VIA EPIA SP 8000, a great replacement for the K6-III+

[TA152H]

Hi all,

Like most of you (I think), I have stuck with my K6-III+ system because I prefer it to the modern systems, because of the heat/power and how quiet it is. For most of the things I like to do, it runs fine and for those that it does not do well enough, I use my other systems.

Back in the 1990s, a company named IDT had a line of processors from a portion of their company called Centaur. These were really bargain basement processors that were made much simpler than anything Intel or AMD was making, with the benefit of being much lower power and much less expensive. They did not reach the performance levels however. They were called Winchips, and their main design goal was getting the most processing power per watt and size of chip (which, of course, are related).

VIA bought both Cyrix and Centaur, and immediately killed Cyrix, and have used Centaur to continue to make processors with essentially the same design goals as before. They have miserable performance compared to Intel's and AMD's best, but they are very low power and still can run a lot of stuff adequately. Sounds a bit like a K6-III, right? Well, that is sort of my point, the same things that inspired me to build a K6-III+ system long after I had an Athlon motivated me to try out a VIA motherboard/CPU combination. The difference being, the VIA chips are brand new and not five year old technology that still has useful work left on it.

Having said that, the performance of my K6-III+ running at 500 MHz is roughly the same as my VIA EPIA SP8000 motherboard/processor combination, from purely computational perspective(meaning forget everything outside the processor, including memory). Considering the C5 processor runs at 800 MHz, this is not a big endorsement for VIA, especially since overclocking a K6-III+ to 550 MHz is very common, and 600 MHz is also sometimes possible. On top of this, the EPIA SP8000 has no AGP slot available, and only one PCI slot and memory slot. If you can tolerate these, in every other way the new system is much better than the K6-III+.

For one, this system comes in a mini-ITX form factor, which makes it quite small. Unlike most K6-III+ systems which are full size ATX motherboards, this frees up quite a bit of real estate on the desk, and is quite helpful. Also, while I mentioned in the previous paragraph that that the purely computational power of the K6-III+ at 550 MHz would surpass it, the actual performance of the VIA system is significantly better because the memory is so much faster and the chipset is so much more advanced. All the I/O is naturally going to be faster, and where the K6-III+ dies on memory access, the VIA system running at 133 MHz FSB running 200/400 DDR memory offers vastly superior memory performance. Oddly, even the cache memory access is faster, even normalized for clock speed. So, overall performance of the VIA system is noticeably better, even if the actual processor is not as fast.

The next nicety is that it uses no fan! These chips use much less power than a K6-III+, which can easily fry without a fan. If you do not mind a fan, you can get the SP13000, which uses a 1.3 GHz processor and easily outperforms any K6-III available. It is also uses less power than a K6-III+ (remarkably, only 8 watts), and thus uses a much quieter fan. I did not want a fan, however.

This system is extremely modern too, with Firewire (1394), USB 2.0, RAID, SATA, FIR, the aforementioned support for DDR 400 memory, and of course mundane things like IDE, on-board sound and Ethernet (although it has no floppy connector). It also supports SSE, which is much more useful than 3D Now!, since it is better supported. As mentioned previously, it has only one memory slot, although this memory stick can be as large as 1 Gig, so it should not be a problem considering the uses of a low performing machine like this is.

To sum it up, this system outperforms my K6-III+ and is considerably more responsive, is much quieter, uses much less power, is much smaller and in subjective terms much more pleasant to work with. On the downside, it is less expandable and is not cheap (the motherboard/processor was almost $250). On top of this, you might need a new case, etc..., so it is not trivial. The power usage is lower, so it is less expensive to use, especially in the summer where you have to use an air conditioner in the room.

So, I just wanted to share this you all. I really like this new system, and if you can live with the limited expandability, I think just about everyone would prefer it to a K6-III+. That is a big "if" though; I used a TV Tuner card and thus can not upgrade the sound or video processor if I were inclined to. Or add the next I/O protocol that comes out in a couple of years (assuming one does, which seems safe ). Even if you do not wish to use a TV tuner and would rather use a video board, you are still stuck with PCI and no AGP or PCI Express (Then again, not much is out for AGP 2x anymore either). If you can live with these limitations, then I strongly recommend this system, or the SP13000 if you can live with a CPU fan and need greater performance.

One last idea for people that like quiet systems. I recently bought a laptop hard drive converter that allows one to use laptop hard drives (2.5 inch) in their desktop machines. Why would anyone want to, especially since you lose any size advantage? Well, they are quieter, use much less power and give off much less heat. Less heat means less likelihood of needing a case fan. Even with a SP8000 (fanless) motherboard/processor, if you use a standard 7200 RPM hard disk, you will need a case fan to dissipate all the heat it creates. Something that runs much cooler is not only quieter, it probably will extend the life of your system because, as we all know, heat dramatically lowers the lifespan of a system. Of course, their performance is significantly inferior to the standard 3.5 inch desktop drives running at 7200+ RPMs. If you are wiling to make the tradeoff, I found the adapter here - http://www.bixnet.com/adfor25to35h.html .

[tazwegion]

Ummm... interesting post there TA152H, kinda' an anti-K6plus theme though

While the platform you suggest IS quiet and reasonably powerful, making them especially suitable for entertainment system platforms (due to their low heat properties)... I personally am a die-hard K6 fan and couldn't envisage such a cash outlay for a meagre power upgrade (Skt. A maybe ) however thanks for the heads up all the same

BTW I gotta' admit a 12 x 12cm (Nano-ITX) motherboard is kinda cute, and do have a cult following

[Jim]

From my standpoint, I don't use ATX socket 7 boards at all; - just ATs. Why? - If I want something with a footprint the size of a "Sasquatch", (a form of "Bigfoot" abominable snowman said to inhabit the interior of British Columbia here in Canada), then I might as well build a poo 4. Also would point out that I have had 5 450ACZs and every one of them has been good for 600, so it is the exception not the rule for them not to clock 600.

[sylvan2626]

Memory is fuzzy on this Via CPU, but here are my questions:

Weren't these processors supposed to get increased (or maybe "enhanced" is a better word) performance from certain software? I think I remember that "the software" would give performance above the CPU's rated speed. Also, weren't these CPUs integrated into the board's design?

Do any of you remember this (or maybe the 60s and 70s have finally caught up with me)?

[TA152H]

Sylvan,

If you like the K6 because it is a K6, that is fine with me. Irrational likes and dislikes make life worth living, and no one has ever talked me out of any of mine. So, enjoy your K6s, they will do useful work for a long time to come. However, if you decide you have to upgrade in the future, you might want to look at the VIA processors since they are much more similar to K6s than any of the power beasts from Intel and AMD. Another alternative, which I just bought, is the Pentium III. At 1.4 GHz, they only use 32 watts. Considering the performance of this processor, exceeding even the K6-III clock normalized, that power use is quite low. I needed something more powerful than the VIA or K6-III chips, but still wanted to stay low power, so this processor made a lot of sense to me for a game server I run.

With regards to the software, maybe it is DOS 6.2 . Seriously, if you look at the architecture of these chips you can see there is no way to get any type of performance out of them. They are very simple, very slow processors. There is no hidden power, or type of software that will run much faster on this processor. It is just slow. Floating point in particular is nothing short of abysmal.

As far as the CPUs being integrated into the board's design, that is sort of true. As I indicated, you buy a motherboard/processor and you can not change it. However, the same motherboard is available with different processors, so in that sense it is not part of the design per se. Because of the size of these boards, there probably was no other way. A full size Socket 370 processor would be just too big.

Jim, your experience with K6-IIIs is better than mine. I tried five processors, and only one would do 600 MHz reliably. AMD did a horrible job on the shrink to 180 nm, as the K6-2 made on .25 could run 550 MHz. AMD was the only make that made the transition to .18 without increasing the clock speed. This was unheard of, and I still have no idea how AMD could make a shrink that bad. In contrast, Intel was able to jack up the Pentium III to 1 GHz, although admittedly it was a superior design. Even so, the K6-III should have made 700 MHz easily, and probably 800 MHz. 550 or 600 just sucks, but I guess they did not care about it with the Athlon being out. Probably just as well, a 800 MHz K6-III+ would have given the Athlon Hell. Pity though, I would love to have that processor.

I am not sure what a Poo 4 is, but I would guess from the 4 that you are referring to a Pentium 4? If so, why?????? Why would anyone want arguably the worst processor ever created? It uses enormous amounts of power, has generally poor performance and is considered a failed design even by Intel. Pentium M, or Athlon 64 would be a better choice except for very specific applications. On the other hand, if you like the Pentium 4 so much, you can get it in small form factors too. If you like Micro-ATX, you can find plenty of motherboards for the Pentium 4 available. I am still scratching my head as to why a Pentium 4 though.

[Jim]

Well, one thing at a time here. First, I have noticed from your posts that most of the K6-3+ (lll+) processors that you buy are the El Cheapo low voltage variety. ATZ APZ etc. ANZ is expensive low voltage variety, and might actually be good for 600; but generally if you want 600 go with the ACZ. How many of those do you have? - All of mine are ACZs. (The price difference on E-Bay between an ATZ and an ACZ is less than $10.00, and I suspect that you made your choice based on voltage rather than price. Admirable, in this world of diminishing resources; but unrealistic to expect the same speed on the same die size at 20% lower voltage). As for a "poo 4" or alternately "pee 4"; yes you are right; and I thought I had expressed my opinion of them clearly enough; but apparently not. The issue is not any particular processor, I really haven't looked into newer processors and could not tell you what is bad or good. The point I was making is that if I am going to switch to ATX, (with its barn door sized footprint), it will not be to build a socket 7 machine. Sacrifice of that much space; and the acceptance of an ATX case, (Pick one, to me they all look "Buck Rogers" - trying to look "Futuristic"), would be something I would only consider doing if there was a huge gain in performance; - which is admittedly possible, - but for me unnecessary.

As for the 700 Mhz, it has been done. I have seen speeds reported stable as high as 740 Mhz. Extreme machines yes. But it is not the fault of the K6-3 that the mobo manufacturers never built a socket 7 board with a multiplier higher than 5.5.

The final point I would make is that these VIA processor on board items you are talking about, do have an attraction; namely they must have a very small footprint with a board size like that. However, it is something like the PC Chips AT boards that would support a Plll 1.3 Gig processor. - Only 1 PCI slot with onboard everything. I considered going that route; but decided not to. Too big a sacrifice for the gain in speed. But between the two, I think I would go with the PC Chips. More speed for the same sacrifices.

Edit: I did miss something here. You said the VIAs could be had up to 1.3Gig. That I missed. I thought you were just saying 800Mhz. Even so I think given the compatability of the PC Chips with my cases, power supplies, etc. I would still go with one of them between the two.

[TA152H]

Jim,

With regards to voltage, I totally disagree. A processor that can run at 450 MHz a 1.7 volts is almost certainly going to be more overclockable than a 450 MHz processor that needs 2.0 volts. When I bought the ATZs, I had the choice between them and 2 volts ACZs and figured the ATZs were better, since .4 volts for 50 MHz is an easy tradeoff to make. The ANZ, however, running 100 MHz for .2 volts, is probably a superior processor, we agree on that. I have little doubt they would run at 600 MHz, but they were not available when I was looking for K6-IIIs. My APZ runs runs 600 MHz without any problems, and I wonder if it is just random luck, or it is because they are taken from a better portion of the wafer. It may also be manufactured later,as my ATZ were 1999 models if I remember correctly. I would guess 2002 versions would probably tend to run a bit higher.

With regards to 700 MHz, having a few chips that can make that is not even close to what I am talking about. They should have been able to create a chip rated at 800 MHz pretty easily from that shrink, if you consider the clock speed improvements of the Athlon and Pentium III when they moved from .25 to .18. Or any processor. AMD did a terrible shrink as the max rated clock speed stayed at 550 MHz. The fact that these chips struggle at 600 MHz in many cases does not point to a problem with motherboard makers, but with the processor itself. The motherboards do support 600 MHz, or really a 6x multiplier, so you can get higher than 550 easily. 2x is interpreted by the K6-III as 6x. Getting much higher than 600 MHz would be difficult even if motherboards supported it, and AMD would never have been able to rate them at 650 MHz without modifications to the processor. The ones that came out would not cut it, despite the rare bird that runs at 700 MHz.

The attraction of the VIA chips is by no means limited to the size of the motherboard. That seems to be your bias, whereas mine is more the power usage/heat and really the level of noise. The size to me is secondary or even tertiary, and to be honest I would have preferred a full size system that ran without a fan, except they do not offer it.

With regards to the tradeoffs, I can easily see why you would consider them too severe. I think a lot of people would, and do. While I felt they were worth it, they definitely are not something I am crazy about or find irrelevant. I simply put up with the limited expandability because I like the other qualities of the system. It certainly agree it is a very major compromise to make.

With regards to size/performance, if noise is not an issue for you, you can get the 1.3 GHz VIA chip. It uses a fan, and a quiet one at that, but since noise is not your primary consideration, you would get much more power at the same size. You still have the expandability issue though.

[Jim]

You seem to be missing a major point. ACZs do not "struggle at 600 Mhz". They do it easy. 5 out of 5 , (all 1999), in my case. How many of them are good for 700, I do not know, because I have never OCed the FSB, because I am using a DFI which, (mobo again), does not have any provision for OCing the FSB. The fact is for whatever reason, ( I suspect that AMD down rated the speed and voltage of the lower grade production), you are wrong about which is the better chip between ACZs & ATZs. You are also wrong about the mobos supporting a 6x muliplier. They don't. It is only because later K6 processors remap the 2x as 6x that 6x is available. I suppose AMD could have remapped the 3.5x as 7x too; but why bother, if the mobo people are so far behind? If I had any Epox boards, I probably would have one running over 600 right now. But since I am posting this with Superpuppy 2 which has an ASUS P5A-B, which does have OC settings for the FSB, and since my spare board for it has just benched good, which this one is not, (12v is out of specs, - 12.7), I will soon be replacing it and heading upwards from 600.

Edit: Addendum: While in therory a lower voltage processor, (all else being equal), should make a better OCer, that will not be the case if the reason they are lower voltage is that they were unable to cut it at higher voltages and higher speeds. These are the processors that are likely "to struggle" when OCed to any significant degree. Be it also noted, that since they have different 3 letter designations, there is no guarantee that they even have the same architecture.

P.S. If you want 700 Mhz, and silence, which I realize is your top concern, get an ACZ, a good quality mobo, and go the water cooled route. That seems to be the way it is done; and it is not just the very rare chip that can make it. The limiting factors seem to be mobo stability and cooling.

[TA152H]

Jim,

As I mentioned, if you had read carefully, the later processors read 2x as 6x.

You were lucky with your ACZs and making the assumption that all work as well as yours is plain wrong. Most people can not get those chips to make 600 MHz, or at least the ones I have heard from. They are the worst rated K6-IIIs, being stuck at 450 MHz at 2.0 volts.

To buy an inferior processor and hope it reaches 600 MHz relies too much on luck. If I wanted 600 MHz, I would buy a 500 ANZ or 450 APZ. My 450 APZ does 600 MHz no problem.

You missed the point completely about higher than 600 MHz. Motherboard makers did not support it because AMD did not make the chips for motherboards, and in any event the chips can rarely make it over that, and even more rare do it reliably over time. Remember, AMD can not put out chips that miss even 5% of the time at their rated speeds. If they did, they would quickly get a reputation for unreliability. So, they have to be very safe in their estimations.

By the way, all processors, even the lowly ACZ, are made on the same wafer as all the others. Even the 500 ANZ is made on the same wafer as the 450 ACZ, but it comes from a part that AMD estimates would more likely have better speeds based on their tests. Of course this pertains to processors made at the same time, as 2002 models may have some small changes to them. You were lucky with yours, but if you are AMD you are not going to take ACZs from a better part than a 1.6v ATZ. My ATZ does 450 MHz reliably at 1.6 volts, it does not need 2 volts. So, in odd instances you may end up with better processors, but as a rule it is unlikely that AMD was wrong every time. Of course, anything is possible.

[Jim]

There is no point in arguing about it. You took the view that the lower voltage chips would be better performers. I took that into consideration; made inquiries; and found that was not the case. I then took the view that the higher priced chip, (more desirable), would probably be the better performer. The results would seem to show that I was right.

If not why is the ATZ only rated at 400? Why not 450? ... Answer because it comes from an inferior part of the wafer, (as you put it), and can not be expected to clock as high reliably in every case. Therefore, downrate them, both in speed and voltage; and sell them as 400s.

Using that logic the ANZ is probably the best of the bunch, - rated 500 Mhz @ $100.00 U.S. Probably hand picked specimens. In any case the so called "Lowly ACZ" will womp the doors off an ATZ. And since the mobo people were still stuck at 550 Mhz while AMD was putting out processors capeable of going over 600, it is no wonder that AMD finaly gave up on it.

(Note): I saw what you said about mobos, but the catch is that you said: "The motherboards do support 600 Mhz, or really a 6x multiplier" That is straight out wrong. (Although you went on to say that the K6 remapped the 2x as 6x.); But the point is that has nothing to do with the motherboard. It is entirely the processors doing; and if you doubt that, try it with an intel. The point I was making is that AMD was way out in front of the mobo manufacturers; and had been for some time. Why else would they have to resort to remapping the 2x as 6x in order to get the available boards to run at the speeds that AMD was producing processors for. Have you ever seen an Unintel processor that remapped board multipliers to achieve its rated speed? (Answer = maybe: The P1 used the 1.5x multiplier as 3.5x; though I am not sure if it was the processor or the mobo doing it). The issue was the mobo manufacturers were still piddeling around at 550, when AMD had processors capeable of over 600 ready to market if ever the mobo people woke up. They didn't and AMD gave up. (If Intel was remapping 1.5 as 3.5, then the parralel is striking. because they gave up on socket 7 too; and maybe for the same reason).

Just like my schtick is footprint, yours is voltage, and it seems that you are willing to put up with lower performance in order to achieve lower voltage. Nothing wrong with that, laudable actually; but don't go trying to say that the average ATZ will outperform the average ACZ , (except in terms of power consumption), because that is just plain wrong.

Using your logic my old K6-2 550 AGR is a worse processor than a K6-2 500 AFX, because the AGR uses 2.3 volts while the AFX uses 2.2. But in point of fact I think the AGRs were sort of hand picked from the run of the mill, which for the most part would not do 550.

Anyway this is getting too argumentative for my taste. Yes there are ACZs out there that won't do 600, (in most cases probably because of an instable mobo), but as regards the remainder, they are the exception, not the rule.

See Article at Ace's Hardware Re K6-3+ 450 ACZ: "All specimens tested were able to clock 600 Mhz" etc.

NOTE: I just went back and reread the ACE's articles myself. I was wrong. It was ACZ 550s they were talking about, both K6-2 & K6-3. Still If you check people's sig's you will find that most machines running 600 or higher are 450ACZs. (And those ACZs that arn't 600 or higher, usually are running higher than 100 Mhz FSB.)

[tazwegion]

Integrated SFF mini-ITX platforms are a great idea if you like small footprint low wattage processing, but in reality motherboards die leaving a homeless (but still useful) CPU, if one of the integrated components go down... where does that leave the remainder, ONLY one PCI expansion slot remember?

[DasMan2]

I am staying AMD K6-III "+" and K6-2 "+" for a long time.

I have had 4 ACZ 's (K6-III "+") all did 600 over higher ...depending on the motherboard the onboard cache affects overclockability ...turn off to eliminate that stumbling block on some boards. More so related to the Bus speed chosen to achieve the 600 speeds.

Heat Sink Fan units have to be good to keep any heat issue from arising.

5+years been running my Gigabyte GA-5AX rev 5.2 at 600 (5 x 120 bus speed) Good memory can make a difference but the onboard cache or even TAG ram or Timing chip can be issues if they heat up.

Again more overall cooling the better

It would be great to get AMD to make their own upgraded Super Socket 7 M.B. with modern features and connection ports ,alot of sales of socket 7 boards and CPU's still go on E-bay ...there must be a market .

Alot of people still buy *Plus* CPU's and usually more than what Duron's sell for AND that has been for sometime now

I have "AT" AMD K6-2/III boards and the cases were dirt cheap and the size is small enough to be "Cute" with cheap Power supplies available all the time. Alot of uses for them easy for along time.

I think that when a very low power fanless heatsinkless? more powerfull CPU/mobo comes out I might think of buying that, although I am keeping my AMD K6-III"+" til they die off totally.

Peace .....

[TA152H]

Jim,

The ATZ is rated at 400 MHz at 1.6volts, not 400 MHz at 2.0 volts. You don't decide a processor can't run at 450 MHz at 2.0 volts, and then downbin it to 400 MHz and lower the voltage. If it can't make 450 at 2 volts, there is no way in Hell it will make 400 at 1.6. 400 at 1.6 is harder. All my ATZs make 450 MHz at 1.6 volts,by the way. The reason AMD made them should be obvious; some laptop makers wanted a very low power processor. So, AMD gave them extremely low clock voltage and low clock speeds. Keeping in mind that computer manufacturers are loathe to run anything out of spec, these processors are the best possible for power use, better even than a 500 ANZ. Could a 500 ANZ run at 400 MHz at 1.6 volts? Without a doubt, probably even lower, but no manufacturer would want to run a processor out of spec. It is not good for their reputation.

By the way, the price for a new 500 ANZ is $75, not $100. This is a bargain compared to a 550 MHz ACR which typically goes for twice that much used. They are probably equal parts too, .2 volts for 50 MHz is about right. I have no doubt at all either one could do 600 MHz.

Your remarks about motherboards and AMD are strange to say the least. I have no idea how you can think the motherboards makers should have created multipliers higher than 6.0. First of all, AMD did not make these processors for desktop machines. They hid the fact they could even work on them until others found it out for themselves, and in any event were very difficult to even buy outside of a new system. They were mobile only CPUs, remember? On top of that, AMD only rated these dogs as high as 550 MHz, so even supporting 600 MHz is overkill. Expecting a motherboard maker to support a dead platform, for processors that should never be used on it, and are not supported for that use by the maker, and at two performance scales higher than the rated clock speed by the manufacturers, is extremely unrealistic. I do not understand at all why you think they should have when AMD never intended for these processors to be used in such a fashion.

Your point with respect to the remapping of 2x to 6x is incorrect. This was not done for 100 MHz systems, but for lower system bus machines. This way a maker could run a 66 MHz machine at 400 MHz, or some other weird multiplier. Remember, on mobile systems lowering the system bus speed can save on power, and performance is often less important. So AMD gave them an easier way to.

Your remark about K6-2s seems intent on obfuscating the discussion, as you use a .1 voltage/50 MHz difference to make the point that a .4 voltage/50 MHz difference is the same. In my experience, it is roughly .2 volts to 50 MHz. So, I would agree completely the 550 is better in the example you gave. If the voltages were 1.9v to 2.3v, I would not.

Dasman, there is a market for super7 for sure, it numbers in hundreds instead of millions. AMD would never waste the time on such a small market though. On top of this, it is a dying market as K6 based machines become less and functional vis-a-vis modern requirements. Super 7 is dead, for sure, although it surely could be made functional if someone really wanted to. A modern motherboard with say a 1 GHz K6-IV running SSE/SSE2 400MHz DDR and running at 20 or so watts (probably less) would be very attractive, except VIA is already making a 1.3 GHz processor for that market, and has either released, or is going to release, a 1.7 GHz processor.

As much as people like the K6, you have to keep in mind it is an inferior processor, and an inferior design. It was released after the Pentium Pro/II/III (so was the K5 for that matter) and never matched it in performance, or even came close. It had yield problems the whole way because the design struggled with high clock speeds, and was just barely good enough to keep AMD in business while they developed their own processor (the K6 was largely designed by NexGen, a company they bought). I do not intend this as an insult to anyone using the K6, I use them too , but I think people become so fond of technology they lose sight of why it was allowed to die. The K6 because, all in all, it sucked. The Pentium Pro/II/III never quite did die, because it was so good. The Pentium M is largely based on it, the Athlon is largely based on it, and the P8 will be based on it. Nothing is based on the K6.

Again, I do not intend this as an insult to anyone, because if I am insulting you guys, I am insulting myself too. I have plenty of K6 machines, and I have over a 100 vintage machines from Apple II's, to TRS-80s, Commodores, original IBM PCs, etc... Irrational likes are what make life worth living, and what make businesses bankrupt. So, while we can enjoy these obscure and interesting processors like K6s, you can hardly blame AMD for dropping relatively bad technology when they created something better. If they did not, I doubt they would be as strong a company as they are.

[ChrisW]

Actually, other than the fact that both the P3 and the K7 are x86 chips, they are widely divergent in design - it's not terribly accurate to say "The Pentium Pro/II/III never quite did die, because it was so good. The Pentium M is largely based on it, the Athlon is largely based on it..." The original Athlon was, in many ways, a revolutionary design.

The original K6 most certainly did not "suck". Don't you remember the good-to-excellent reviews it garnered? Anand Lai Shimpi essentially began Anandtech with a review of the K6. A 600mhz K6-3 can beat an Athlon 700 in some office application scenarios.

The K6/K6-2 CPU was good up until around 400mhz... and then the reliance upon motherboard-bound cache started to strangle it. The K6-2+/K6-3(+) iterations alleviated that problem, but the fabricating expense combined with the looming obsolesence of SS7, and AMD's desire to leap from "follower" to "leader" did in the K6.

[Jim]

Whether or not the K6-lll+ was always intended to be a mobile processor is for me at least an open issue. I was not privy to the inner councils at AMD. Sources that I have seen have indicated however that they did intend to release a 600 Mhz version of the chip. (Which is probably why so many of them can reach that speed). I real ize that when AMD remapped the 2x as 6x they were doing it to achieve 6x66; but the point that you miss is it is not normally the responsibility of the processor manufactureer to supply the multipliers. You still seem to be giving the mobo people credit for the ability to reach 600, (6x100), when they don't deserve it, because the highest multiplier they supplied was 5.5 not 6. Further, alot of their boards are not stable at 600. Not even socket 7 laptop boards have 6 x multipliers, so the bottom line is the mobo manufacturers were not holding up their end of the deal. And when the mobo people don't hold up their end of the deal processor manufacturers seem to have a history of saying screw you. Edit: Those words were poorly chosen. Of course the processor manufacturers and the mobo manufacturers have to work fairly closely together, or it would be impossible to produce compatable components. Exactly what passed between the mobo people and AMD, (and Intel for that matter when they gave up on socket 7), is not something I am privy to. In the case of the K6-lll, however, which was nearing the end of its upgrade cycle, there was little incentive for AMD to once again do the mobo people's job for them. Yes the K6 in some respects is an inferior processor. It certainly was a deadend processor, because its architecture limited it to a top speed of about 750 Mhz. (See articles at Ace's Hardware). Nonetheless for those whom speeds below 750 Mhz are satisfactory it is a good performer; and a challenge to get as close to 750 as possible.

As to your ATZs, fine, they are better than ACZs, 450Mhz, (correction 550Mhz), is better than 600+, and a Sopwith Camel can fly rings around a TA152H1C. All equally true.