Mondodimotori wrote on 2024-08-19, 12:46:
Yeah, I actually don't know in wich direction the fan of the new PSU will blow.
...
How do I check on the box where the PSU fan blows? Or it's something I discover after installation?
For PSUs with a 80 mm fan at the rear (exhaust), the direction is always outwards (out of the case)... but that we all know. 😉
For PSUs with 120/140 mm fans at the "top", the airflow direction is always inwards (PSU sucking air from the fan side and blowing it out the back grille out of the back of the PC case.)
For PSUs with 90 mm fans at the top.... it really depends on the age and OEM of the PSU. Very, very PSUs had the fan where it's sucking air from the back (rear of the PC case, where the AC cord is) and blowing it inwards into the case. As mentioned, I only ran into one PSU like this (the Morex above). Every other ATX PSU with a 90 mm fan was just like the PSUs with 120 mm fans. But with MicroATX, I've also seen exceptions to this. Case in point was the AcBel OEM PSU in an IBM NetVista 6578 PC I picked up from next to the trash last summer. But aside from these exceptions, every other PSU had "standard" airflow direction (typically PSU taking air from inside the PC case and dumping it out the back.)
So all in all, I think you can expect most PSUs with a side/top 80 or 90 or 120 mm fan to be sucking air in and blowing it out the back.
As for PSUs with multiple fans (i.e. 80 mm in the rear and 80 or 90 mm fan on the top/side), the airflow is still more than likely going to be the "conventional/standard" way.
Mondodimotori wrote on 2024-08-19, 12:46:I found a few used Enermax ATX eg565 for decent prices, but the situation with caps in unknown. They are listed as "working", and the sellers says that the caps "looks good", but who knows how much will they last.
Enermax usually used CEC Holdings (PCE-TUR and PCE-TUL) electrolytic caps, which would sometimes hold up OK and sometimes not. Apart from the caps, they are really well-built PSUs.
If you intend to ever do any soldering or recapping and want to learn, starting with a PSU would be easier compared to recapping a motherboard. So if the price isn't too crazy on those Enermax PSUs, they might be worthwhile getting just for that... and having a nice well-working PSU, of course 😉
.
My personal cheap Go-To choice for socket A systems is the Dell-branded Chicony/HiPro model HP-P2507F3P and HP-P2507FWP. These are 250 Watt PSUs "only", with a low-key (but very solid) 22 Amp rating on the 5V rail. Yes, these PSUs also need a recap sometimes, but they are very easy to work on and have large footprints on the PCB for the caps, thus making it trivial to find replacement options.
Delta and LiteOn had similar equivalents for Dell and HP machines, though they didn't give the same large cap footprint on the PCB.
Mondodimotori wrote on 2024-08-19, 12:46:
But I also found a couple of new PSUs in a local store that have more than 30 amps and 180/200W on the 3.3 and 5v rail. All I need there would be a 20 pin adapter (if they aren't alredy 20+4 pin) and some SATA to MOLEX power adapters. I was actually planning to get those when the store opens again after summer vacation.
New PSU with 30 Amps on the 5V rail? - Most likely it's garbage with a label that's a complete lie (unless it's some 800W+ modern unit with DC-DC rails... but I doubt it.)
If the store allows it, maybe take a picture of the PSU and show it here. Or at least mention the brand and model.
I suspect it will be a gutless wonder PSU. But let's see.
Mondodimotori wrote on 2024-08-19, 12:46:
So I believe I absolutely need some ventilation there, even with a better CPU cooler, because the more heat I vent away from the CPU, the more it'll go elsewhere
Exactly!
Moreover, socket A systems need very good case cooling, because their power consumption (and thus heat dissipation) doesn't really change too much between idle and full/100% load state. My Athlon XP 2500+, for example, draws about 100 Watts from the wall when completely idle, which is quite high. Even my Pentium 4 Prescott 2.8 GHz PC doesn't use that kind of power when it's idle (idles around 75-ish Watts.) When I put a synthetic load on these CPUs, the 2500+ only goes up to ~115 Watts power consumption (again, measured at the wall), while the P4 Prescott isn't that far off at 125 Watts.
But the point I'm trying to make with the above example is that socket A systems are constantly pumping out a lot of heat (whereas the P4 Prescott only dumps a lot of heat under full load, but it otherwise puts out less heat when Idle.) Thus, good case cooling is a must for socket A systems.
That being said, I don't know how easy it would be to achieve this with your current case.
If you're not against some DIY (kludging/janky) case modding, perhaps you might be able to cut some holes into the case and add fans there?
Alternatively, just run with the PC case side "cracked open" on one side... though that may not always be enough.
Mondodimotori wrote on 2024-08-19, 12:46:
Do you actually suggest a smaller PSU than the standard ATX?
If you can find one, and one that has good ratings on the 5V rail too, then YES.
But that might prove to be somewhat difficult route too.
For example, if you go with uATX or flex-ATX PSU, you might need to make a "shim" plate for the back of the PSU in order for it to mount properly into the ATX hole. Depending on your machining skills (or enthusiasm to try it), this might be either something trivial or a complete show-stopper to go this route. But otherwise, I see this as the most viable solution if you really are dead-set on using the current case you have.
Mondodimotori wrote on 2024-08-19, 12:46:
About the caps on the mobo... At a visual inspection they seems to still be good, no bulging and no leaking. I'm not practical with recapping (yet), so I'll try ask around if I find someone locally that does those kind of electronical work. But it wasn't a priority after I inspected them.
Understandable.
I tend to do the same with my personal equipment quite often, even though I have no problem doing any recapping whatsoever (and have the equipment and caps to do it too.)
It all depends on how critical and/or rare the equipment is, though... and how vulnerable to damage if a cap failure occurs. Some hardware will still work OK (or at least tollerate) with bad caps. But some hardware isn't like that at all. Case in point: standard PCI-E version of the EVGA GeForce 7600 GT - when the Sacon FZ caps fail on this video card, in 20-50% of the cases, the RAM chips can get taken out too. And then it becomes a nightmare to repair it. So for such hardware, I do a recap right away. But most hardware is not like this and will typically just crash / become unstable with bad caps. So it's not necessary to be worried and replace everything right away.
The Serpent Rider wrote on 2024-08-19, 13:02:
Yes, it will work with 200/266 FSB on lower speed just fine, but luckily early Thoroughbred/Barton CPU had unlocked multiplier anyway.
The problem is finding an early T-bred or Barton with an unlocked multi. 😉
Not rare, I know... but certainly not as easy to find too.
Mondodimotori wrote on 2024-08-19, 13:10:
So you'd just fix up the multiplier to match up the original frequency?
Exactly.
BitWrangler wrote on 2024-08-19, 18:28:
You can test a flipped CPU fan too, some ppl used to swear that often worked better.
I've tried that myself a few times as well, but was always disappointed each time.
Only CPU coolers that seem to work well with this method are ones that have high airflow and high static pressure (read: LOUD ONES.) Think: "Dustbuster" cooler on the nVidia GeForce 5900XT series.
The coolers that blow air towards the board also add the benefit of cooling other components on the motherboard (like the VRM inductors and caps), so I always prefer this more standard method.
Of course, with the limited case space & cooling of the O/P's case, flipping the fan on the CPU heatsink and using a PSU with a large 120 mm fan on the top might actually work quite well. So your suggestion is certainly worth trying. 😀
Repo Man11 wrote on 2024-08-19, 19:06:
the case you're using was (IMO) designed for Pentium 2/3 systems.
+1
Cases with PSU "on the side" (covering) the motherboard view were indeed a lot more popular during the Pentium / Pentium II / and early Pentium era. They pretty much went out of fashion after the late 90's, in part due to poor cooling, and in part due to being harder to change hardware / work on.
Probably worth mentioning that our family's Pentium II PC used this style of case too. In 2004, I got some new components to upgrade that PC... or rather a new motherboard and CPU. It was a socket A system with a Duron Applebred (T-Bred) 1400. It ran too hot for its own good, so I immediately ditched the PC case's side from day 1. That helped, but only a little bit. My problem was exactly like Mondodimotori's: the CPU heatsink fan was right up against the PSU side/bottom and starving for airflow. My initial solution was to add a 2nd 80 mm fan blowing air from the side of the motherboard and onto the CPU heatsink, which worked quite well - it lowered the load CPU temps from 55C to 43C. Later, I added another fan on the other side and that shaved off another 2C. But at that point, the PC was a complete "hackery": PC side panel permanently off, random fan(s) wired inside the case to move air across the CPU heatsink (bonus points: they were salvaged fans from PSUs, and their wires were just shoved into spare Molex connectors 🤣 ). But hey, at least it all stayed cool, even in the summer, when my room at the time was known to reach around 27-28C -with- A/C.
swaaye wrote on 2024-08-19, 20:49:
Ah I missed the part about it being with the case open and PSU not in the way. Back around 2000 I had a friend with a similar case with the PSU in the way, and he too was trying to run a Athlon 1.4. It would hit 80C and this will cause a K7 to become unstable. The Tdie max spec is probably 85-90C.
80C is wayy too much for these socket A CPUs, as is 70C. Heck, even 60C is no good.
Remember, socket A CPUs don't have built-in temperature probes.
The way most socket A motherboards measure CPU temperature is through an NTC thermistor or thermal diode in the CPU socket... and typically not touching the CPU package at all... meaning, those temperature readings are a little "passive" and quite slow. This is why a socket A CPU can die instantly without a heatsink, even if the motherboard has an over-temperature shutdown mechanism. By the time the NTC thermistor or diode under the CPU gets hot enough for the motherboard to register an overheating condition, the CPU die would be long gone and dead.
So for this reason, socket A CPUs should be kept at a lot lower temperature. IMO, 55C is about as high as I'd like to see one... and usually isn't that hard to achieve in a decently-large case with OK cooling and a good CPU heatsink.
swaaye wrote on 2024-08-19, 20:49:Frankly I would get a new case.
+1
Actually, let's be specific here: it doesn't need to be a "new" case or even a "modern" one. By the time Socket A systems rolled out, there were beige cases on the market that had slots for 2x 80 mm fans on the back or one 90/92 mm fan. Combine that with a decent PSU with a 120 mm fan, and it should be enough to keep such a system happy / cool while still keeping the retro beige looks of the early 2000's.