The Serpent Rider wrote:

Why even bother? If it's your main system, there's vastly better options to upgrade for a similar price.

There is something about trying to take a piece of hardware to the max that excites and entertains me. Of course it may be impractical, but I don't care. The majority of things people do and report on this forum are impractical from a financial and usability perspective.

Koltoroc wrote:

32bit windows is realistically on the way out. For a few years now there is software that doesn't come as a 32bit version any longer, particularly newer games to the point that 32bit windows is useless for a modern gaming machine. Granted; this is likely not relevant here, but it IS where this is going. Nvidia has a few months ago stopped supporting 32bit windows alltogether so the drivers are likely to be stuck at the versions from a few months ago. You can expect more hardware manufacturers to follow suit.

The main "downside" of 64 bit windows or advantage of 32bit is that 64 bit windows did away with the support for 16bit code. The only time I ran into this as an issue were old late 90s game installers that tend to be 16bit programs. That is something to be aware of if you use a lot of old (and I mean REALLY old) windows programs. In my experience it is not an issue unless it is some exotic piece of software that vital for someones business but is long out of production, hasn't seen support in decades and nobody knows how exactly it works.

Thanks for this; I am horrifically outdated with how computers and Windows are trending. I tend to use hardware/software until I'm forced go move on and this is usually at the point in which I cannot accept the speed of the internet.

I do not intend to play any games, modern or old, on this system. Nvidia stopped support 32-bit Windows? I had no idea. My main objective is for GPU video acceleration, so this is disappointing. Is AMD still in the 32-bit game? I don't care about 16-bit code on this system as I have a dozen other systems which can use that if needed.

I have the Asrock AM2 CPU card for my motherboard, but was going to hold off on finding a CPU for it for as long as possible. It has its own DIMM slots, which says DDRII-800. Is it correct to assume that the RAM limit of this CPU expansion card follows that of the original motherboard, that is, 4 GB? I only have DDR1 and DDR3 RAM. I assume DDR3 won't work in there? The AsRock AM2 upgrade card does have a 4-pin fan header, which is nice for throttling. Any idea on what the fastest AM2 CPU is which this card will use properly?

My only other non-P4 desktop is an original Mac Pro (4-core), which contains dual Xeon 5160's at 3 GHz. The system has 4 GB of RAM and runs Ubuntu 16 LTS. I have a pair of X5365 Xeons I could put in there to make it an 8-core system at 3 GHz, but not sure if there is much consumer benefit in doing so. The electrical drain would be higher (130 W vs. 300 W TDP). It is tempting to switch over, however, I get frustrated with how Linux Excel, Office, and other files don't always display and edit as intended. I suppose I could try and install Widnows on it, but this is a whole different topic. My intent here is to use the motherboard I have, correct the overheating issue, and obtain some browser video acceleration.

ph4nt0m wrote:

Once you install Linux, you don't care what M$ supports and what not. A dual core Opty with 8Gb of RAM can handle all common everyday tasks.

Can it use the graphics card to accelerate videos in a web browser? I am tempted to switch, though I have been tempted for decades.

There is something about trying to take a piece of hardware to the max that excites and entertains me.

You already maxed out with Opteron 185.

ASRock AM2 card supports all AM2 CPU's that are based on K8 architecture.
Basicly, 6400+ x2 is best you can get, while 65nm stuff is easiest to cool down.
Adapter supports 8GB of DDR2 memory :

While board itself works fine with up to Kepler GPUs (at least, that is how far I tested it) :

agent_x007 wrote:

Adapter supports 8GB of DDR2 memory :

Are these screen shots of a motherboard using an AM2 CPU card? Which motherboard was used? EDIT/answer: ASRock 939Dual-VSTA.
...
How is it possible that a motherboard can support up to only 4 GB, yet it will work with 8 GB when using the AM2 card? How much RAM does the BIOS display before POST?
...
Looks like the 65 nm chips contain half the L2 cache of the 90 nm Athlon 64 X2 6400+ . Perhaps a reasonable trade-off for less heat.

How is it possible that a motherboard can support up to only 4 GB

The answer is simple: unbuffered DDR stopped at 1gb.

How is it possible that a motherboard can support up to only 4 GB

unbuffered DDR1 modules go only up to 1GB, DDR2 go officially to 2GB. I say officially because there are 4GB "High density" modules. these modules are seemingly organized similar to FB dimms and can be used on AMD CPUs because of some peculiarities of how the IMC is implemented. You can get those on ebay for rather cheap, search for AMD only DDR2 memory.

A word of warning here, this type of memory is officially unsupported and may not work, may be unstable or might just work at half capacity. IIRC they won't work on intel at all. The Motherboard/bios has something to do with possible compaility, but what exactly seems to be unknown.

Since you like pushing old hardware to the limit and beyond seems to be your thing I thought that might something you may be interested in experimenting with.

Also, for low profile coolers I would look at noctuas coolers, they are rather good but also rather expensive. However, finding low profile coolers that work well with higher TDP CPUs is somewhat difficult and I would be weary of cheap ones.

The Serpent Rider wrote:

How is it possible that a motherboard can support up to only 4 GB

The answer is simple: unbuffered DDR stopped at 1gb.

Wouldn't the BIOS and the ULi M1695 northbridge need to support 8 GB of RAM? Does the northbridge on the 939dual-SATA2 or 939dual-VSTA support 8 GB?

Koltoroc wrote:

there are 4GB "High density" modules. these modules are seemingly organized similar to FB dimms and can be used on AMD CPUs because of some peculiarities of how the IMC is implemented. You can get those on ebay for rather cheap, search for AMD only DDR2 memory.

So I could possibly take this MB to 16 GB?

feipoa wrote:

ph4nt0m wrote:

Once you install Linux, you don't care what M$ supports and what not. A dual core Opty with 8Gb of RAM can handle all common everyday tasks.

Can it use the graphics card to accelerate videos in a web browser? I am tempted to switch, though I have been tempted for decades.

I think Linux supports it for ages. The actual performance is browser and driver dependent of course. The last time I've had issues with it when working around Intel 945GM quite a long time ago.

https://wiki.archlinux.org/index.php/Hardware … eo_acceleration

feipoa wrote:

Wouldn't the BIOS and the ULi M1695 northbridge need to support 8 GB of RAM? Does the northbridge on the 939dual-SATA2 or 939dual-VSTA support 8 GB?

there is no "northbridge" as such on Athlon64 systems. The Chipset is connected to the CPU over Hypertransport and is not directly connected to the system memory, because the Memory controller is inside the CPU, the chipset has no influence on memory support. That is why the AM2 upgrade card can work in the first place.

the concept of northbridge and southbridge died with the advent of integrated memory controllers.

the bios might have some say in this, but memory sizes described in manuals are usually just the stuff available at the time of printing, not what is actually possible. Most of the time it is not a problem however.

feipoa wrote:

So I could possibly take this MB to 16 GB?

possibly yes. But as I said, compatibility is, lets say, questionable. Impossible to know for sure without actually trying.

Neat. Kinda like with the Cyrix MediaGX. Took some time for that to catch on for the masses.

I have removed the heat spreader on my Opteron 185 and have a few questions.

1) Has anyone performed a delidding and left the heat spreader off, thereby placing the heatsink/fan in direct contact with the surface below? What was the outcome? Wouldn't this be better for cooling, that is, to remove one of the layers with heatsink compound?

2) What do you glue the lid back on with? I assume ordinary contact cement might not fare well with the 70 C heat over itme.

3) If I am not going to glue the lid back on soon, is there any purpose in removing the excess glue on the underside of the lid? See 3rd image.

4) Has anyone glued these back on with permanent heatsink epoxy? Wouldn't this offer a longer lasting solution? Or is it simply that the epoxy-based thermal compounds far worse at conducting heat compared to the paste?

5) Anyone try using thermal pads instead of paste? Do you have any numbers to share in terms of effects on temperature reduction or addition?

6) After re-pasting the lid, how are you smashing the lid down hard? Seems like the more pressure the better, to a) reduce air gaps and b) reduce thickness of the paste. I was thinking it might not be a good idea to press the lid down hard with the CPU inside the motherboard because of MB PCB deflection.

Wouldn't this be better for cooling, that is, to remove one of the layers with heatsink compound?

Yes, it will be better. But you can potentially damage the crystal. Cooler bracket might need a modification too.

If I am not going to glue the lid back on soon, is there any purpose in removing the excess glue on the underside of the lid?

No,

What do you glue the lid back on with?

Silicon adhesive for example.

What, like the stuff I use at the shower basin? That stuff don't see 80 C though.

I ultimately used Ceramique 2 and stuck the IHS on without the edge adhesive. I put the heatsink fan back on, but boy was the IHS moving all over the place. Took a few tries before I had the IHS on relatively centred.

Fired it up. Let it sit in the BIOS with the Hardware Monitor page visible for 10 minutes. CPU temp eventually crept up to 66 C and stayed at this value. Previously, it would get to the mid-70's and hang. So, progress.

Booted XP and ran a youtube video in Chromium. I played a video in HD on full-screen. The video, by selection, is about this motherboard, the AsRock 939dual-SATA2. https://www.youtube.com/watch?v=g3-4J_HiUO8 I let it run for a few minutes and took a screenshot showing the CPU temperature. CPU usage was around 93% as an average and the max CPU temp was 72 C. Previously, it would get to around 82 C before hanging. So, things have improved and I'm hoping a better heatsink/fan will help further still. At this point, I might also want to invest in some better thermal compound. Is Arctic MX-4 is the good stuff these days? Also, anyone have a part number for the edge adhesive?

feipoa wrote:

I have removed the heat spreader on my Opteron 185 and have a few questions.

1) Has anyone performed a delidding and left the heat spreader off, thereby placing the heatsink/fan in direct contact with the surface below? What was the outcome? Wouldn't this be better for cooling, that is, to remove one of the layers with heatsink compound?

It would improve the thermals a bit, but less then you think. The problems are finding a suitable cooler and the chance to damage the die. Finding a suitable cooler is an issue because they are designed with the heatspreader present. The cooler must still be able to exert enough pressure to make good contact despite the loss of height on the CPU. Coolers using the default mounting bracket are generally not a good choice. I would avoid it.

feipoa wrote:

2) What do you glue the lid back on with? I assume ordinary contact cement might not fare well with the 70 C heat over itme.

silicon glue, just like the stuff used originally

feipoa wrote:

3) If I am not going to glue the lid back on soon, is there any purpose in removing the excess glue on the underside of the lid? See 3rd image.

probably not. If you glue it back on, you should remove it however.

feipoa wrote:

4) Has anyone glued these back on with permanent heatsink epoxy? Wouldn't this offer a longer lasting solution? Or is it simply that the epoxy-based thermal compounds far worse at conducting heat compared to the paste?

Bad idea. Thermal conductivity is lower, and you kinda lose the ability to replace the compound after. thermal epoxy is mostly suited for lower power components like ram chips.

feipoa wrote:

5) Anyone try using thermal pads instead of paste? Do you have any numbers to share in terms of effects on temperature reduction or addition?

even worse idea than the thermal epoxy.

feipoa wrote:

6) After re-pasting the lid, how are you smashing the lid down hard? Seems like the more pressure the better, to a) reduce air gaps and b) reduce thickness of the paste. I was thinking it might not be a good idea to press the lid down hard with the CPU inside the motherboard because of MB PCB deflection.

just install it on the motherboard, install the cooler and let its pressure do the job. To make it easier spread the thermal compound on the die to cover all of it as thinly as possible. You don't want parts of the die uncovered with thermal compound.

Works also if you glue it down again. Install CPU, put thermal paste on the die, put glue on the edge of the heatspreader, put heatspreader on the CPU, clamp down with your cooler (no thermal compound to prevent it sticking), and let it sit until the glue is dry (24hours would be my suggestion regardless). After the glue has dried completely remove the cooler, apply thermal compound and then apply the cooler. i would not run the system while the glue is drying to prevent weird results from the glue.

Realistically, you don't have to glue the heatspreader back on. The Pressure of the cooler is definitely enough to keep it in place.

feipoa wrote:

I have removed the heat spreader on my Opteron 185 and have a few questions. […]
Show full quote

I have removed the heat spreader on my Opteron 185 and have a few questions.

1) Has anyone performed a delidding and left the heat spreader off, thereby placing the heatsink/fan in direct contact with the surface below? What was the outcome? Wouldn't this be better for cooling, that is, to remove one of the layers with heatsink compound?

2) What do you glue the lid back on with? I assume ordinary contact cement might not fare well with the 70 C heat over itme.

3) If I am not going to glue the lid back on soon, is there any purpose in removing the excess glue on the underside of the lid? See 3rd image.

4) Has anyone glued these back on with permanent heatsink epoxy? Wouldn't this offer a longer lasting solution? Or is it simply that the epoxy-based thermal compounds far worse at conducting heat compared to the paste?

5) Anyone try using thermal pads instead of paste? Do you have any numbers to share in terms of effects on temperature reduction or addition?

6) After re-pasting the lid, how are you smashing the lid down hard? Seems like the more pressure the better, to a) reduce air gaps and b) reduce thickness of the paste. I was thinking it might not be a good idea to press the lid down hard with the CPU inside the motherboard because of MB PCB deflection.

Use some liquid metal and silicone sealant to assemble it back. I've used Coollaboratory Liquid Pro and Dow Corning 3140 RTV. Also soldered those missing capacitors just in case. A dead Sempron served to this purpose.

RTV, yes... I have some of that for when I perform a differential service on my car and need to pop the cap off (though I've recently been using resealable gaskets with great results). Differential RTV is good at high temps. I also have a bunch of thin thermal tape I've been waiting to use. Perhaps I could use that to tack the lid down on its 4 corners, and 4 centres?

In cases whereby the heatsink/fan clamp doesn't suffice, I have thermally epoxied the heatsink down the "crystal", as you call it. I've only done this with plentiful CPUs, like Tualatins. Opteron 185 is probably too rare and I'm in the mist of testing out heatsinks, so not the best idea. Anyone have temp. numbers when using MX-4 vs. liquid metal?

Koltoroc wrote:

feipoa wrote:

1) Has anyone performed a delidding and left the heat spreader off, thereby placing the heatsink/fan in direct contact with the surface below? What was the outcome? Wouldn't this be better for cooling, that is, to remove one of the layers with heatsink compound?

It would improve the thermals a bit, but less then you think. The problems are finding a suitable cooler and the chance to damage the die. Finding a suitable cooler is an issue because they are designed with the heatspreader present. The cooler must still be able to exert enough pressure to make good contact despite the loss of height on the CPU. Coolers using the default mounting bracket are generally not a good choice. I would avoid it.

What about a copper shim? It would conduct heat better than the original aluminum but I don't know if the end result would amount to much. Copper is also prone to corrosion, but since these things are used in a dry indoor environment maybe that's not a big issue.

shamino wrote:

What about a copper shim? It would conduct heat better than the original aluminum but I don't know if the end result would amount to much. Copper is also prone to corrosion, but since these things are used in a dry indoor environment maybe that's not a big issue.

If by "shim" you mean the heatspreader, it is made of nickel plated copper.

The IHS definitely felt a lot heavier than solid aluminum.

While waiting for my fan, I've ordered some MX-4. I read several reviews on the different pastes and this seems to be decent. Plus, I needed to add something small to my cart to get free shipping on amazon.ca. Will use this on the IHS and use some Permatex gasket sealer. I used to have some Honda Bond for gaskets. That stuff was particularly good for high temp, but cannot seem to find it at the Honda dealer anymore. Probably renamed it or something.

RTV sealant seems too flexible to use as an adhesive. It tears if subjected to repeated flexing. But a CPU heatspreader is unlikely to experience much twisting or pulling anyway so it would probably work fine.