I had a K5-PR100 from '96 to '97 and have really fond memories of it. Replaced it in '97 with a new system with K6-200 though. The K5 back then was good value for many and could hold it's own against Intel chips. I recently found an old magazine where they were complaining about the methods Intel used back then to shut the competition out.

This year, I got a system with a K5-PR166 on an HX board. At first I was impressed, but I've since swapped it for a 200 MMX since some games were unplayable on the K5 (Blood for example - runs fine on the Intel, abysmal on the AMD). Since I now have a faster K6-2 machine, I'm thinking about swapping the K5 back into this one.

The K6 was a strong contender against the P55C when it first came out, but for me personally it really took off with the K6-2, taking Socket 7 systems to the next level.

The K5 are really fast integer processors. AMD did a good job in the internal translation decoder for the RISC micro instructions. The problem is that the CPU design had as a result additional layers of complexity that negatively impacted its ability to clock higher. These problems could be somewhat mitigated by additional fabbing process enhancements but AMD's fab processes were far from being state of the art.

All the delays and the MHz mania didn't comercially help an otherwise very advanced CPU design.

The very rare AMD K5 PR200 is quite a speed demon integer wise. I don't know if anyone tried to overclock one of these past 133MHz (real MHz) but i wonder how a 75MHz FSB X 2 would perform. This would be same setting as a 6x86 PR200, but at these speeds the K5 would be the winner by far, and give higher end Pentium MMX's a run for their money.

The NexGen 6x86 that later became the AMD K6 was not a straight name change, since the NexGen design was optimized for IBM's fabbing process. So there was a slight redesign for AMD production.

Had a bit of a disaster, pulled the K5 to test a couple of IBM 6x86 CPU's i got the other day and 2 pins fell off the K5 as i lifted it out of the socket 🙁 both pins of course disapeared into the deep pile carpet, shame, a sad end to a great CPU, kinda parallels history though I guess, a CPU dead before its time.

carlostex wrote:

The very rare AMD K5 PR200 is quite a speed demon integer wise. I don't know if anyone tried to overclock one of these past 133MHz (real MHz) but i wonder how a 75MHz FSB X 2 would perform.

I have a PR200. I tried to run it a bit faster than 133 MHz, but it wasn't stable. I think I tried 1.5x95 (143 MHz) on my ASUS P5A.

It certainly is an interesting CPU. Its integer performance per clock is probably similar to P6. Unfortunately it was about a year behind the curve.

I have a K5 system clocked 1.5x75 MHz on a i430TX board and it runs very well for it's clock.

BSA Starfire wrote:

Had a bit of a disaster, pulled the K5 to test a couple of IBM 6x86 CPU's i got the other day and 2 pins fell off the K5 as i lifted it out of the socket 🙁 both pins of course disapeared into the deep pile carpet, shame, a sad end to a great CPU, kinda parallels history though I guess, a CPU dead before its time.

I'd look up the datasheet. they might just be GND pins that are duplicated elsewhere. Or just try it and see if it works.

swaaye wrote:

carlostex wrote:

The very rare AMD K5 PR200 is quite a speed demon integer wise. I don't know if anyone tried to overclock one of these past 133MHz (real MHz) but i wonder how a 75MHz FSB X 2 would perform.

I have a PR200. I tried to run it a bit faster than 133 MHz, but it wasn't stable. I think I tried 1.5x95 (143 MHz) on my ASUS P5A.

It certainly is an interesting CPU. Its integer performance per clock is probably similar to P6. Unfortunately it was about a year behind the curve.

They are ridiculously difficult to work on anything higher than the rated clock speed. I have a PR133 that I tried running at PR166 speeds (116.5ish I believe?) and it wouldn't even POST. Such a shame too 🙁. Both AMD's K5 and Cyrix's 6x86 were really interesting desings, it's just a damn shame they had so many yield issues after a certain clockspeed.

BSA Starfire wrote:

Had a bit of a disaster, pulled the K5 to test a couple of IBM 6x86 CPU's i got the other day and 2 pins fell off the K5 as i lifted it out of the socket 🙁 both pins of course disapeared into the deep pile carpet, shame, a sad end to a great CPU, kinda parallels history though I guess, a CPU dead before its time.

cpu 003.jpg

Ever thought of stealing some pins from another CPU and soldering them on?

I have an AMD K5-PR200, however I am not enough of a risk taker to see how far it can be overclocked. The K5-PR200 maintains value in that the lower grade chips cannot be overclocked to 133 MHz. A similar concept is true for the Cyrix 5x86-133/4x, although not so much concerning the rate Cyrix MII-433 GP because lesser grades can run at 300 MHz reliably.

I have tried overclocking the K5-PR166. The most it would overclock and still complete a heated round of benchmarking was about 4 MHz.

Just yesturday I had a pin fall off a CPU. They an be soldered back on with a steady hand.

Another socket 7 chip which has great clock-for-clock performance is the RiSE mp6. This chip also has good clock-for-clock FPU results.

feipoa wrote:

I have an AMD K5-PR200, however I am not enough of a risk taker to see how far it can be overclocked. The K5-PR200 maintains value in that the lower grade chips cannot be overclocked to 133 MHz. A similar concept is true for the Cyrix 5x86-133/4x, although not so much concerning the rate Cyrix MII-433 GP because lesser grades can run at 300 MHz reliably.

I have tried overclocking the K5-PR166. The most it would overclock and still complete a heated round of benchmarking was about 4 MHz.

Just yesturday I had a pin fall off a CPU. They an be soldered back on with a steady hand.

The K5 PR200 is valuable because it is rare. It was released at the same time as the K6, so AMD did not give it much publicity because they didn't want it stealing sales from the K6. They aren't easy to find at all and you pay big money for them when they do show up. $150-$200 is not an unusual price when they do come up.

You don't think the fact that no other K5 chips can run reliably at 133 MHz adds value to the K5-PR200? If the PR166 ran reliably at 133 MHz, I would have had little desire to obtain the PR200, unless it was sufficiently cheap. Since no other K5 chip could run at 133 MHz, I was willing to pay more for the K5-PR200. From others who think similar to myself, you add another interested party group for K5-PR200, which adds to its value.

Does anyone know if the 5k86 really is faster/more efficient than the SSA/5 stepping? Since the later 5k86 uses a higher PR rating than it's clockspeed, while the earlier one has a PR rating exactly at it's clockspeed.

oerk wrote:

Does anyone know if the 5k86 really is faster/more efficient than the SSA/5 stepping? Since the later 5k86 uses a higher PR rating than it's clockspeed, while the earlier one has a PR rating exactly at it's clockspeed.

Imho - no.
i doubt PR120 is 33% faster than PR90 (both use same clock speeds). Same for PR100 and PR133. Of course Amd have fixed some troubles in the 2nd K5 core revision. But they aren't wizards. 😀

Imperious wrote:

BSA Starfire wrote:

Had a bit of a disaster, pulled the K5 to test a couple of IBM 6x86 CPU's i got the other day and 2 pins fell off the K5 as i lifted it out of the socket 🙁 both pins of course disapeared into the deep pile carpet, shame, a sad end to a great CPU, kinda parallels history though I guess, a CPU dead before its time.

I'd look up the datasheet. they might just be GND pins that are duplicated elsewhere. Or just try it and see if it works.

I did try it, sadly it doesn't work, at least one other pin is loose as well, this CPU came from ebay wrapped in a napkin in an envelope, so lots of bent pins, took me hours to straighten them out, obviously some were badly weakend by the whole experience. Both pins dissapeared into the carpet too, so no hope there, not that I can solder anyway. If anyone is confident and wants to repair it then it's yours free, PM me if so. Rather it stood a chance of being used again.
Best,
Chris

feipoa wrote:

You don't think the fact that no other K5 chips can run reliably at 133 MHz adds value to the K5-PR200? If the PR166 ran reliably at 133 MHz, I would have had little desire to obtain the PR200, unless it was sufficiently cheap. Since no other K5 chip could run at 133 MHz, I was willing to pay more for the K5-PR200. From others who think similar to myself, you add another interested party group for K5-PR200, which adds to its value.

Actually, no, I don't believe that's a consideration for most people who end up buying one. The fact that they are almost never seen plays a bigger part in driving the prices up, like the Cyrix 5x86 133mhz or MII 433GP.

The sickly K5 has found a new home with ODwilly, posted today, hopefully he can breath new life into it. So no longer available.
I ran a Cyrix 6x86MX 166(133mhz) against a K5 166(117mhz) in Phil's ultimate VGA benchmark, exactly the same hardware, K5 came out 7 FPS quicker than the Cyrix.

So we have pretty much agreed that the K5 was a good design, especially considering the clock limitations and AMD's ability to make the thing, just a bit too late.
What do we know about AMD's K6? Not the Nexgen that was released, but what was in development before the nexgen purchase. I am really interested in what "might" have been here, as a lot of the k5 architecture seems better than the nexgen k6 clock for clock, so the next step must have been pretty cool(clearly not "great" enough though, or nexgen buy out wouldn't have happened).

In the next month or two, I'm going to be moving some stuff out of long term storage (ie near 20 years), and was thinking about some of the things I will keep and get rid of. When it comes to my computer and game stuff, I'm keeping all of that, and it should have my old NexGen P90. It was having issues when I pulled it, but that was nearly two years after the AMD buyout. I decided that one day it would have value as a collector piece or a parts board. If I recall it was ISA and PCI. If I can find and repair it it will make a fine little oddball PC in my collection. I'm kinda curious if there is a known list of games that don't work well with the NexGen boards. I remember I had TON's of issues with MANY games on that board. But it could have been because of the hardware I had installed on it. I built it as a budget workstation for myself when I was working at my family business, as I was given a very tight budget. Any recommendations as to what sound and video cards pair well with it? Does 98SE function on it well? I never liked '95 and would rather not go back to it. Maybe straight up DOS?

Mainly thought I'd ask about it since I've got a few months to look into getting some hardware for it before I pull it out of storage. Just wanted to ask any collectors of the NexGen platform if they had recommendations. Right now I'm thinking about using my CT2760 Rev 3 board (I have the non rev board in another rig and its staying put) for audio, and a PCI Diamond Stealth III S540 for video. I'll try a VooDoo 1 if its known to not have issues with this board. Please feel free to enlighten me on how to best outfit this odd little fellow.

i used to have good impression on k5's efficiency. yeah its doing surprisingly well in 3d benchmarks including quake2/3 and 3dmark01, despite having no multimedia instructions.
but the clock-to-clock lead against p54c and k6 is minimal, and comparing k5-pr200(actual 133mhz) with p54c-200 or k6-200 is simply a joke.

I overclocked a K5-133 to Pr200 on a pcchips m571

The trick is that the board has unlisted voltages beyond 3.5 Volts, I believe I had it around 4 Volts

On the board you will see a bunch of voltage jumpers, if you bridge more than one you get more voltage

I had a good heatsynce and fan (for the original cyrix200 ) and ran it that way for some time at a higher than normal bus speed.