skv400 wrote:

The reason for the 486DX 50MHz and 486DX2 66MHz results is the same because 50MHz is the DX-50 of FSB 50.

Err... no. MIPS are Million Instructions Per Second, which are not affected by FSB. A DX2/66 must have more MIPS than a DX/50. The fact that it doesn't just shows that 7zip is not a reliable benchmark and also doesn't prove that the CPU became any faster.

Also, the way you attached the heatsinks is very ineffective. The die sits in the middle of the CPU, and that's were most heat is emitted. It would have been much better to use only a single heatsink placed in the middle.

What is your motherboard revision?

derSammler wrote:

Err... no. MIPS are Million Instructions Per Second, which are not affected by FSB. A DX2/66 must have more MIPS than a DX/50. The fact that it doesn't just shows that 7zip is not a reliable benchmark and also doesn't prove that the CPU became any faster.

Also, the way you attached the heatsinks is very ineffective. The die sits in the middle of the CPU, and that's were most heat is emitted. It would have been much better to use only a single heatsink placed in the middle.

I don't know how unreliable the test is, but technically according to 7zip,

Compression speed strongly depends from memory (RAM) latency, Data Cache size/speed and TLB.
Decompression speed strongly depends on CPU integer operations.

So the bus speed will at least affect the compression speed, and therefore the combined speed of the two, which is the MIPS recorded by the OP.
I agree that the placement of the heatsinks was odd... I'd get two more pieces and cover the whole top. It's such a good CPU that I wish you are in the US so I can send you my spare heatsinks.

This cpu when sold at retail used to come with a cooler in box for a good reason and not sure why people would skimp on cooling.

nforce4max wrote:

This cpu when sold at retail used to come with a cooler in box for a good reason and not sure why people would skimp on cooling.

I don't know either,but i'm kind of a cooling maniac.
I also dislike huge heatsinks, on modern system i rather go liquid cooled than a air cooled humongous heatsink.
All my atx systems have at least one intake fan, and my p4 have one intake and exhaust fan(besides the PSU fan)

nforce4max wrote:

This cpu when sold at retail used to come with a cooler in box for a good reason and not sure why people would skimp on cooling.

It had the standard Cyrix green heatsink. Now-a-days I can't seem to find them anywhere. They are even rarer then the cpus 😉

derSammler//I acknowledge that Heatsink is installed in an inefficient location.
However, it is mounted at this position to protect the characters on the CPU surface.

kixs//This Board's revision is 1.01. and i have one more same motherboard. i don't know that board's revision but maybe same.

nforce4max//Unfortunately I have only a CPU.

Not too long ago I picked up a pristine 33/66 with original factory heatsink. I don't understand why 99% of them don't have heatsinks, considering all of them came with one in the box.

skv400 wrote:

Compared to the 10MIPS before the cooler, 12MIPS after the cooler has been improved by 20%.

This makes no sense to me.
If you did nothing else other than adding a cooler to the system, there is no way the performance could have changed.
Cooling only affects modern systems because they have active temperature sensors and clock throttling circuits. So the CPU can control the clockspeed based on temperature.
This sort of technology did not exist back in the 486 days.
Ergo, a CPU performs exactly as it is configured, until it crashes because of heat-related instability.
Apparently it didn't crash without the cooler, because you got measurements.

Anonymous Coward wrote:

Not too long ago I picked up a pristine 33/66 with original factory heatsink. I don't understand why 99% of them don't have heatsinks, considering all of them came with one in the box.

Probably because most CPUs weren't boxed models, but OEM ('tray') ones?
These are sold without heatsink, because the idea is that the OEM adds a cooling solution.

Aren't you thinking of the 486DLC? That was an OEM CPU. I am pretty certain that the DRx and SRx chips were primarily for retail sale. At least, the majority of them should have been. They were designed specifically to upgrade older motherboards.

Anonymous Coward wrote:

Aren't you thinking of the 486DLC? That was an OEM CPU. I am pretty certain that the DRx and SRx chips were primarily for retail sale. At least, the majority of them should have been. They were designed specifically to upgrade older motherboards.

I don't know... I'm just saying that most, if not all, CPUs generally came in either boxed/retail version or tray/OEM version.
There are many CPUs that 'should have been' for retail, but various shops would just order tray models and sell those directly to end-users (because that allowed them to sell the CPUs a bit cheaper, and also give the end-user choice in what cooling solution to use). It could explain why you often see them without heatsink.
Upgrading older motherboards is of course something not exclusively aimed at end-users. Generally shops would offer this as a service. In which case, buying the CPUs in bulk/tray, and buying heatsinks separately may have been common.
In other words: we need to find out how Cyrix sold these CPUs to resellers.

IIRC (as in 24 years ago) the Cyrix 486SRx/DRx came in plastic packaging with a floppy and hung from racks.

derSammler wrote:

skv400 wrote:

The reason for the 486DX 50MHz and 486DX2 66MHz results is the same because 50MHz is the DX-50 of FSB 50.

Also, the way you attached the heatsinks is very ineffective. The die sits in the middle of the CPU, and that's were most heat is emitted. It would have been much better to use only a single heatsink placed in the middle.

That 3M thermal transfer tape has lousy heat conduction (I mean, it’s better than using double sided tape, but still). That tape’s heat conduction is equivalent to sticking about a dozen mica insulators between the heatsink and CPU. Lousy.

The OP really needs thermally conductive epoxy, which contains finely ground steel, aluminum or silver in electrically conductive varieties and powdered ceramic or silicates in non-conductive varieties.

In the past I used Arctic Alumina Thermal Epoxy to bond small heatsinks to chips, but they don’t seem to make it anymore. Now I use MG Chemicals Thermal Adhesive with good results.

I’ve also used JB Weld (which is an off the shelf steel infused epoxy you can buy at any hardware store) in a pinch, when I was out of the real stuff and on a time crunch, to attach very small heatsinks and thermocouples with no problems.

Note: The MG Chemicals adhesive is *electrically conductive*! Only apply a thin layer and leave at *least* a 10mm gap around the edges of the chip, because the epoxy will spread out when you push the heatsink down and you do *not* want it falling over the edge and shorting something out.

Anyone know what Cyrix was using to attach their green heatsinks? It appears to be a thicker piece of aluminium foil with adhesive on both sides.

A 386-class system with 64 MB of RAM. That isn't very common. Can the 256K cache all the RAM? Is the cache write-back of write-through?

Does the game you played depend on the FPU?

Could you run Doom timedemo3 in fullscreen mode? This is a more common benchmark that everyone tends to use when comparing 486 systems.

I also do not understand why all the DRx2-66's which turn up are heatsink-less. Did the adhesive or thermal pad on them have a shelf life which has expired? Anonymous Coward, how stuck would you say that heatsink is on yours? I realise this question is overly subjective.

I would also like to determine a good substitute for what Cyrix used. I bought a lot of thermal pads once upon a time but was afraid to use them. I don't want the lettering to come off the CPUs in some years when the heatsink needs to be removed. I've had issues with some thermal epoxies, when the heatsink was removed, so was half the letters.

Scali's explanation concerning the lack of heatsinks is certainly plausable, however, it does make me wonder what the OEM's and repair shops were using to cool these chips. I haven't come across a 486-like removable heatsink solution for PGA-132 CPUs. For this reason, I have assembled two of these which seem to work well: Re: Thermal tape for heatsinks, particularly for DRx2-66 and SXL2-50+ My second attempt looked a lot cleaner than the photos shown there (which was my first attempt).

I've used thermal epoxy from Omega Engineering. The shelf life was about 3 years. That stuff gets damn hard like JB Weld though. I also have the MG Chemicals syringes, which I have only used once to glue the heat-spreader back onto a K6-200 CPU.

EDIT: I'm going to need to find some more removable thermal adhesive to attach a heatsink to my TI 486SXL2-66 QFP144 chip and not damage the wording. If someone has positive experience with something which does not damage the wording yet holds well enough to secure a heatsink with rotating fan, please advise.

I would say the heatsink on my DRx2-33/66 is secured about as well as the ones on my cx5x86s. I haven't pried at it, but it doesn't seem like it would just randomly come off. I would guess that had the heatsink been attached in the first place, that most of the loose ones you see floating around would have had that diamond pattern from the texture on the thermal pad that you often see on the 5x86s and cx486s that have had their heatsinks removed.

My best guess is that most of the people who bought the kit were just too lazy to install the heatsink in the first place. I am fairly confident that the DRx chips were not offered as OEM upgrades, especially the 66MHz part which had relatively poor yields compared to the other speed grades. Had they been sold to OEMs, they obviously still would have required a cooling solution, which likely would have left a glue mark if the heatsinks had been removed. My guess is that OEMs would have used a thermal epoxy that would have left a serious mark on the surface of the CPU.

Does it have that mesh thermal pad? I mean, if you look at the side, do you see a thin white thermal pad?

Also, why do these QFP144 TI486SXL2-66 chips all seem to come without a heatsink, while the PGA168 versions of the same chip have a heatsink? Were there any clip-on QFP144 heatsinks I can use on this thing? Or some other wild z-clip type of thing? Do they not need a heatsink at 3.6 V? I was thinking about removing one of the heatsinks from my PGA168 SXL2-66 chips and using it on the QFP144, but I still wouldn't want to use thermal epoxy.