386SX wrote on 2022-02-02, 19:36:

If I use the Jumperfree mode for the vcore, an unsupported cpu like an XP or even the Geode, would the mainboard still read and set at least the required voltage core or it would boot in some "random" voltage? Cause I might set it manually but the lower still would be 1,1v compared to the 1,0v of the cpu, I suppose it'd not do much difference but just to know it would not boot with some 1,8v burning the cpu. 😉
Also interesting in the manual the FSB jumpers even for the 1.02 rev is said to go up to 133Mhz/33 so I might also set the correct FSB. I was expecting it from the next chipset version not this.

The manual probably is just stating some overclocking options. I never tried overclocking the A7V and just ended up putting an Athlon 1100 in there and left it at that.
I mean it ran pretty much fine for what it was, a 1000MHz Thunderbird is roughly similar in performance to a 1GHz Coppermine except that it uses like >50% more power.
1.1v should be fine. The Geode is basically just an underclocked Athlon XP (AMD wanted a CPU with as low power consumption as they could get away with based on standard sA CPUs they had available at that time). I find it extremely unlikely that 1.1v is gonna fry it.
I don't know for sure what will happen if you put that CPU in there. I've always been on the more cautious side so I'd probably not even risk it. But I'm not you, it's your stuff and you can do whatever you want with it 😜

Some update.. 😁.. situation better than expected but quite confusing.. updated to latest beta 1012 bios, some more options in the bios eventually will be helpful. Positive thing is that all the cpu unsupported I'm trying actually boot thing I wasn't expecting included the Geode! The Jumperfree mode seems to work setting the vcore (at least only that) at some theorical close value (the 1.0v of the Geode became 1,15v..). The cpuid seen is quite confused, sometimes it see them as XP sometimes as Athlon, but the problem are the multipliers.. The Geode NX 1500 become an Athlon XP 600Mhz (600Mhz for sure, I checked in recovery mode in Win cause now the previous win installation seems broken I don't know if for the bios update changing IRQ or whatever or the CPU with SSE who knows).. the Athlon XP-M 1400+ become a Athlon XP 500Mhz.. with a FSB of 100Mhz I think not much can be increased there.
I tried to set (overclock as Tetrium said) 133Mhz (also with latest bios it could go up to 200Mhz.. I wonder which revision could get there not this one I think.. 😁) but I suppose is simply too much for the board and no cpu can boot with that and force me to change cpu cause I discovered the clear cmos jumper is factory removed, shorting it seems difficult close in the above right corner on this very large PCB, removing the bios battery didn't reset the bios in minutes, so I directly changed CPU and it reset itself the FSB values.
I should go for the manual jumper mode but if the multipliers are locked.. the FSB seems to not like overclock so I can't go far. Suggestions?

Anyway positive thing beside I might have to reinstall everything if Win installation can't boot with a generic error, is that in recovery mode I checked CPUZ and other tools and it see all the features beside read both the original string but the o.s. see it as a Duron cpu. But even the 3DNow! Pro is checked and with the Geode should basically support the complete feature list but still 600Mhz feel quite low and not that power saving at all.. at the wall with the Geode NX that should ask for 9W I don't know why but the final power read 55 watts. Maybe Powernow! feature need a Win program to enable that I don't remember cause the only time I've seen it workin was with Linux and the K6-2+. I don't have much installed and mostly low power, a Vibra 128 PCI, a Radeon 7000 VE, the hard disk and dvd reader but I suppose much power goes to the mainboard power circuit and the powerful PSU but not really a modern one that might loose even 10 watts for itself.

The clear CMOS jumper is removed on A7V133 as well, or at least it was never present on any of those boards I ever seen so this is probably normal 😜
We just used a flatbed screwdriver to clear the CMOS but usually before the board was mounted inside a case. These boards are kinda large so once inside a case part of it could be obscured by the PC case's innards so it might be better to complete work on the board before mounting it inside a case for final use.

The 55W might be due to (I'm guessing here) to the CPU voltage not having been lowered? So it may be running underclocked, but not at its specified CPU voltage.
I don't remember whether the A7V worked with powernow but I never bothered with this feature. I wouldn't be surprised if the A7V doesn't support this feature though. I'd just be happy to get that board running stably 🤣!
But anyway, 55W doesn't seem like a huge problem to me (it also depends on the rest of the components and is not just the CPU when measuring from the wall outlet). I'd want to know how warm the heatsink gets after a while to make sure nothing dies too soon. But if you installed the CPU HSF properly and are using a nice beefy CPU HSF, you shouldn't have a problem with temps on what is basically an Athlon XP underclocked to 500MHz or 600MHz.

Btw I do remember these boards often wanting to set the FSB to 100MHz whenever they thought they had a failed overclock or even after we were swapping components around.

Sounds like they're booting at minimum mobile multiplier and you'll need a util to up the multiplier by software.

Power... PSUs have a kind of upturned bathtub efficiency curve that PSU manufacturers don't like showing you the bottom and top ends of, because they're horrendous. Meaning you can save all the power you like but it's gonna suck it from the wall anyway and blow it out as heat from the PSU once you go below a certain point. To save power then, you need a lower wattage PSU... that's why I don't agree with ppl saying "Get a newer PSU it's wayyyy more efficient" because a) you probably gotta go to 700W or so to get enough amps on 5V, and b) having gone to 700W your 50W draw at 70% efficiency has fell off the bottom of a 700ws curve and is now 50W at 50% efficiency or something.

Tetrium wrote on 2022-02-02, 21:24:

If with thermal runaway you mean frying because of overheating, I fried one of these chips by forgetting to add TIM to the other […]
Show full quote

Pickle wrote on 2022-02-02, 20:23:

one thing you might consider is it possible to have thermal runaway with some of these chips (I pretty sure i cooked my original Tbird 1.2 Ghz when i had the cpu fan off. Thankfully it was cheap to replace)
i think this was resolved with the XP series.

If with thermal runaway you mean frying because of overheating, I fried one of these chips by forgetting to add TIM to the otherwise properly installed HSF and the CPU fried within seconds (didn't even make it to the POST screen, it stayed black and the CPU was dead almost instantly).
Iirc the CPU I fried was a Palomino though. This was on one of the A7v133 boards which may have something to do with this but the thermal protection was still wonky from my perspective, much less reliable than for instance contemporary Pentium 4 thermal protection.
AMD added something resembling a real thermal protection only when they released Athlon 64 I think. Anyway I wouldn't take the risk with Athlon XP.

In particular never do the socket 3/4/5/7/370 trick of testing for CPU life by putting your thumb on top and flicking the power on to see if it gets to POST screen, meaning to turn it off if it gets ouchy hot... on socket A that's like striking matches to see if they work. Always fully install heatsink before applying power on these.

Thanks, yes the stability of this board would already be a success. 😁

I suspect too that the board might not support the Powernow! feature and also the Geode seems like running a bit overvolted of 0,1 volt at least. I suppose 55 watts at the end are right cause there're anyway old component inside.. the video card might ask for 20 watts I suppose looking at its thin heatsink, the hard disk is a 40GB old IDE disk and I suppose is also not very low power. Also the mainboard and the PSU as said I'm sure take a part of that value even without anything. The CPU heatsink is nothing great, more a <1Ghz Socket A style cooler, far from the usual XP 3200+ ones so it get warm even with the Geode. The XP-M 1400 TDP should be 25 watts and the final difference is more or less right compared to the 10 watts of the Geode. I also tried the XP 3000+ and it boot but again at low freq but much higher power demand, like 95-100 watts even at I don't remember which freq but 100Mhz x (locked multiplier) of the XP 3000+. But it might be at the end one of the faster. I've other ones to test, many Sempron, a XP 2000+ (AXDA2000DUT3C) then a XP 2800+, the Duron 1200 and 1300, the Athlon 1400 but at that point I think I have to search for a better heatsink.. 😉

BitWrangler wrote on 2022-02-02, 22:46:

Sounds like they're booting at minimum mobile multiplier and you'll need a util to up the multiplier by software.

Power... PSUs have a kind of upturned bathtub efficiency curve that PSU manufacturers don't like showing you the bottom and top ends of, because they're horrendous. Meaning you can save all the power you like but it's gonna suck it from the wall anyway and blow it out as heat from the PSU once you go below a certain point. To save power then, you need a lower wattage PSU... that's why I don't agree with ppl saying "Get a newer PSU it's wayyyy more efficient" because a) you probably gotta go to 700W or so to get enough amps on 5V, and b) having gone to 700W your 50W draw at 70% efficiency has fell off the bottom of a 700ws curve and is now 50W at 50% efficiency or something.

Yeah I've seen that on some modern PSU (nothing high end but enough acceptable as a 600W model) how many watts it take once the power demand is low.. like even 10 watts compared to some DC-DC PicoPSU (of course it's not fair to compare those but useful to do simple math on low power systems) with mini-itx low power boards. Also voltages goes usually where they wants with very low power. This PSU is a Prescott oriented Enermax, still working strong with a quite complex PCB and 5 volts is not a problem here (while the problem is the PSU connector lifetime, I had to fix its final wires with removable glue to the plastic cause they sometimes would easily detach from the 20 pin connector).
Not that I'm trying to build a low power Socket A config, it would be difficult beside some industrial micro atx board supporting the Geode and oriented to low power config but still just doing tests to find a balanced high end solution, not wasting power into the fastest ever that would probably break the mainboard or the PSU soon. More a fast, stable and still balanced retro system for Win 98 / ME. 😀

About the multiplier, I don't remember if they are locked shouldn't they need to be unlocked before or the software can change that anyway?

With mobile the min and max are locked but you can go anywhere in between... however, the mobile bridges were locked out on non-mobile CPUs in the Barton era and it was entirely locked, vs the mobile unlock that was possible in Tbred era, and direct multiplier bridge manipulation possible with XP and earlier.

Interesting, I have few memories of these details cause I never unlocked any and I think I had the XP 1800+ than switched to the XP 3000+@3200+ with the FSB in those good times using the Radeon 9500 Pro L modded version.. 😁

Anyway I'll try the others cpu I have to find which gives the better options. I don't know if the 133Mhz limit might impact more on the FSB limits or the RAM maybe supported only at lower freq beside they are PC133 for sure and time correct around 2001/2002 modules.
The Geode would be interesting @ 1Ghz anyway I suppose really taking the best sides, low power and all the cpu features, SSE included. I wonder how some sw like Windvd or Powerdvd would work supporting all those features at the same time, they could be a good bench to test this cpu compared to the original Athlon 1000 with only the MMX and the basic 3DNow!.

BitWrangler wrote on 2022-02-02, 22:46:

Tetrium wrote on 2022-02-02, 21:24:

If with thermal runaway you mean frying because of overheating, I fried one of these chips by forgetting to add TIM to the otherwise properly installed HSF and the CPU fried within seconds (didn't even make it to the POST screen, it stayed black and the CPU was dead almost instantly).
Iirc the CPU I fried was a Palomino though. This was on one of the A7v133 boards which may have something to do with this but the thermal protection was still wonky from my perspective, much less reliable than for instance contemporary Pentium 4 thermal protection.
AMD added something resembling a real thermal protection only when they released Athlon 64 I think. Anyway I wouldn't take the risk with Athlon XP.

In particular never do the socket 3/4/5/7/370 trick of testing for CPU life by putting your thumb on top and flicking the power on to see if it gets to POST screen, meaning to turn it off if it gets ouchy hot... on socket A that's like striking matches to see if they work. Always fully install heatsink before applying power on these.

I never test using that method anyway, for me seeing the POST screen appear is enough test for to see the CPU working.
And indeed, with the s370 and earlier I did get away with sometimes installing only a heatsink for a quick testing, but with Athlon and Athlon XP I would never risk it.

One funny thing btw is that I tend to use less sought after parts for any initial testing and for sA this was a Thunderbird 800 or 900. And this little chip did manage to survive several quick posts using only a heatsink (properly installed but without TIM) and even a few accidental overclocks (FSB 133MHz instead of FSB 100MHz by accident) and survived through that all ^^
Mind you, all of these tests were always just long enough to either see something appear on screen or to hear the beep tones, then cut off power by using a mechanical switch (so disabling the power completely to the whole assembly and not just cutting power from the back of the PSU).

Usually I'd use some old TIM for testing btw which is absolutely better than not using any TIM at all but this stuff was the really cheap stuff, so excellent for quick testing 😜

Interesting I was reading the specs of the various Athlon XP I have and I found an XP 2600 having Thoroughbred (Model 8 ) core with 16x multiplier that is seen correctly by the bios cpuid as Athlon(TM) XP 2600+ (why this is read ok the others not I don't know..) and in the bios there's a Operating system freq of 800Mhz that I don't know how it calculate that value but in Win the o.s. see it as Duron cpu but CPUZ report it as Athlon XP correctly operating at 1614Mhz FSB 100 x 16. Power demand seems high almost 100 watts. The bios seems to understand it's an XP cpu but Win98 even with patches seems to see it as a Duron. While it's working and fast for sure.. 1600Mhz is quite high but @ 0,13um. CPU clock is real, cpu benchmarks is higher than ever but I wonder if it might be stable at such freq.

Another cpu detected by the bios cpuid correctly the Athlon XP 1700+ (100Mhz x 11 better than nothing).

Let's reinstall the broken Win installation so I might test every cpu without using recovery mode. I'll let the XP 1700+@1100Mhz for now and later testing the others. I don't know how the Sempron cpu are seen eventually. Not that change much, power demand seems mostly the same. I should find a situation where cpuid is read correctly by bios and the os, freq stays in the original 1000 to 1400 one and having a less power demanding solution with the new cpu features. Like taking the best sides of these cpu without necessary only the fastest.

It's seems to be not as easy as I hoped.. I see unstable insallations before, during or after the Win 98 setup process with error protections or driver installation blue screen (with no PCI cards). I suppose are these cpu but still I see those even going back to the Athlon 1000. I'm flashing the previous latest 1011 bios and restart.

386SX wrote on 2022-02-03, 13:23:

It's seems to be not as easy as I hoped.. I see unstable insallations before, during or after the Win 98 setup process with error protections or driver installation blue screen (with no PCI cards). I suppose are these cpu but still I see those even going back to the Athlon 1000. I'm flashing the previous latest 1011 bios and restart.

Btw, I don't want to come across as rude or something, but it would help if you would edit your replies more often instead of adding a few lines of new text with each consecutive reply.

I wonder if a recap would help with stability on this board if you can't get a Windows OS stable with a 100% compatible CPU.

ODwilly wrote on 2022-02-03, 13:41:

I wonder if a recap would help with stability on this board if you can't get a Windows OS stable with a 100% compatible CPU.

Back when I was working a lot with these boards, it didn't seem to be the caps (these boards were way younger then they are now).
These boards were finicky right out of the box, but still it might not even be a bad idea!

No problems in editing the posts, I usually edit updating them but sometimes get faster to write in the Quick Reply. 😉

I think it's not a capacitors problem (even if I'd not be surprised), at least it wasn't before trying the unsupported cpus, now seems like the system has difficult time reinstalling the o.s. even with previous bios. I've noticed that disabling the Promise Fastrack100 IDE controller crash the installation process beside it seems Win ME hang during the Plug&Play initial installation and reboot itself. I switch back to install Win 98 first edition patching it later. Is it better the leave Plug&Play "yes" or "no" in the bios for Win 9X and ME?
It's strange now I got these problems before with the default bios/cpu I had not. I might try previous bios versions too who knows, maybe not everything get always fixed as expected.

386SX wrote on 2022-02-03, 14:19:

No problems in editing the posts, I usually edit updating them but sometimes get faster to write in the Quick Reply. 😉

I think it's not a capacitors problem (even if I'd not be surprised), at least it wasn't before trying the unsupported cpus, now seems like the system has difficult time reinstalling the o.s. even with previous bios. I've noticed that disabling the Promise Fastrack100 IDE controller crash the installation process beside it seems Win ME hang during the Plug&Play initial installation and reboot itself. I switch back to install Win 98 first edition patching it later. Is it better the leave Plug&Play "yes" or "no" in the bios for Win 9X and ME?
It's strange now I got these problems before with the default bios/cpu I had not. I might try previous bios versions too who knows, maybe not everything get always fixed as expected.

I always ended up ignoring the Promise controller altogether, basically pretending it isn't there.
I don't remember if I disabled it in BIOS but it would be my first thing to try.
I installed ME on several A7V133 systems and also on at least one single A7v (but probably different revisions) so it should definitely be doable.
I'm unsure of what BIOS version but it was usually the last available one.

However, installing ME is something new in itself in some way. I even wrote something of a guide about it but it essentially goes like this:
When wanting to install ME on a new ('new', hehe) system, I always formatted the drive in at least 2 partitions with the 1st partition for just ME (minimum 1 gb but it didn't have to be really massive unless I really needed to install a lot of junk on there and left all my junk files on the desktop which is really just the C drive). I used 1 stick of RAM (one stick only!) and only a PCI Virge and no other cards (no sound and no NIC, only VGA CPU and RAM and of course HDD, FDD and ODD). Disabled all unnecessary onboard stuff I was never gonna use (this includes any extra IDE stuff and sometimes but not always onboard sound but do disable onboard sound if you intend to use a dedicated sound card).
Install ME fully from disk (or harddrive, whatever floats your boat) until installation is finished.
After I'm on the desktop I sometimes would copy all needed files over like drivers.
Then I shut windows down, unplug anything and remove the Virge and add/swap memory modules till the desired configuration has been achieved.
If I feel uncomfortable about something I'd literally insert the expansion elements one at the time, completely unplugging the system each and every time. I'd then boot her up again, install the driver and if succesful shut down the system, unplug and add the next component, boot up again, install another driver, rinse and repeat till all drivers are installed.
Just keep installing components until the desired end-configuration has been achieved. Tweak windows and install the drivers and once these have been done do some stability tests like benchmarks or play some more demanding games on it for a couple hours.

Before installing, set up the BIOS. Only change the BIOS again after the desktop has been reached and you want to start adding components.

Sure it takes a bit longer perhaps, but it's so much easier this way than having to try and fix spooky odd bugs later.

the promise controller can be disabled from the bios. I am only running a cf adapter and dvd drive and the other controller was also UDMA 100 capable, plus the promise detection slows down the bootup. I think the only reason it really would be needed is for a RAID configuration.

Thanks for the Win ME "guide" 😉 , usually I had not many problems with ME only some in few mainboards but depending really on chipset, configurations, components.. but at the early installation process I wasn't used to it (maybe with some difficult vga) more after installation was at least completed. Anyway I've flashed the 1004C version I've found and same version I had before. Now I start from a S3 Virge/DX PCI and a single module of ram. Win 98 and then upgrading step by step.

On the Promise controller even disabling it in the bios, it seems like Win installation sort of crashed until I enabled it again. So it might have some resources still detected by the o.s. who knows. (I think in the old bios it wasn't possible to disable it like the AGP speed choice, but I might be wrong)

EDIT: interesting, the combination to old 1004C bios, plus the S3 Virge DX but changing PCI slot seems have actually helped. Also not having the floppy drive (is a new atx case) I wasn't sure if disabling it in the bios or leaving it enabled was actually helping or not in the newer bios version. With this one seems to actually help disabling it in the bios. Also I changed ram modules.. now a single one double sides 256MB PC100. It's the first mainboard I've seen that get better with an older bios version.. 😁 Also the unsupported cpus might have had something to do with this but at the end maybe some of them might actually work, at least the Duron with SSE to not risk asking too much from this board as compatibility.
Now I'm into Win 98 after installation..deciding which vga I should go for... a time correct one or risking to try something more powerful like a 9500 Pro 64MB? Or before the vga I should better choose the final cpu to use.. The Athlon 1000 at the end works ok. From what I've read on the driver page of the Promise site, it seems like updating the controller bios might need a o.s. reinstallation.. I suppose the controller bios is included inside the bios update? Cause I think I've seen the controller date at boot changing from 2000 to 2001 when upgraded to 1011/1012 bios. Controller is at boot "Ultra100 2.01 Build 28" now.

EDIT2: also lot of IRQ / resources not-free problems.. seems like some bios updates solved something of that but before updating I'm trying to change PCI slot or sound cards to find one that doesn't require too many things. Still any combinations of vga and a single pci card can get different results.. this is a benchmark for the patience. I suppose many people weren't exactly happy with this wonderful compatibility.. I don't remember but sure I didn't have this in the 2000 probably still the Super Socket 7.