Superpuppy 2

tazwegion · Post by **tazwegion** » Wed Dec 21, 2005 12:59 am

Jim, seeing as socketA coolers are interchangable with most (if not all) SS7 & socket370 platforms, perhaps you could start your quest there... my personal favourite (bang-for-buck) heatsink is the AMD issued Athlon style unit and have successfully used these on either SS7 or S370 platforms... for an example, here is the Thermalright SK6+ (all copper) heatsink (Skt.A)

Though not 100% comprehensive... I find the cooler reviews over @ FrostyTech to be most interesting

Stedman5040 · Post by **Stedman5040** » Wed Dec 21, 2005 4:26 am

Jim,

I don't know exactly what the arrangement of the capacitors are like around your zif socket, but it is extremly tight around the P5A. For instance the use of a three prong clip is not possible due to the close proximity of such a capacitor to one of the securing plastic clips on the zif socket. Also along the side of the socket lever there are three capacitors lined very close to the lever.

All that taken into account I was able to secure a Globalwin CAK38 copper cooler onto the socket. OK it has a delta screamer fan on top (sounds like a vacuum cleaner), but this could easily be replaced with something more kind to the ears. I got mine off e-bay in the UK. These coolers keep Socket A chips up to at least xp2000+ real cool and are also suitable for the standard room heater the Athlon1400.

My K6-2+/500 that I have would not boot into windows at 600MHz until I used one of these sledgehammer coolers. They do work real good. The Coolermaster HCC-002 also fits the P5A board and is as good as the CAK38. I picked mine up from e-bay for less than £5, about 10 Canadian dollars.

Using the CAK 38 on my K6-III+/450 at 550MHZ and 1.8V my temperatures stay around 20C according to the Aopen monitor. I don't believe it, but I need to perhaps compare the results with say a standard socket7 cooler. This was with an ambient room temp of about 18 to 20C. Using the same arrangenment but at 2.0V on the P5A temperatures get up to 32C without temperature compensation and up to 40C with temperature compensation.

Stedman

tazwegion · Post by **tazwegion** » Wed Dec 21, 2005 5:58 pm

Stedman5040 wrote:I don't know exactly what the arrangement of the capacitors are like around your zif socket, but it is extremly tight around the P5A. For instance the use of a three prong clip is not possible due to the close proximity of such a capacitor to one of the securing plastic clips on the zif socket. Also along the side of the socket lever there are three capacitors lined very close to the lever.

Funny you should say that... my 'first ever' aftermarket cooler (Just Cooler) was purchased of eBay and had been previously fitted to an Intel PIII, it too had a 3 prong clip which required trimming to fit my Epox MVP3C2, allowing then for 600Mhz on a K62 CXT core processor

Jim · Post by **Jim** » Wed Dec 21, 2005 6:34 pm

K, guys, thanks. First @ Taz: Without disrespect that is the kind of heatsink that I chop up to make passive sinks out of. Reason being, yes it is a great passive heatsink, (or even active one for people living out in the country); but it is also a dust trap extraodinaire when used in an active capacity here in "THE BIG SMOKE". The fins are so closely spaced that they naturally trap dust very well. I am willing to sacrifice some surface area to get wider fin spacing so that the dust blown into the sink has some chance to escape.

Second: Am really looking I guess for a Coolermaster in copper. Third: yes the caps are just as close on the P5A-B, and probably a bunch of other boards. You have to do a bit of creative sawing or filing to use that type of heatsink clip.

Edit : Should have clearly stated that Superpuppy 2 is limited, (by Processor location vs HDD location & AGP Location), to heat sinks w/ a 50MM square base. No limit on height; but only 50mm square. Checked E-bay, and did not see anything that looked applicable. Also checked "Frosty" and did not see anything useable there either. Probably have to cut down a larger sink or go to liquid cooling. (Lots of liquid sinks will fit).

tazwegion · Post by **tazwegion** » Thu Dec 22, 2005 1:47 am

Well seeing as your in a pickle with case dimensions etc. have you considered an intake port directly above/beside the CPU's HSF combo... either fan-forced or passive?

BTW with a heatsink such as that SK6+ (with closely placed fins) a high CFM fan is a must have item to help insure dust doesn't collect as easily

Jim · Post by **Jim** » Sun Jan 15, 2006 8:21 pm

K, Superpuppy-2 has had its rebuild and is back online, and seems to be working OK once again. As originally built it featured a mobo modified as shown :

and open it looked like this :

Well since I had problems running OSMark I decided to do what I could to try to get the machine to be able to run that program. As those of you who follow my posts know, there is a type of heatsink that I refer to as a "Dust Trap Extraordinaire"

These sinks work exceptionally well because of their large surface area for as long as they stay clean; but they don't stay clean very long.

I got this one out of the garbage at a computer shop and asked if I could have it; and was told : "Sure, but the fan doesn't work". Got it home and tried it; and found the fan worked fine; BUT :

Well I cleaned it up and kept it, because I had another identical one that I messed up trying to cut a relief for the processor locking arm on a FIC VA_503+ board. The messed up one got the chop with a hacksaw.

The removed section then got filed with a large flat file until all the saw marks were gone and it was approximately flat. Next I used a sheet of 5/8 inch thick glass as a base to lap the removed section using wet sandpaper. First 800 grit, then 1000 grit, then 1500 grit, then 2000 grit.

Once I had a good finish on the underside of the removed section, I used a razor saw lubricated w/ dish soap to cut it up into individual heatsinks for various mobo chips. When that was done, I relapped each individual sink using 2000 grit lubricated w/ dish soap to clean up the edges where they had been sawed.

Then I installed them on the P5A-B using thermal transfer epoxy.
Here we see the cache chip and another (function unknown to me chip) w/ their heatsinks :

Here is another view showing those same two sinks and one of the chipset fans :

And here is a view of the PLL heatsink :

Here we have Mosfet sinks 1 & 2 :

Here we have Mosfet sinks 3 & 4 :

And here is a closeup of my little homemade chipset fan power splitter :

And finally an overall view of the revised board :

With appologies for lousy camera work on some of these shots but they were the first I have taken w/ my webcam; and I don't have much in the way of lighting to use.

EDIT : Thought I would stick in the best results I ever got from SP2 before it went down on account of damaged traces.

Having installed three 256 Meg sticks of Hynix HYM71V32635HCT8P-K, (PC133 CL2), in Superpuppy 2, I decided to see if I could get 112 MHz FSB stable.

Superpuppy 2 : ASUS P5A-B Rev 104 w/ K6-3+ 450 ACZ @ 5.5 x 112 MHz @ 2.3v. w/ mobo cache enabled. (First attempt; - has gone well, will try to reduce to 2.2v. later).

Everest :
Memory Read = 361 MB/s
Memory Write = 177 MB/s
Memory Latency = 182.8 ns

Sandra :
Performance Rating (Estimated) = 740
Arithmetic = 1521 D ALU MIPS ; 731 W FPU MFLOPS.
Multimedia = 3544 INT MMX ; 4217 FLOAT 3D NOW!
Memory Bandwidth : (Run repeatedly on account of rising results)
INT MMX : 210 ; 213 ; 229 ; 231 ; 238 ; 241 ; 240.
FLOAT FPU : 210 ; 216 ; 228 ; 234 ; 237 ; 237 ; 237.
Cache & Memory Bandwidth : 1164 - 2k ; 1179 - 4k ; 1183 - 8k ; 1164 - 16k ; 1120 - 32k ; 1036 - 64k ; 915 - 128k ; 630 - 256k ; 268 - 512k ; 236 - 1Meg ; 236 - 4Meg ; 235 - 16Meg ; 237 - 64 Meg ; 236 - 256Meg.

Rerunning Everest AFTER Sandra gave the following results :
Memory Read = 365 MB/s
Memory Write = 201 MB/s
Memory Latency = 184.3 ns

All tests run under WIN98-SE ; w/ Power Tweak 2's tweaks ; w/ : CAS Latency = 2T ; RAS to CAS Delay = 2T ; RAS Precharge = 2T ; RAS Active Time = 6T ; Row Cycle Time = 9T. - & my usual mess of 10 taskbar items running.

Currently standing in for Superpuppy-3

Jim · Post by **Jim** » Sun Jan 22, 2006 9:33 pm

Having decided to rebuild Superpuppy-3 as a "Dust-Tight" machine, I have also decided to use one of those closely spaced fins - "Dust Trap Extraordinaire" heatsinks. Looking for a copper one, similar to that posted by TAZ at the top of this page. Any assistance in locating one of same would be greatly appreciated.

Jim · Post by **Jim** » Wed Nov 15, 2006 4:30 pm

Superpuppy 2 has been down w/ damaged mobo traces for a while now. Still works but only w/ 16 colour graphics. The new board arrived today. Gonna have my hands full for a while finishing SP3, then doing a complete rebuild of SP2 using the case I modded for SP4. Only when all that is done will I tackle Superpuppy Z - (My last dog).

Jim · Post by **Jim** » Sun Nov 26, 2006 9:34 pm

Question for all the more knowledgable out there. I have now got 2 excellent ASUS P5A-B Rev 104 boards, (w/ caps in "purr-fect" condition), also have an ASUS P5A-B Rev 105 on the way. Since it takes a heck of a lot of work to mod the boards sticking heatsinks on everything, I would like to know people's opinion as to which board I should use. Does anyone know if the chipset internal tagram works on Rev 105s?

Uranium235 · Post by **Uranium235** » Sun Nov 26, 2006 11:37 pm

Does anyone know if the chipset internal tagram works on Rev 105s?

Depends on the revision of the ALi V chipset. Revision "F" and later revisions had their internal TagRam enable allowing the boards to have a cacheable memory area of 512MB. The P5A-B 1.05's and 1.06's were the ones that worked slow with the K6+ CPU's though.....

Jim · Post by **Jim** » Mon Nov 27, 2006 2:05 am

Yeah, so I have heard. -- Except there is one guy, rls48j, who instead of flashing his bios to the 1011, spent months working with WPCredit to get his 106 working properly. So maybe I can do it too. THANKS for the info about the chipset ! - Will check the board when it arrives.

DonPedro · Post by **DonPedro** » Mon Nov 27, 2006 4:00 am

@jim, uranium235

my information is - and I can verify this on any chipset "F"-revision board I own, that the internal tag-ram bit is disabled.

@jim
I would stay away from the 1.05 and 1.06 boards if you plan to spent time on pimping them up. I have contacted rls48j about the 1.06 solution: he now says there is none! what he achieved through some tweaking was that the performance went up, but he failed clearly to make the 1.06 as fast as it should be. I have two 1.06s (with different clock-plls) but both start showing mem-write performance up from k6-III-450 (NOT the +-version). the +-version cpus are a waste for these boards. the k6-3-400 runs perfectly.

Jim · Post by **Jim** » Mon Nov 27, 2006 7:29 am

THANK YOU very much for the info Peter ! Now comes the question : You say the chipsets internal tagram is "Disabled". Do you mean "Disabled" as in the manner in which a K6-2+ has half its on-die cache "Disabled" ? (i.e. somehow physically disconnected), or do you mean "Disabled" in the sense of being "switched off" ? I ask because I know WPCredit has a setting to "Enable / Disable" the internal tagram.

DonPedro · Post by **DonPedro** » Mon Nov 27, 2006 9:18 am

jim, to be honest I don't know whether it is electrically "dead" or just "switched off". but I think it is switched off because the other way would have meant to change the chip's design. if they had done it this way, then they - at the time back then - could have also "repaired" the chip's faulty layout. but they have decided to add an external chip. I think they did it because it was cheaper than throwing away millions of chips they already had produced. and so they just "switched it off" by software. when stocks plenished they produced the "G" revision with repaired layout. I think that I tried once to turn on the internal tag-ram on an "F" revision board via wpcredit and the system crashed, but about that I am not 100% sure.

but the internal/external tag-ram thing is not worth to think about a second if you disable onboard 2nd level cache anyway in case you are planning to use a cpu with on-die cache (+-series, -III). the cpu's on-die cache is able to cover an area of 4GB ram (it has an in-build tag-ram mechanism). this is equally true for the k6-2+ which features only 128kb on-die cache. by disabling onboard cache you win performance on some applications, on others you loose. this depends on the application's pattern of read/write usage of ram. also I would like to note that the amount of ram that can be cached (addressed) is one thing (performed by the tag-ram), cacheing/holding data in cache another (the cache itself)! this means that there are applications that do profit from increasing SIZE of the cache-ram (512kb, 1mb, 2mb). these applications would be hurt most if the onboard cache-ram is disabled. my next bench-suite I am about to begin will deal with observing this kind of behaviour. I have 3 via boards (shuttle, dfi, epox) with respective amount of onboard-cache (512kb, 1mb, 2mb). I hope you can help a lot on the wpcredit tweaks here. since the info on this seems to be scattered through many threads through time and space some summary would be very very welcome

but this is another story and off-topic ....

Uranium235 · Post by **Uranium235** » Mon Nov 27, 2006 10:53 am

One sure way to tell if the internal TagRam is being used is if the motherboard's external TagRam was physically removed from the board. The internal TagRam had bugs in the earlier chipset revisions necessitating the need for an external TagRam. This bug was allegitly fixed from revisions F and later but not all motherboard manufacturers chose to implement the internal TagRam for whatever reasons.
Sorry I wasn't more accurate before.