I don't know about the POD, but the IntelDX4 100 MHz will be faster than a AMD 5x86-133 clocked at 100 MHz. But the difference is very small.

I've only ever seen a single Intel DX4 CPU, and that was back in the day. It was very fast compared to other 486 machines, but then again, I'm talking about a high end IBM Aptiva system. The only AMD based 486 / 586 machines I've seen are boutique builds with cheap motherboards, no WB cache support, no EDO support and poor PCI performance.

I too am interested in a performance comparisson between them - especially considering I've never seen / used a socket 3 pentium overdrive (back here, it was cheaper to swap your MB + CPU to a socket 7 pentium then but a POD - especially since boutique builders would late you trade in for newer parts).

hello
i am also interested in dx4-100 performance comparision between amd enhanced486(16k wb), amd 8kb wt and intel 16k wb, as 3x clock ones are more dependent on caches than 2x ones.
i doubt if amd had any 16k wt 486s, i think they had only 8kb wt, 8kb wb and 16k wb.

jesolo wrote:

I was wondering that, if I wanted to upgrade the AMD CPU to an Intel 486 DX4-100 (also the write through model) or an Intel Pentium Overdrive 83 MHz (in write through mode), which one of the two would be best upgrade?

if you want the best performance then just go ahead with pod83, its faster than any true 486s. even amd486dx5-133 oc 150or160 is only comparable to pentium-60or66.
however if you want a pure 486 and not a pentium under the hood, go with dx4-100 or dx5-133.

noshutdown wrote:

jesolo wrote:

I was wondering that, if I wanted to upgrade the AMD CPU to an Intel 486 DX4-100 (also the write through model) or an Intel Pentium Overdrive 83 MHz (in write through mode), which one of the two would be best upgrade?

if you want the best performance then just go ahead with pod83, its faster than any true 486s. even amd486dx5-133 oc 150or160 is only comparable to pentium-60or66.
however if you want a pure 486 and not a pentium under the hood, go with dx4-100 or dx5-133.

I don't know about that. I was just doing benchmarks and the amd OC to 160mhz beat a POD 83mhz in everything but quake.

would the performance of a Cyrix 5x86 100mhz without any of the register enhancements on be about even with a Intel 100mhz DX4?

soviet conscript wrote:

noshutdown wrote:

jesolo wrote:

I was wondering that, if I wanted to upgrade the AMD CPU to an Intel 486 DX4-100 (also the write through model) or an Intel Pentium Overdrive 83 MHz (in write through mode), which one of the two would be best upgrade?

if you want the best performance then just go ahead with pod83, its faster than any true 486s. even amd486dx5-133 oc 150or160 is only comparable to pentium-60or66.
however if you want a pure 486 and not a pentium under the hood, go with dx4-100 or dx5-133.

I don't know about that. I was just doing benchmarks and the amd OC to 160mhz beat a POD 83mhz in everything but quake.

if you're gonna compare OC to regular stock speed then you might as well OC the POD to 100mhz. Result will look differently then 😀

Otherwise I agree with the pure 486 statement. Depends what you want, the POD83 is the fastest of the CPU's you listed.

I have the VL/I-486SV2G and with the POD83 it is still topping the charts in Phil's VGA benchmark, both in regular stock speed and overclocked mode.

And I'm still trying to figure out what you did with that, because I always hit a bottleneck with the BUS and RAM leading to such speeds being unattainable on a vastly superior motherboard... Though I can get within a few percent so it's probably within the margins.

POD is indeed likely to be the fastest overall when used with the correct hardware.

Edit: Just to clarify, because I'm probably seen as "the resident dick head" around here and that might come off as such. I'm not bashing you, I'm actually applauding you... Don't get too used to that spot though, I might just have to do something about it 😉 I liked being at the top of that chart and I think I have a few aces left up my sleeve.

Is OCing a POD to 100mhz generally OK? I know with a amd 5x86 you can do it rather casually with not many worries but with the POD I read it's a bit risky so I never bothered. Also the reason I mentioned the OC other then it can done routinely is it was stated a OC AMD at 160mhz is slower then a POD at 83mhz, which unless I'm mistaken and can't read benchmarks is not true except for floating point stuff. @HighTreason if it makes you feel any better your certainly not the first name that comes to mind when think of the resident "dick head".

These days, I'm actually wary in overclocking my retro hardware.
Since these retro parts are becoming more scarce by the day, I would like to keep my hardware in working condition for as long as possible.
Overclocking my CPU and other components just tells me that it could reduce the life span of my hardware. The last thing I want is to blow or fry something on my motherboard and then not being able to replace those parts.

Coming back to my original post.
What I'm trying to achieve is to build up the fastest 486 class machine that also underclocks very well (via the turbo switch - "de-turbo").
Some people might ask why but, this is purely for interest sake to see whether I can play most of my DOS games on one PC and not have to build up another fast 386 or slower 486 PC (486 DX-33).

So far, based on my own tests and only using the turbo switch, I've confirmed that AMD 486 CPU's with L1 write through cache underclocks much better than AMD 486 CPU's with L1 write back cache - the latter underclocks with a very slight margin (it's hardly noticeable on some benchmark tests).
I've tested this on an AMD 5x86-133 (in write back mode) & AMD 486DX4-120 (write back) - These were performed on an Asus PVI-486SP3 & Gigabyte 5486AL respectively.

So far, the "best" CPU I've come across that underclocks very well is the AMD 486 DX4-100 (write through) CPU. The AMD 5x86-133 (in write through mode) also underclocks very well but, in turbo mode, it's actually slower than my AMD DX4-100.

I'm not sure how Intel's DX4-100 (write through & write back) or POD-83 CPU's behave since I've don't have any to test (yet).

In my tests the POD often won't de-turbo successfully. Though that may simply be the boards I have tried it in, I never tested too much for that feature anyway as I never intended to use it with that CPU. I use the U5S for de-turbo when I want a super-slow 486.

soviet conscript wrote:

Is OCing a POD to 100mhz generally OK?

Some have done it successfully on here, but I've not been able to do it without issue. It seems to run synthetic benchmarks just fine, as well as the DOOM time demo. However, I cannot get the Quake benchmark to run at all - it just bombs out. This has been the experience shared by others who have tried to overclock this one as well - even in a higher-end board like the MB-8433UUD.

That being said, I know Vetz has one processor that WILL run stable at 100Mhz while another does not.
Re: Artex's Build of the Week: Intel DX4 100 Write Back Enhanced - the Classic!

HighTreason wrote:

And I'm still trying to figure out what you did with that, because I always hit a bottleneck with the BUS and RAM leading to such speeds being unattainable on a vastly superior motherboard... Though I can get within a few percent so it's probably within the margins.

POD is indeed likely to be the fastest overall when used with the correct hardware.

Edit: Just to clarify, because I'm probably seen as "the resident dick head" around here and that might come off as such. I'm not bashing you, I'm actually applauding you... Don't get too used to that spot though, I might just have to do something about it 😉 I liked being at the top of that chart and I think I have a few aces left up my sleeve.

I have problem understanding why I got so good results as well. My guess is that the VLB architecture allows for better timings than using a 486 PCI motherboard. Yes, the VLB graphics card won't be as fast as a newer PCI, so it has a disadvantage there.

HighTreason wrote:

In my tests the POD often won't de-turbo successfully. Though that may simply be the boards I have tried it in, I never tested too much for that feature anyway as I never intended to use it with that CPU. I use the U5S for de-turbo when I want a super-slow 486.

Noticed the same here. It's either unstable or produces very little slowdown. Actually gonna switch out the system to a case without turbo button because of this.

Artex wrote:

That being said, I know Vetz has one processor that WILL run stable at 100Mhz while another does not.
viewtopic.php?f=25&t=41804&p=395449&hilit=Overdrive#p395453

That is correct, but I don't think it's dangerous to do. You don't increase the voltage to the CPU and with the fan it doesnt run very hot at 100mhz.

Yes, I noted an increase when using VLB on my board too, though employing certain tweaks to the BIOS closes the gap on the PCI a little.

As for the turbo switch, how about connecting it to the BUS COCK pins if possible? My board has a row of three-pin headers which proved ideal as I can switch between 83MHz and 100MHz as desired when the system is not powered on, having a three digit LED display helped in my case, I just couldn't resist. May just be my boards, but none of them have the actual turbo switch do anything when the POD is installed, only the keyboard shortcut seems to have any effect.

Just overclocking without altering voltages is generally considered OK, though I have often questioned the effect on gates in the chip being triggered more often, I doubt it has much relevance. Every once in a while I still see marked-up systems (Such as a DX4-100 that has been operating at 120MHz since it's inception in the mid-1990's) which still seem to work fine, one didn't even have a heatsink. Old CPUs are generally tough anyway and I suspect it would become unstable at the higher speed long before you actually killed it if anything actually did happen.

Yes, I must admit that overclocking some CPU's shouldn't cause any harm.
A good example (based on some posts I've read) is your AMD 5x86-133, with later models actually being quite capable of running at 40 Mhz x 4 = 160 MHz (apparently, they were unoffiicially rated to run at those speeds).
The trick is just to make sure that you don't stress out your other hardware components (like your PCI graphics card that might run on your overclocked PCI bus).

Anyone ever had any luck in underclocking an Intel 486 DX4-100 (both write back and write through) via the turbo switch?

Speedsys/3DBench/PCPBench/Doom/Quake

1AMD DX4-100 V8T :   37.09	67.1	15.9	39.81	9.7
2Intel DX4-100 WT :  41.88	71.9	19.8	42.68	10.4
3POD83 WT :          ???     68.5	18.6	41.73	16.1

Phil's test suite, MB 1433/50 UIV, ET4000/W32p.

jesolo wrote:

So far, based on my own tests and only using the turbo switch, I've confirmed that AMD 486 CPU's with L1 write through cache underclocks much better than AMD 486 CPU's with L1 write back cache - the latter underclocks with a very slight margin (it's hardly noticeable on some benchmark tests).
I've tested this on an AMD 5x86-133 (in write back mode) & AMD 486DX4-120 (write back) - These were performed on an Asus PVI-486SP3 & Gigabyte 5486AL respectively.

There are definitely some issues with write-back and the turbo function getting into each other's way. When I switch my Am5x86 to write-back on my UMC 8498F based board, turbo just halts the system until toggled back.

Some interesting findings I've made with my AMD 5x86-133 (in write back mode) on an ASUS PVI-486SP3 motherboard with a PCI graphics card.
In its "standard" (turbo switch on) mode, it performs as expected, based on benchmark tests performed with Norton Sysinfo 8.0, Landmark v2.0, and NSSI 0.60.
As stated previously, depressing the turbo switch button (i.e., running it in "de-turbo" mode) hardly shows any difference on the CPU (integer) benchmark tests (although, on Landmark v2.0, the graphics performance do drop significantly).

However, I decided to perform some in game benchmark tests as well (Doom timedemo & Wing Commander 1).
Interestingly, in "de-turbo" mode, these games do actually show a significant decrease in performance, despite the above mentioned benchmark tests not reflecting any significant change in CPU speed.

Where I get around between 35 fps & 40 fps in various Doom timedemo's, the performance drops to around 15 fps in "de-turbo" mode.
Although I cannot perform any benchmark results with Wing Commander, I could clearly see that the intro sequences were much slower in "de-turbo" mode (but still slightly faster than the speed on a 386DX-33).
Subsequent tests with Pcpbench & 3Dbench also confirmed a drop in graphics performance.

Based on the above results, I'm estimating the "de-turbo" speed of this CPU to be more or less equivalent to the CPU speed of an AMD 486DX-40 (pure integer performance, not necessarily "real world" performance).

Don't quite understand why this is but, I presume it has something to do with when your CPU's L1 cache is in write back mode.
The CPU benchmarks do not seem to "read" integer performance in "de-turbo" mode that well but, the graphics performance does decrease.