have to use a fan at any rate (Opteron 180).

With Ninja plus and a decent 120mm fan on the Ninja and cooler air being brought from outside by anoter 120mm fan, I'm getting 80C on the Opteron 180 (not overclocked, def voltage) when both cores are loaded 100%.

Felger Carbon · Post by **Felger Carbon** » Tue Apr 11, 2006 4:01 am

halcyon wrote:With Ninja plus and a decent 120mm fan on the Ninja and cooler air being brought from outside by anoter 120mm fan, I'm getting 80C on the Opteron 180 (not overclocked, def voltage) when both cores are loaded 100%.

It would be of interest to us SPCR types if you let us know what temperatures the Thermalright HSF gave when using the same fan as the Ninja plus, running the same RPM. Thus comparing apples with apples, not kumquats with tangerines.

diver · Post by **diver** » Tue Apr 11, 2006 6:51 am

If you take a look at the Madshrimps Feb 06 CPU cooler test there are a bunch of results that show nearly equal performance for the SI-120, Ninja and Big Typhoon on CPU and PWM temps. This was both for a 3200 stock and hotly OC'ed at 1.7V. All used the same fan at the same RPM.

As I mentioned before, in the SPCR dual core Pentium article the XP-120 produced worse PWM temps than the Ninja.

Some of you may want to dismiss that because the Intel CPU runs hot, but so does an AMD @ 1.7V.

The evidence just does not support the notion that blow down CPU coolers will always produce lower PWM temperatures, YMMV. There is something else. Blow across coolers can be ducted so that the hot exhaust goes right out the case. It made a big difference in my system, dropping all 3 sensor temperatures, reducing PSU fan speeds considerably and eliminated crashes resulting from PSU overheating during multi hour Prime95 stability testing.

Sizzle · Post by **Sizzle** » Tue Apr 11, 2006 8:08 am

Take a look at the OC'd P4 article. The fan is mounted very low on the Ninja. In my setup, that was not possible. My fan could only be lowered on the hs to the top of my memory sticks. Leaving a good 3/4 inch between the buttom of the fan and the surface of the motherboard. That explains to me why my PWM temps were so high with the Ninja.

diver · Post by **diver** » Tue Apr 11, 2006 8:17 am

Sizzle-

All I am saying is that one should not generalize on this blow across/blow down issue. Each system is different and sometimes you just have to experiment. My data points show one thing, your system shows another. It could be the location of your fan, or something else. If you changed coolers and your temps are lower, that is progress.

dfrost · Post by **dfrost** » Tue Apr 11, 2006 6:15 pm

Try blowing up (away from the mobo) with an XP-120. Doing so with my P4 3.0C (no OC) Northwoods system gives the best balance between NB and PWM temps. Blowing down consistently increases PWM at load by 7C, and decreases NB temp by 7C (no change in CPU temp). In this system, the PWM bits are under the overhang part of the XP-120, while the NB is nestled between the heatpipe bends.

cmthomson · Post by **cmthomson** » Tue Apr 11, 2006 6:44 pm

Sizzle wrote:Take a look at the OC'd P4 [EDIT: P 830 D] article. The fan is mounted very low on the Ninja. In my setup, that was not possible. My fan could only be lowered on the hs to the top of my memory sticks. Leaving a good 3/4 inch between the buttom of the fan and the surface of the motherboard. That explains to me why my PWM temps were so high with the Ninja.

Actually, the fan is mounted well above the motherboard even when the wires were bent around the clips (certainly above the DIMMs, since the fan overlaps them). The bottom of the fan is about 1.5" above the motherboard.

EDIT: The XP-120 should work very well with a CPU of this wattage. My problem was that the XP-120 could transfer 100W+ of heat to the air, but only by making it very hot, which cooked the downstream VRM parts.

mattthemuppet · Post by **mattthemuppet** » Tue Apr 11, 2006 7:40 pm

Sizzle - interesting observation, but, as TomZ wrote above, it's a shame you changed several things at once. It makes drawing any firm conclusions, even for your system, rather fraught with pitfalls.

Sizzle · Post by **Sizzle** » Wed Apr 12, 2006 4:34 am

mattthemuppet wrote:Sizzle - interesting observation, but, as TomZ wrote above, it's a shame you changed several things at once. It makes drawing any firm conclusions, even for your system, rather fraught with pitfalls.

This is true, but looking at my changes, I would say it should add more heat.

The Phantom certainly is going to add more heat to the area than a decent fan cooled PSU. I had the Phantom in the case with the Ninja before and my PWM and SYS temps were higher than they were with the SmartPower in the case.

On paper the Scythe moves slightly less air at the same voltage than the Papst fan I had.

I will have to dig this up, but I believe the PWM temp sensor on the An8 Ultra is by the memory slots. Outside of the range of a fan mounted on the Ninja over the memory slots.

mattthemuppet · Post by **mattthemuppet** » Wed Apr 12, 2006 3:23 pm

that's true, it just would have been nice to know for definate, rather than inferred from the other component changes. Besides, I'm a scientist - more than one uncontrolled variable in an experiment sends me into paroxysms

jaganath · Post by **jaganath** » Thu Apr 13, 2006 1:27 am

Besides, I'm a scientist

Out of curiosity, can I ask what field you are in? (no sheep jokes, I promise

) ie physics/chem/biology?

jmke · Post by **jmke** » Thu Apr 13, 2006 2:04 am

Aris wrote: EDIT: also, this is a common known issue with all tower heatsinks. since they blow across the motherboard and not down on it, alot of the motherboard components tend to run alot warmer than normal.

common for who? I have not experienced such a thing in my testing...

Post by **Ralf Hutter** » Thu Apr 13, 2006 5:22 am

jmke wrote:
Aris wrote: EDIT: also, this is a common known issue with all tower heatsinks. since they blow across the motherboard and not down on it, alot of the motherboard components tend to run alot warmer than normal.

common for who? I have not experienced such a thing in my testing...

FWIW, I've experienced it, not only in testing on the workbench, but in built-up systems as well.

In the interests of keeping the Vreg/NB sections (and, I suspect, the DIMMs as well) a bit cooler, I have been staying away from Ninja-style tower heatsinks and using SI-120's, XP-90's or XP-120's, depending on the application.

jmke · Post by **jmke** » Thu Apr 13, 2006 5:27 am

I'm not saying it's not possible; but the statement is exaggerated, not all tower coolers are created equal and neither are motherboards so the position of the PWM area will differ;

blowing air down onto the motherboard in a closed case (build-up system as you say) seems to increase PWM area temps more then a some other fan configurations.

Each case is different so findings may differ (as seen in the post of the thread starter) however calling it "common knowledge" and using it as a "rule of thumb" to avoid Tower heatsinks is a bit short sighted:)

With a few small cardboard ducts a tower heatsink will seriously increase noise/performance ratio over those "blown down onto the motherboard" units. -- it drops temps a few degrees (http://www.madshrimps.be/?action=getart ... rticID=406) on the PWM area of this DFI board.

"
System 1: P4 2.4C, Intel 875PBZ, Scythe Ninja + 5V Nexus. 2 x 512MB Mushkin PC3200 Level II RAM, SP1614C, SLK3700BQE. "

that SLK3700 is very similar to the Antec Sonata (2) internal layout wise, what PWM temps are you seeing?

Post by **Ralf Hutter** » Thu Apr 13, 2006 5:42 am

jmke wrote: "System 1: P4 2.4C, Intel 875PBZ, Scythe Ninja + 5V Nexus. 2 x 512MB Mushkin PC3200 Level II RAM, SP1614C, SLK3700BQE. "

that SLK3700 is very similar to the Antec Sonata (2) internal layout wise, what PWM temps are you seeing?

At idle, PWM temps=CPU temps. Same exact system with an XP-120 gives me 2-4°C cooler PWM temps. At load with the Ninja, PWM temps are within 1-3° of CPU temps, with the XP-120 they're around 10°C cooler than the CPU temps.

This system has been slated for removal/replacement of the Ninja (with an XP-90 + 92mm Nexus) for months now, but I rarely use it and haven't got around to it.