Archer57 wrote on 2025-08-21, 13:11:

That's why i dislike hot chipsets, yes. Also AFAIK all NF4 are defective, but i may be wrong - there is no definitive information on what's defective and what is not.

I don't know where you are getting this from? But if it's still running great after 20 years its obviously not defective.

Archer57 wrote on 2025-08-21, 13:11:

Yes, on modern hardware, not sure/have not tested how it'll work on this old stuff given CPU and other parts of the system do play a role here. It is also completely irrelevant for a storage drive.

My A8N SLI deluxe isn't that new, in fact it predates yours by almost a year. Same chipset minus the southbridge. I attached a screenshot of crystaldiskmark with the pcie 1.0 @ 1x drive benchmark.

Archer57 wrote on 2025-08-21, 13:11:

Well, i'll test what the temperatures are on specific system. So far from what i've seen in this case airflow is good enough for things to not become too hot as long as no 500W GPUs are used.

Also number of fans + meshes does not equal good cooling. Often the opposite. The airflow has to work in a sensible way, like front intake back exhaust. It does in this case. When cases are made of mesh almost entirely with stupid number of fans - there is no directional airflow at all - just a bunch of fans blowing in random directions.

This board will deteriorate though, nothing i can do about that. It gets too hot to hold just laying around on a table...

Just get a slight breeze in the case, I made my nForce4 chipset with a Zalman ZM-NB47J heatsink go from barely touchable for 5 seconds to pleasantly lukewarm with just one of those cheap aliexpress pci bracket coolers running at 5v.

Archer57 wrote on 2025-08-21, 13:11:

I'll probably find something different eventually. Perhaps something based on K8T890. I prefer simpler boards anyway and this one is more suitable for a bench, fooling around with SLI and everything, than for a permanent build.

That's a good board yeah.

Raldsen wrote on 2025-08-21, 15:27:

I don't know where you are getting this from? But if it's still running great after 20 years its obviously not defective.

Well, i got 3 8800GTs for this build which are absolutely 100% defective and are still working after all the years also. As well as some AGP stuff like 7600GT and 7800GS. Bumpgate is not a death sentence, if conditions are right affected hardware can last.

Does not matter ultimately, if i am wrong and it is not defective - that's great.

Raldsen wrote on 2025-08-21, 15:27:

My A8N SLI deluxe isn't that new, in fact it predates yours by almost a year. Same chipset minus the southbridge. I attached a screenshot of crystaldiskmark with the pcie 1.0 @ 1x drive benchmark.

Just out of curiosity i've popped in JMB585 + (slightly full) 860pro and it works like this:

35MB/S random 4k are not impressive for pci-e at all...

But yeah, this only shows that free pci-e x4 can be useful. Hmm, may be i should leave this card in here instead of using buggy onboard SATA which works as SATA1 only with modern drives...

So i've ran 3dmark05 game benchmarks with "repeat each test 5 times", got this temperatures (i am using this tool in aida simply to get statistics, did not run tests themselves):

I do not see anything horrible here, judging by HDD temperature inside the case does increase by a few C, but that's to be expected and i'll leave it like this. Also i did not notice significant/noticeable performance drop with cool&quiet enabled, it does drop both frequency and voltage so i'll leave it on for now.

How long is 5x in 3d mark 05? I wonder if something like 30 minutes is sufficiently relevant. The case must be closed as it would normally be, in the location where it is supposed to be. The worry is basically in warm days like when ambient temperature reaches 26-28'C and the system is running for at least an hour it could start to overheat. Probably more relevant temperature is motherboard rather than hard drive. A hard drive can be easily replaced although the data will be lost.

AlexZ wrote on 2025-08-21, 17:38:

How long is 5x in 3d mark 05? I wonder if something like 30 minutes is sufficiently relevant. The case must be closed as it would normally be, in the location where it is supposed to be. The worry is basically in warm days like when ambient temperature reaches 26-28'C and the system is running for at least an hour it could start to overheat. Probably more relevant temperature is motherboard rather than hard drive. A hard drive can be easily replaced although the data will be lost.

Yeah, it'll be around 30 minutes or so. But it is abnormally high load, would not be like that in actual games. May be a longer run would be more precise, but honestly - i am not a fan of stress tests, i'll try to run aida in background when i actually play something and see how it goes.

Ambient is around 26.

The reason i am looking at HDD - it is not used/does not generate much heat (this is secondary storage, OS is on SSD, and it is 2.5 inch HDD) and is useful as "ambient inside the case" sensor. Not because i am worried about its longevity. "Motherboard" sensor... not sure what it is measuring but it constantly stays between 40 and 50 regardless of extra cooling, motherboard laying on a table or inside the case, etc. Looks like unreliable data.

Videocard also has "board" and "ambient" sensors, but those very obviously measure board temperature, when the board heats up they both go up even in the open.

Temperatures look good. That motherboard sensor could be chipset temperature, VRM or somewhere close to pcie slot. You would have to check value just after booting.

My Athlon 64 3400+ is in case Zalman Z9 Plus. After 10x run of 3d mark 2005 (about 1.5h to complete), with 27.5'C ambient temperature, the maximum temperatures were:
- CPU - 50'C (38-41'C when idle, without cool&quiet, 1300 rpm)
- GPU - 64'C (47'C when idle, 1250 rpm at 35%, custom fan profile set in BIOS)
- Samsung 860 EVO - 28'C
- Western Digital 1TB magnetic HDD - 42'C (located in internal aluminum HDD box with little cooling fan removed).

Archer57 wrote on 2025-08-15, 18:08:

By the way i absolutely love the place they've chosen for that red sata port...

that is Aopen Ax59pro FDD port level of dumb

AlexZ wrote on 2025-08-21, 19:39:

Temperatures look good. That motherboard sensor could be chipset temperature, VRM or somewhere close to pcie slot. You would hav […]
Show full quote

Temperatures look good. That motherboard sensor could be chipset temperature, VRM or somewhere close to pcie slot. You would have to check value just after booting.

My Athlon 64 3400+ is in case Zalman Z9 Plus. After 10x run of 3d mark 2005 (about 1.5h to complete), with 27.5'C ambient temperature, the maximum temperatures were:
- CPU - 50'C (38-41'C when idle, without cool&quiet, 1300 rpm)
- GPU - 64'C (47'C when idle, 1250 rpm at 35%, custom fan profile set in BIOS)
- Samsung 860 EVO - 28'C
- Western Digital 1TB magnetic HDD - 42'C (located in internal aluminum HDD box with little cooling fan removed).

What's curious - i've started aida right after boot, yet it shows minimum of 45C... not sure where the sensor is, but it is weird.

But yeah, temperatures look fine to me so far, of course i'll have to see how it works over time, but i suspect it'll be fine. This system will likely use <200W under load which is nothing by modern standards and that 120MM fan does have much better performance than typical optional 80MM fans in beige boxes from late 90s early 00s many of this systems would have ended up in.

Sadly i can not see SSD temperature because my aliexpress special SSD has fixed 40C readout.

Z9+ is a good case, i would not hesitate using it for a modern system with 1000W PSU "recommended", in fact it is probably better than many modern cases. It is large though, as many of those cases are. I want small and stackable, which is one reason for my choice...

Living wrote on 2025-08-22, 02:43:

that is Aopen Ax59pro FDD port level of dumb

That one is even more crazy 😁

Many weird decision on those older boards, especially during transition from AT to ATX, but later too. Even modern boards still have annoying layout issues from time to time, but nothing so crazy.

At least i can kind on see why asus did this - they wanted esata on the back io panel and probably did not want to route the traces through whole board, so they placed the chip along with the second port next to the back panel...

Living wrote on 2025-08-22, 02:43:

Archer57 wrote on 2025-08-15, 18:08:

By the way i absolutely love the place they've chosen for that red sata port...

that is Aopen Ax59pro FDD port level of dumb

totally off topic
Hah! Right now I'm conducting some benchmarks with this board. I disabled the floppy, screw that.

Archer57 wrote on 2025-08-21, 15:49:

35MB/S random 4k are not impressive for pci-e at all...

It's very typical for all SSD's, even the average unraided consumer NVMe's at PCIe 4.0 or even 5.0 only usually does about double that. Even your SATA3 SSD only did 13MB/s more. I did a benchmark on my MSI K9A2 Platinum system and compare the SATA2 speeds with my previous benchmark. Now you see that a NVMe even at PCIe 1.0@1x is quite a bit snappier than a SATA2 SSD.

Raldsen wrote on Yesterday, 11:18:

It's very typical for all SSD's, even the average unraided consumer NVMe's at PCIe 4.0 or even 5.0 only usually does about double that. Even your SATA3 SSD only did 13MB/s more. I did a benchmark on my MSI K9A2 Platinum system and compare the SATA2 speeds with my previous benchmark. Now you see that a NVMe even at PCIe 1.0@1x is quite a bit snappier than a SATA2 SSD.

Yeah, and that's sad. All this SSDs are marketed as "better, faster" but in reality they are progressing backwards. Many NVME SSDs, even 4.0 or 5.0 can be slower than good SATA SSDs, especially if the load is heavy enough. Ultimately performance depends on SSD itself (controller, flash, etc) and PC, not the interface.

And that's easily visible by comparing your pci-e x4 results to my sata3. Though 860pro probably is the fastest sata ssd ever made, and sadly they are not sold anymore. That's why i am not a fan of buying cheap nvme stuff - it is not practically any better.

I generally expect at least something like this from NVME, this is, at this point, old 3.0 x4 SSD on LGA2011 workstation:

You are making me curious though, i'll probably try something decent on this hardware and see how well it works.

So, here we go...

Being crazy person that i am i decided that after all the time spent making this stuff work properly second system deserves being built too:

I have all the hardware anyway (and still have a few of this case + PSU kits), even if i am not going to use both systems at the same time storing it like this, in functional condition, makes sense.

I do not have second 4800+ CPU, so this one is 4200+ Manchester instead, again 2 gigs of RAM (hmm, i am running out of suitable 1GB sticks...), HD3850 AGP and audigy4. Along with SF-2281 SSD i spent a couple of evenings rebuilding firmware for, which can be switched to sata1 mode and works with VT8237R perfectly.

The only complication - i do not have anything to use for storage, ancient 40GB IDE drives make no sense here and newer sata2-3 stuff does not work.

I'll post aida screenshots for both systems a bit later...

Also one more bug i discovered in this hardware while building it - if i put soundcard in the last slot weird stuff happens - the sound does not work properly and one CPU core is constantly at 100% load. This stuff is fun...

If you have a spare case then why not. Athlon 64 X2 4200+ is a good basis for comparison with Athlon XP 3200+, same clock speed, same cache. Maybe you can disable one core somehow for exact comparison. It's probably best to delegate Athlon XP to Windows 98 and have this for Windows XP. It should be ok for any Windows XP era game (2002-2006) as long as you don't need 1920x1080. 1920x1080 was historically Windows Vista/7 era.

Probably the last pci slot is damaged. You can try putting a video card there to find out.

I'm waiting with my AM2+ build for Zalman Z9 U3 case. If I don't get it within 2 weeks I will probably use Cooler Master N500 that I already have (https://www.newegg.com/cooler-master-atx-mid- … N82E16811119280). Zalman Z9 is a sturdier case than N500. Historically, these cases were used for FX Vishera. Zalman Z9 U3 has a floppy slot and GPU length limit 290mm so it's more suitable for AM2+. N500 could be used for modern PC. PNY GTX 780 OC is unfortunately 296 mm long and I will have to cut off 6mm from the card to fit in. It is possible without damaging the card but I'm not sure if I want to make such a destructive modification. Alternatively, I could fall back to Gigabyte Windforce GTX 770 or MSI Gaming G4 980 (non Ti, which should be about as fast as PNY, I got it specifically because it is shorter). There is also GeForce GTX 680 on sale locally but the owner wants $57 which is too much and it's a founders edition which is not a great choice as they run hot.

AlexZ wrote on Yesterday, 19:42:

If you have a spare case then why not.

Yeah, i've bought 5 of this cases and power supplies some time ago, there were some crazy discounts and they ended up ~$50 for a case+PSU. This cases are not adequate for really hot hardware, but 200-300W are totally fine, which works for a lot of stuff. Turns out a single 120mm exhaust fan + some holes in front of the case can make a massive difference and beat vast majority of period correct cases.

AlexZ wrote on Yesterday, 19:42:

Athlon 64 X2 4200+ is a good basis for comparison with Athlon XP 3200+, same clock speed, same cache. Maybe you can disable one core somehow for exact comparison. It's probably best to delegate Athlon XP to Windows 98 and have this for Windows XP. It should be ok for any Windows XP era game (2002-2006) as long as you don't need 1920x1080. 1920x1080 was historically Windows Vista/7 era.

Hmm, interesting. May be i'll play around with it. So far i've compared 4800+ with 4200+ and even if 4200+ is overclocked to the same ~2.4Ghz (which is really easy without touching the voltage and may be i'll do it) there is still ~10% difference in stuff like 3dmark.

I still need to compile and post the benchmarks i've ran on this board with 4800+ and some AGP cards, some of those are quite interesting even though i already see i did not do everything exactly right and i am certainly not going to redo it.

AlexZ wrote on Yesterday, 19:42:

Probably the last pci slot is damaged. You can try putting a video card there to find out.

Perhaps. It is visually fine though and the card is detected, drivers work, it just causes abnormal CPU load once the drivers are installed and sound is glitchy. Feels more like incompatibility or resource conflict of some sort.

There is also a mention of post cards not working in this slot on the retroweb.

AlexZ wrote on Yesterday, 19:42:

I'm waiting with my AM2+ build for Zalman Z9 U3 case. If I don't get it within 2 weeks I will probably use Cooler Master N500 that I already have (https://www.newegg.com/cooler-master-atx-mid- … N82E16811119280). Zalman Z9 is a sturdier case than N500. Historically, these cases were used for FX Vishera. Zalman Z9 U3 has a floppy slot and GPU length limit 290mm so it's more suitable for AM2+. N500 could be used for modern PC. PNY GTX 780 OC is unfortunately 296 mm long and I will have to cut off 6mm from the card to fit in. It is possible without damaging the card but I'm not sure if I want to make such a destructive modification. Alternatively, I could fall back to Gigabyte Windforce GTX 770 or MSI Gaming G4 980 (non Ti, which should be about as fast as PNY, I got it specifically because it is shorter). There is also GeForce GTX 680 on sale locally but the owner wants $57 which is too much and it's a founders edition which is not a great choice as they run hot.

Yeah, for that kind of cards the choice of cases makes sense. Are not this extreme overkill for the platform though?

Funnily enough for the cases i am using manufacturer claims up to 320mm videocard and 330mm total length (no drive bays...), i have not measured it though and even i would not put GTX780 or something into one, unless the card itself exhausts hot air out of the case.

Also cutting the card does not sound like a good idea IMO. I'd be more likely to rip out the drive cage from the case if i had to do something like this...

I've also been trying to find a cheap S939 CPU, like 3200+-3500+ venice or something, purely for testing purposes and may be to compare with S462 (i have the second board like this) but so far - no luck. Seems like nobody is selling them at all. I've found 4000+ san diego for ~$50, but not going to grab that as i do not really have a use for it...

Archer57 wrote on Yesterday, 12:40:

And that's easily visible by comparing your pci-e x4 results to my sata3. Though 860pro probably is the fastest sata ssd ever made, and sadly they are not sold anymore. That's why i am not a fan of buying cheap nvme stuff - it is not practically any better.

It's very impressive for a consumer SSD, better than the 970 EVO I used in that PCIe2.0 x4 benchmark.

Archer57 wrote on Yesterday, 12:40:

I generally expect at least something like this from NVME, this is, at this point, old 3.0 x4 SSD on LGA2011 workstation:

The attachment nvme.PNG is no longer available

Me too 🤣, that's gotta be something enterprise grade like an Optane.

But if you want less buggy SATA, why not ditch both Nvidia and VIA for 939 and go for something with an ULi or ATi southbridge. Like the MSI RX480 Neo2-F or MSI RD480 Neo2. I got an ASRock 939Dual-SATA2 which has been spotless with SATA SSD's, but very bad caps, like a bad cap festival with every shit brand present.