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

Discussion about systems that do not use the K6-x processors.
sylvan2626
Senior K6'er
Posts: 139
Joined: Tue Apr 05, 2005 3:25 pm
Location: Florida, USA

Post by sylvan2626 »

The K6 CPUs are "locked" by design at a max multiplier of "6". Wouldn't it be great if someone figured out a way (like the Athlon pencil trick or the wire trick) to sidestep that barrier?

Yeah, I know that I am a noob in the K6 CPU arena and am just now plowing over old ground, but still - what if? And someone could modify the K6Speed program....

My 400ATZ runs stably at 1.3V and with low temps using a heat sink only. So my operating speeds range from 400MHz up to 550MHz and the core voltage ranges from 1.3V up to 1.8V. Cooling provided by heat sink or by FHS at higher speeds.

Can't these Via EPIA boards use a 2-in-1 PCI riser? Pretty sure that they can, so I wonder if a 3-in-1 riser exists? I like the speed offered and the SSE support, but I like my multiple PCI slots as well.
TA152H
Junior K6'er
Posts: 47
Joined: Tue Jul 15, 2003 5:49 am

Post by TA152H »

Jim,

I feel like you are not listening or simply are being stubborn.

Motherboard makers support released processors!!! AMD did not release the K6-III+ family for the desktop market, everyone knows this. I could not even buy one at the time and no one knew if you could pop them into a desktop motherboard. Eventually the web sites tried it and found out it worked, and worked well. The number of people actually doing this is extremely low compared to the number of machines that sell in a supported platform. We are the minority, not the standard. We are too few in numbers to make enough financial sense for big companies to invest money in. Plus, it was an unsupported platform, which they rarely go out of their way for. Plus, 650 on a K6 is not easy. I am not trying to be insulting, but you seem to view everything from the perspective of the hobbyist without realizing that businesses have an entirely different perspective. You think it was a motherboard makers responsibility to support a processor AMD never released, and for a market that died and had moved to Socket A. This is just not realistic. There is nothing special aboutthe K6, the K7 does everything it does. Back in the 80s and even 90s, computers died out that had no replacements. So, at least we can upgrade to a newer x86 and run everything we did on our K6s. If you had an Apple IIe, Amiga, Atari ST, etc..., your system simply died and nothing replaced it. So, we are much better off. Then again, I had to work on my Athlon XP and this damn thing gets so hot I can not stand being in the same room as it. I have to move it to my air conditioned room to even make it tolerable. So, yeah, I wish they made higher clocked K6s too, but not as much as I wish Tandy made a Color Computer 4 :P.

With regards to "my" ATZs, they suck. With regards to ATZs in general, they are probably a lot better than mine. I have heard most do 600 easily. None of mine do. I hope that clears it up. My APZ is great though. It does 600 stable at 2.0v in all tests. Is almost stable at 1.9v too, but fails on the worst ones.

ChrisW, the P3 is the 386, and yes it was very different from the Athlon. The Athlon was not in the least revolutionary, it was a modified P6 core. The P7 (Pentium 4) was revolutionary and sucked. The K7 was an excellent use of existing technology and simply an improved P6, well, in some ways. It is far more similar than you are admitting. It had the same number of pipes, was only slightly deeper pipelined, had split caches, was de-coupled (although they all are now) and pretty much looked just like a deeper P6. Pentium M is even closer.

Where the Athlon differed significantly was the cache, but then again, not really. AMD chose a big L1 cache, that was also relatively slow. Because of this, the L2 cache had to be "exclusive", meaning data contained in the L1 was not contained in the L2. This is one important reason why the Coppermine P6s were significantly faster than the Athlon at the same clock speed. The exclusive cache arrangement has a lot of intrinsic compromise, as does making a big/slow L1 cache. AMD did the same with the K8, so obviously they feel the advantages are more important.

With regards to the K6-III 600 and Athlon 700, you ignore a very important fact. There was no K6-III 600!!!! There was an Athlon 700, and higher. The Athlon was not intended to increase IPC (Instructions per cycle), but to run at higher clock speeds. So, while in some applications the K6-III could match a slightly higher clocked K7, it never matched, or even came close to, the best K7 available at the time. In floating point, it was much slower than the K7 even at the same clock speed. So, you can never compared clock speeds and say one is better because it runs faster at the same clock speeds, especially when the main purpose of the K7 was to achieve higher clock speeds without much attention paid to IPC. Later versions, like the Athlon XP did increase the IPC however, and even reduced power. The original was all about clock speed.

I forget where the Athlon topped out on .25, as I never bought one and still think they are crappy chips. I did buy an Athlon XP, which runs at 1.46 GHz. I think they went as high as 1.7 GHz. The powerhouse K6-III stops at a rated speed of 550 MHz, at a higher voltage. The Athlon XP also has a better IPC than the K6-III, although it is close.

But, that is comparing a later chip with an earlier chip, which I would not have done but since you mentioned the K7, I figured it was fair game. Consider the P6 core, which actually predated the K6. It had a better IPC, and ran at much higher clock speeds too. At .25 it topped out at 1 GHz, although they released a 1.13 GHz that was unreliable and had to be pulled. The Coppermine easily outran the K6-III at the same clock speed, and especially in floating point. So, older, better IPC, and better clock speeds. Granted P6 was a great core, and in my opinion the best Intel created considering when it was released (1995), but still the K6 was so lousy in comparison it is hard to call this anything but a bad design that they used until they could make something better. You can be objective and still like something. I like the K6-III a lot, but my Pentium III 1.4 GHz blows the doors off of it, and is actually an earlier design. Kudos to Intel.
Jim
K6'er Elite
Posts: 1745
Joined: Wed Jan 21, 2004 7:10 pm
Location: Toronto

Post by Jim »

I am not sure when AMD first remapped the 2x as 6x, but I do know it was to enable mobos that only had 66, 75, 83 FSBs to reach 400 Mhz without resorting to OCing the FSB, which is of course a waranty voiding no no. Similarly Intel apparently had to resort to remapping the the 1.5 multiplier as 3.5 in order for their VERY LAST socket 7 processor, (the MMX 233), to reach its design speed of 233Mhz without resorting to OCing the FSB. So much for the mobo manufacturers supporting released processors, because in those two cases they clearly did not do so; (in a timely manner in the case of Intel; and AT ALL in the case of AMD); - There NEVER was a 6x multiplier on any socket 7 board to my knowledge; never mind a 6.5x which would be useful assuming that either the K6(?) processors are not locked at 6x, (as someone here has said) - (which I have never heard of, though there are many things I don't know), or that there is some way around it, which the same person suggested may be possible. There is a series of 4 articles at Ace's Hardware dating from 2000 that I think you should read, one of which speaks of the imminent release of a 600 K6+, which as we all know never happened. Why? Well you take your guess; but mine is that AMD had chips that could reliably cut it at 600, but the mobo people did not have boards that could. (Laptop or otherwise). The same articles discuss tests in which all tested examples of 550s were successfully overclocked up to 660; which would seem to indicate that 600 was certainly within reach. And as for your "Incredibly bad die shrink" theory, the same series of articles delves deep into the limitations of the K6 (?) processors and explains in some detail as to why they were incapable of exceeding 750 Mhz. In sum it is my opinion that had there been reliable socket 7 mobos that could be 99.99% depended on to reach 600 Mhz, with a 6x multiplier available, (to indicate continued support), then AMD would have released the K6-lll+ 600; and possibly even something slightly higher. I doubt if they ever would have gone over 650; though even that is open to speculation, given the way they initially arrived at 550.

Finally, re being stubborn, I still have not seen you admit that a K6-lll+ 450 ACZ is a much better processor than a K6-lll+ 400 ATZ, though I did catch the bit where you said your 400 ATZs "Sucked".
Last edited by Jim on Tue Jul 12, 2005 9:17 pm, edited 2 times in total.
Superpuppy 3
K6-3+ 450 ACZ (6x100)
DFI K6BV3+/66 Rev B2 (2 Meg) w/ 2x28mm Chipset Fans
2x256 Meg PC 133 Hynix SDRAM
1x 20G Maxtor (7200)
2x 80G Maxtor (7200) Ducted w/ 2x486 Fans Mount
52/24/52/16 LG CDR/RW/DVD
8/4/3/12/24/16/32 LG Super Multi
ATI 9000 aiw Radeon AGP
SB Audigy 1 MP3 Sound
CMD 649 IDE Controller
NEC USB 2 Card
ChrisW

Post by ChrisW »

"ChrisW, the P3 is the 386, and yes it was very different from the Athlon. The Athlon was not in the least revolutionary, it was a modified P6 core. The P7 (Pentium 4) was revolutionary and sucked. The K7 was an excellent use of existing technology and simply an improved P6, well, in some ways. It is far more similar than you are admitting. It had the same number of pipes, was only slightly deeper pipelined, had split caches, was de-coupled (although they all are now) and pretty much looked just like a deeper P6. Pentium M is even closer. "

Yes and no; they are all 386's in terms of instruction sets, but the K7 was not (just) a modified "P6" core. In fact, in most ways the K7 architecture is a RISC/Superceded-RISC architecture and probably has as much in common with Motorola's G4 processor as it does with Intel's "P6". It's not totally dissimilar - it stradles the two.

"This is one important reason why the Coppermine P6s were significantly faster than the Athlon at the same clock speed" Not necessarily true. I'll let you pick 4 benchmarks where a Cu P3 will win, and then I'll pick 6 where the K7 will. A better bet for the P3 would be the Tualatin, which maybe would turn the tables on the K7.

"If you had an Apple IIe, Amiga, Atari ST, etc..., your system simply died and nothing replaced it." Nope! I fixed my Amiga A500 :D
TA152H
Junior K6'er
Posts: 47
Joined: Tue Jul 15, 2003 5:49 am

Post by TA152H »

Jim,

The 450 ACZ is not better, in general, than the 400 ATZ. I can not admit it, because it would be incorrect. If you ever overclocked processors, you would realize that you get way more than 50 MHz from .4 volts. I am certain there are instances where 450 ACZs are better, but overall, they would not be. What you are saying contradicts a lot of my experience, .4 volts for 50 MHz. Just makes no sense at all.

With regards to 600 MHz, virtually all the motherboards supported it at that time. They had problems at 550 MHz though, and if they updated it to support it, they had no problems at 600 MHz. Based on my experience, the K6 just could not do it very well. Even my APZ could not run at 600 at 1.9v, and 2.0v is probably the limit AMD would have gone. 650 MHz would have been pretty difficult. But, as I have to keep repeating myself, these were not supposed to be used on desktop machines. AMD ONLY sold them for portables. Now, if you want to argue if AMD could have made the K6 run faster with a better shrink, meaning a new revision, I would be entire agreement with you. I think with a lot of work they could have reached 800 MHz. But, the chips out there can not even get close. AMD really did not care either, they had a better chip out.

Chris, you are misusing terms. The P3 refers to the 386. It does not refer to the Pentium III. You are misusing the abreviation. P4 = 486, P5 = Pentium and Pentium MMX, P6 = Pentium Pro/II/III, and the current P7 is the Pentium 4.

Much to learn, you still have (to quote Yoda). You use terms you obviously do not understand. The first de-coupled x86 processor was the Pentium Pro (P6). What you call RISC is actually correctly called de-coupled with respect to the x86. Since the x86 instruction set is so difficult to work with, they decoupled the processors so you have a decoder front end, and essentially a RISC engine behind it. The Pentium actually executed x86 instructions and was the last x86 processor to do so. Even the K5 was internally a RISC processor. The Pentium Pro certainly was. Avoid using words like Superceded since they make no sense, what you really mean is Superscalar. Superscalar simply means it has more than one execution pipeline. The 386 was just a plain old processor. An instruction came in, it executed to completion, and then it started on the next. The 486 added pipelining, meaning you would fetch an instruction, start working on it in stages, and at the same time load other instructions too, at different stages. So, say our instruction moves to stage 2. On a 386, it had to go all the way through before we would start. On the 486, you would start another instruction in stage 1. The next clock cycle they would move to stage 3 and 2 respectively, and we would start executing yet another instruction. When you hear people say "deep" pipelining, it means there are lots of stages before an instruction completes. On a 486, it was 5 stages. The Pentium 4 has something like 31. This is why it runs at such high clock speeds. With the Pentium, Intel added an additional pipeline, so they had two (called U and V). This is what is referred to as Superscalar. More than one pipeline. The Pentium's was very limited, as the V pipe could only handle simple instructions, and both pipes were "lockstepped" with each other. This means if one stalled on an instruction, the other one also waited in that same stage and both pipes stalled. The Pentium Pro added a third integer pipe, removed all the limitations, and even allowed out of order execution. The Athlon adds NOTHING to that. It is deeper though, and this allows better clock speeds, in general.

The Tualatin is just a Coppermine shrunk to .13, some with a bigger cache. They do not exhibit superior performance to the Coppermine on a clock normalized basis, unless they are a larger cache version. You can find benchmarks showing anything you want, but in general terms, the Coppermine outperformed the K7, was much smaller, used much less power, and the architecture was released four years earlier. In my book, that is really impressive.

What I meant by the "system died" was that they are no longer made. It was not referring to individual systems. If Jim gets sick of his K6 and wants something better that runs his favorite software, he can get it. If you had an Amiga 4000 or whatever, and you needed more, well, Commodore is out of business. Or, if you had a great software library for your loved Apple II type system and wanted to upgrade. Hmmmm, out of luck on that one after the IIgs. So, people who lament about how their K6s were not extended at least have the option of buying modern processors that do the same thing. Those who have older systems that were entirely discontinued with no offsprings have suffered a far worse fate. That was kind of my point on this whole thread. I really hate the new processors because of the heat and noise issues, they make me insane. So, I wanted to mention a processor that was more like the K6 in that it made for a much more pleasant machine. I wish they made new K6s that were low power and performed pretty well, but since they do not, I figured this was the next best thing. I hate my Athlon and getting a motherboard for a Pentium III Tualatin 1.4 GHz is anything but trivial. Even Intel does not support it well :(.
Jim
K6'er Elite
Posts: 1745
Joined: Wed Jan 21, 2004 7:10 pm
Location: Toronto

Post by Jim »

I give up. You still seem to think that socket 7 mobos support 600Mhz. How? 5x120Mhz? Isn't that a bit warranty voiding? You just don't seem to get the idea, that it was because of LACK of mobo support that both AMD and Intel had to resort to remapping old, slow, rarely used, multipliers higher!! Who paid for the design work that went into that? ASUS? EPOX? DFI? JETWAY? I don't think so; and when you don't pay for your share of the family get together dinner, you tend to get left off the invite list. Intel dumped socket 7 immediately as a result. You think they were pleased that there wasn't a single mobo manufacturer with a 3.5 multiplier ready? AMD was more patient; but the end result was the same. Super Socket 7 boards were designed for a maximum, (legal), FSB of 100Mhz. That is what the chipsets were designed for, that is what the RAM was initially good for. How pray tell can you say that mobos supported 600 when their highest multiplier ever was 5.5? You are giving the mobo people credit for work done by the processor people, who were probably rather POed, that they had to do the mobo people's job for them. What job? Create new higher multipliers!! (Which probably would have been simple enough; but not so simple for the mobo people was creating boards that were guaranteed stable at those speeds at that time.)

I don't claim to know a whole lot; you are throwing around technical terms that are way over my head, but I do know that multipliers are generally found either as jumper settings on mobos or as bios settings for mobos. Where is this 6x multiplier you are talking about? What board has it? If I drop my MMX 233 into that board and stick a (waterproof plastic) bag of icecubes on top of it, can I use that board to clock it up to 400 Mhz? (6x66), or even 360Mhz? (6x60). You know, I am beginning to think you are just trying to irritate me. Nobody can be that knowledgeable and yet be so obtuse.

"Shrink" Doesn't that mean reduce the size of the core? (Presumably by photoreducing the masks from which they are produced?) If so, where do you get the idea that reducing the size of a design which is inherantly limited to a certain speed range, (approx 750 Mhz according to people more knowledgeable than me), because of the way it is designed, is going to significantly improve that speed unless you change the limiting factors within the design? i.e. redesign it, not just shrink it.

OverClock: I think that term refers to running a processor at a speed above its stated design speed, either by using a higher than designed FSB, or by raising the multiplier to achieve the same effect. If so, is it not "Over Clocking" to run a 450 ACZ at 600? How is an ATZ which struggles to hit 600Mhz at 2.1v any better than an ACZ which hits 600+ easy as pie at 2.1v? I must be really stupid with my measly 162 IQ, because I don't get it. One other bit of food for thought. Whatever did become of the "600 Mhz Processors" which according to the article at Ace's hardware were soon going to be released? They would have needed higher than average voltage wouldn't they? --Tra La
Superpuppy 3
K6-3+ 450 ACZ (6x100)
DFI K6BV3+/66 Rev B2 (2 Meg) w/ 2x28mm Chipset Fans
2x256 Meg PC 133 Hynix SDRAM
1x 20G Maxtor (7200)
2x 80G Maxtor (7200) Ducted w/ 2x486 Fans Mount
52/24/52/16 LG CDR/RW/DVD
8/4/3/12/24/16/32 LG Super Multi
ATI 9000 aiw Radeon AGP
SB Audigy 1 MP3 Sound
CMD 649 IDE Controller
NEC USB 2 Card
ChrisW

Post by ChrisW »

TA152H, I'm with Jim-- I give up. You are simply too full of yourself to listen to anything other than your own arrogant, bombastic voice.
TA152H
Junior K6'er
Posts: 47
Joined: Tue Jul 15, 2003 5:49 am

Post by TA152H »

ChrisW,

I am arrogant for pointing out you do not know what you are talking about and explaining why? Or are you an idiot for using terms you do not know, and then directing some vague insult based on no information at me? Stick to what you know, and stop trying to sound like you know more than you do, and maybe it will not be necessary for someone to spend the time to correct you.

Jim, I have no idea what I am doing that is making it impossible for me to get across one simple point. There is not 600 MHz K6-III, so no motherboards have a 6x multiplier. Nor would there be, because AMD made it completely unnecessary. What possible reason would a motherboard maker have to create a 6x multiplier when AMD remapped 2x to 6x (since one would never run the processor at 2x, it was a painless and logical choice), and AMD never released a 600 MHz chip!!!! Good grief, this is not rocket science here, and for some reason we are just not communicating on a very simple point.

Your remarks about Intel are completely incorrect, and I wish you would stop putting me in a situation where I have to come off as insulting. For one, my friend, Intel makes motherboards, and lots of them. Saying Intel and referring to motherboard makers as "they" is not accurate. Now, why would they move from Socket 7 to Socket 8 and Slot 1? Well, simple, the K6. The K6 was good enough to compete with the Pentium, and not good enough to compete with the P6 core (Pentium Pro and Pentium II in this case). So, naturally Intel made the move to the P6, just like they did before that with the 486 when AMD cloned the 386, and the Pentium when AMD cloned the 486. Remember, Intel makes the chipsets, and stopped at the 430 TX for the Socket 7. This alone handicapped AMD since VIA was remarkably poor at making chipsets (some think they still are, but in my opinion they have improved) and the move to Super 7 was without Intel's chipsets. Consequently, we have garbage like the MVP3. It had nothing to do with motherboard makers, and if it did, Intel would fire their own employees. It is just what Intel does when their competition catches up with their old technology, which is how things used to be.

With respect to the 400 ATZ and 450 ACZ, you are beating an old horse and you refuse to listen. Your premise is that all 450 ACZs are exactly like yours, and all 400 ATZs are exactly like mine. Considering the small sample set, I think that is quite a leap. My contention is that getting 50 MHz rating for .4 volts is so easy, they would generally be from the part of the wafer considered the worst. Naturally, there will be great ones, and some lots will be particularly good, but in general the 400 ATZ should exhibit better characteristics. Anyone that overclocks will tell you that .4 volts for 50 MHz is a lousy trade. You got lucky, be happy with it. If you still want to believe the 450 ACZ are miraculously the best, go right ahead. At this point the whole conversation bores me and is played out.

With regards to a shrink, I think we need to clear up a few things. There is nothing about a K6 that limits it to say 750 MHz. You could say there is a design limitation that reduces it to 750 MHz at a certain lithography. When you shrink a processor, you are shrinking the line width, meaning a more refined lithography. You have some problems, like greater leakage which can reduce clock speeds (they were a major issue moving to 90nm), but at .18 they were minor. When you shrink the lithography, and processor, everything becomes closer, and you should be able to operate at faster clock speeds. If your design remained exactly the same, at a finer lithography, discounting problems like leakage, you would run faster with no changes. Everything is closer and smaller. Now, the initial shrink is never as good as subsequent ones, where you can find the limiting areas, and there is always at least one, and address them. So, yes, you do get better as you work with it, but even without these changes, a processor should run faster. For the K6, they did. Stretching the K6-2 to 550 MHz was miraculous, AMD did an incredible job manufacturing it at those speeds considering the design. On .18 it was easy. My point is while Intel and the K7 almost double clock speeds, AMD simply enjoyed better yields. Of course, they could overclock better, and certainly could have reached 600 MHz if they really wanted to, it still represents a terrible shrink vis-a-vis the Athlon and Pentium II/III.

I can only speculate why AMD did not make the K6-III at 600 MHz. While I was in contact with AMD quite a bit at that time, it was all about the Athlon. The K6-III was not even considered for business machines, because the chipset support was so poor. AMD made their own chipset for the K7 family, and it was considered much more solid. So, I am guessing that while there is no technical reason the K6 could not have been made at 600 MHz, there was probably a marketing issue. It would probably been confusing to customers, and it may not even have been something anyone wanted (meaning in numbers) since increased clock speed meant shorter battery lifespan. Since these were mobile processors, performance was generally not as important as battery life. So, there may not have been a market for it for laptops. Since the chips were not sold for the desktop, power was always a major factor in the attractiveness of the processors.

I guess one could ask why they never moved the .18 K6s to the desktop. I think that would be easy to answer with greater accuracy. It would confuse customers, and they probably wanted to get away from the platform since it was so weak. The chipsets for the Athlon were so much more stable, and back then being known for reliability was very important for AMD since they were not as well known as they are now. So, I think they were happy to see the Super 7 platform fade into obsolescence.

Keep in mind, none of this came from my talks with AMD. It is pure speculation, I do not know anything that you do not.

Oh, on a different note, I found a really interesting store today! They still have Pentium Pro 200's, with 1MB cache!!!!!!! Brand new. They have all sorts of cheap/old processors, all new. K6s as well. I never thought I would see a Pentium Pro 200 1MB processor new in my life. These bad boys went for over $2000, and are extremely rare. If anyone is interested in the store's URL, let me know and I will post it. I have to find my wallet and order the processors before I do though :P. I still can not believe I found these things. I have been looking for them for a long time. Now to find some motherboards :P.
Jim
K6'er Elite
Posts: 1745
Joined: Wed Jan 21, 2004 7:10 pm
Location: Toronto

Post by Jim »

I have two of the Pentium Pro mobos, even have fans. - Just QDIs though. Brand new unused. Part of the four ton truckload of stuff that I bought from QDI. - Do I want a processor for them? No Way. You are giving a much better explanation of your reasoning here than you did before. I nonetheless think you are wrong about ACZs. Perhaps they were not as desirable as laptop chips, where as you say battery life is a major factor; but they are much better for desktop purposes and it ain't just the rare bird that is. Darn near all of them are. Which leads me to suspect that perhaps they are the missing 600s. (Respecified as 450s so as to enable the sale of all of them, defective or not) As to the no 600s so no 6x multiplier, that is a chicken and egg argument, which comes first? Is a processor manufacturer supposed to produce a new processor then wait a couple of years for the various periphal manufacturers to get their act together? I would have thought that this was all done in close consultation with each other, so as to have it all hit the market simultaneously. There is no question that the K6-lll+ was and is capable of 600 Mhz. If therefore AMD did not release it, I would have thought that it was because someone else was unable to produce the necessary complementary components in such a manner as to meet the required standard of reliability. Given my experience with mobos, they are my lead suspect. Your comments about "Chipset Support" narrow it down even further. Bottom line is it was not the K6-lll+ that caused Super7 to end at 550, it was the mobos. I have had several that were not stable at the top range of their rated speed, and it seems you have too.

Edit: This is bugging me to the point where when time permits, I am going to underclock the ACZ on this machine, and we shall see how well it runs at lower speeds and reduced voltage. One thing I do know; it takes the 5 ACZs that I have had .1v over their rated 2v to clock an additional 150Mhz. That would seem to indicate that they don't really need 2v to reach 450Mhz; and I very much doubt if ATZs are any better. If they were, they would cost more.
Superpuppy 3
K6-3+ 450 ACZ (6x100)
DFI K6BV3+/66 Rev B2 (2 Meg) w/ 2x28mm Chipset Fans
2x256 Meg PC 133 Hynix SDRAM
1x 20G Maxtor (7200)
2x 80G Maxtor (7200) Ducted w/ 2x486 Fans Mount
52/24/52/16 LG CDR/RW/DVD
8/4/3/12/24/16/32 LG Super Multi
ATI 9000 aiw Radeon AGP
SB Audigy 1 MP3 Sound
CMD 649 IDE Controller
NEC USB 2 Card
TA152H
Junior K6'er
Posts: 47
Joined: Tue Jul 15, 2003 5:49 am

Post by TA152H »

Jim,

With regards to the 600s, AMD could easily have released those chips if they wanted to, since they were indirectly supported by all the motherboards since they remapped 2x to 6x. Now, 5.5x was the big problem as I mentioned, but that was AMD's fault since they told motherboard makers they were not going past 5x. I still consider getting a .25 K6-2 to 550 MHz an amazing achievement, and so did AMD since they never expected to do it. In point of fact, the K6-III+ is easily used in desktop motherboards despite AMD not supporting it at all, and motherboard makers making no effort to support the processor since AMD said it was mobile only. So, if AMD were inclined to make it, how can you say motherboards were what held them back? The motherboards that ran fine at 550 MHz, also ran fine at 600 MHz, because of remapping 2x to 6x. So, obviously AMD could have done this easily, they did not want to.

One thing you seem to disregard more casually than you probably should is the politics of the industry. A manufacturer is very, very concerned about confusing a customer and making an unclear product line. They go to great pains to avoid this, and even so fail (particularly Intel has in the past). I doubt very much AMD wanted to have the two lines close to each other, or even remotely competitive. I would guess this is why the K6-III+ never saw desktop machines, or was never supposed to. K7 for the desktop, K6-II+/III+ for the mobile market. Very clean, very easy to understand. You start making some powerful K6s for the desktop, now customers are confused. Hmmm, this K6-III runs integer apps as fast as the K7, and runs floating point really poorly? What does that mean? Which should I get? What is floating point? etc... I think that, and the fact the platform was so poor were the main reasons for AMD wanting to quickly kill the K6. No one seriously considered the K6 for anything but a home computer, and then preferably someone else's. Not that the K6 was not a reliable chip, but the platform was horrible compared to Intel's, because of VIA and ALI and their lousy chipsets.

If you notice now, AMD tried desperately to avoid confusion with their newest lines in a similar way. Once they started pushing their Athlon 64 to the mainstream market, they "discontinued" the Athlon XP, without actually discontinuing it. Still, they wanted to eliminate confusion between Athlon XP and Athlon 64, so they relabeled them Sempron and changed their PR rating. When they added a K8 processor to the Sempron line, they castrated it so it only ran in 32-bit mode. All this is to have clearly defined processor lines that do not confuse customers. FX line for the ultimate performance, Athlon 64 for excellent mainstream performance, and Sempron for the value segment. But, like I said, everyone screws this up to a degree. They now have these x2 processors and the FX line. One runs with two cores, and one runs at higher clock speeds. Kind of an unclear choice, but AMD is trying to say that FX is the gaming platform, and x2 is better for business. People will believe this oversimplistic rubbish, and that is all that matters. Everyone knows which processor they should buy, even if it is not quite correct. Oh well.

Despite your dislike of the Pentium Pro, I should mention that you can get Pentium IIIs very cheap at this store. For example, a Pentium III 733/133 is only $19!!! You can get an 833/133 for $26. This is a great deal compared to the much slower K6 family, and you get a stable platform on top of it all. I wish I had seen these prices before I bought my VIA now :P. Hey, maybe that will make a good header for another thread. Coppermine, the perfect replacement for the K6! I'd do it, but I'd need an asbestos suit I'd get flamed so often. Still, I can not believe the price/value of these processors. About two weeks ago I bought a 1.4 GHz Tualatin for around $160. I got poor bang for my buck vis-a-vis these cheapos. Guess that is what they mean when they say "live and learn", or for that matter "sadder but wiser".
Jim
K6'er Elite
Posts: 1745
Joined: Wed Jan 21, 2004 7:10 pm
Location: Toronto

Post by Jim »

I can't seem to get this point through to you. "The motherboards that ran fine at 550 also ran fine at 600 because ..." Sorry not true. A lot of them are unstable at their higher end speeds. See a post by a guy tagged "Nauctrnl" who is running his Compac Presario 1685 at 550Mhz, (no problem), but it craps out at 600. And this is a K6-2+ 550 ACZ we are talking about. I have 2 of them; and they do 600 so easy it is pathetic. We are talking mobo here, says me. You want my guess, some diddely little thing just gets further and further out of spec as the speed and temp go up. (Or alternately the mobo just doesn't support 600 Mhz). Hell I am no expert, I am reasonably intelligent; and I have some hands on experience, enough to know that a lot of boards don't cut it at higher speeds. That 4 ton truck load that I bought from QDI, was a mix of unsold old stock and RMAs. Trust me, I do have a fair bit of experience in benching boards.

PS. I have nothing against the Pentium Pro; but as I understand it, (correct me if I am wrong), it is a 200Mhz processor. My machines are old working machines. They have to be able to play DVDs, surf the web, and any of the various other things I do. I don't think a 200 is up to that. If you want my boards, you are welcome to them, (no charge), because I will never use them. I also have a selection of various types of old processors that you are equally welcome to, if you are interested in them from a historical standpoint; because they have very little utility these days.
Superpuppy 3
K6-3+ 450 ACZ (6x100)
DFI K6BV3+/66 Rev B2 (2 Meg) w/ 2x28mm Chipset Fans
2x256 Meg PC 133 Hynix SDRAM
1x 20G Maxtor (7200)
2x 80G Maxtor (7200) Ducted w/ 2x486 Fans Mount
52/24/52/16 LG CDR/RW/DVD
8/4/3/12/24/16/32 LG Super Multi
ATI 9000 aiw Radeon AGP
SB Audigy 1 MP3 Sound
CMD 649 IDE Controller
NEC USB 2 Card
TA152H
Junior K6'er
Posts: 47
Joined: Tue Jul 15, 2003 5:49 am

Post by TA152H »

Jim,

Every motherboard I have tried works fine at 600 MHz, if they work at 550 MHz. But, it is certainly possible that some do not. However, there would be more than enough that do, as most people on this site could tell you. The way it works is the CPU makers tells the motherboard makers the processors they are making before they are out, they send specs and samples, and then the motherboard makers make whatever updates they need to. This is what happened at 550 MHz when AMD surprised everyone with the K6-2 at 550 MHz and did not give the motherboard makers enough lead time. Consequently, proper motherboards were not out much in advance of the chip. If AMD had told motherboard makers they were going to make a K6-III+ for the desktop and at 600 MHz, it would have been exceedingly simple to support it. So many do even without AMD mentioning the chip, and the few that may not would require very minor revisions since the motherboard runs essentially the same. My guess is there is absolutely no difference between 600 MHz and 550 for a K6-III, since the problem with the original K6-2 550 was supplying stable power at that high a voltage. Still, I am not 100% sure on that, so I will not rely on that as a fact. What is fact is that plenty of motherboards ran them fine without any effort, and it should be clear that even if there were motherboards that failed at 600, they would required very little change.

Also, this would be the first time a processor was not released because of motherboards. It is just not the way things are done, you have the tail wagging the dog. The processor maker tells the motherboard makers what they are coming out with, and they support it. If they do not, you can be pretty sure a few of their competitors will.

With regards to a K6-2 550 ACZ+ not doing 600, it is very possible it is the processor failing and not the motherboard. Not all processors, even the ones named the same, are created equally. If they were, yield rates would be 100%, and all processors would overclock the same. I have had wildly different results with the same processors more than once. Still, it could be the motherboard too, my point is you do not know until you actually run a chip at 600 MHz on the motherboard with the same cooler, etc...

With regards to the Pentium Pro, you are correct they stopped at 200 MHz. They also do not have MMX, although there are plenty of dual processor motherboards which help somewhat. Intel never went higher to avoid competing with the Pentium II line, which they ran faster than at the same clock speed (faster L2 cache being the reason).

The Pentium IIIs might be worth looking at for you then. They have very nice performance and are very cheap. I ran Dungeon Siege on my K6-III+ at 550 MHz and it ran quite poorly. The Pentium III 667 I have ran it so much better, with the same video card, it shocked me. Same thing with Civ III, both of which are nearly unplayable on the K6-III+, whereas Dungeon Siege is very playable on the Coppermine, and Civ III is slow, but not hair-pulling slow. Civ III is laggardly even on my Athlon XP 1700+ though, although noticeably quicker than the Coppermine at 667.
User avatar
tazwegion
Veteran K6'er
Posts: 349
Joined: Sat Jun 21, 2003 11:13 am
Location: Victoria, Australia

Post by tazwegion »

Jim wrote:I give up. You still seem to think that socket 7 mobos support 600Mhz. How? 5x120Mhz? Isn't that a bit warranty voiding? How pray tell can you say that mobos supported 600 when their highest multiplier ever was 5.5?
Sorry to enter the frey late in the picture... :P

Jim, I own an IBM PRO263 SS7 SiS 530 chipset (as used in the 21xx series Aptiva's), and strangely enough it HAS a 6 x multiplier (on mobo)... though only a max fsb of 100Mhz :(

While currently out on loan (to be upgraded next month) I'll endeavour to get a 'digi-shot' of the m/boards jumper region ASAP, BTW this is the ONLY m/board I've ever seen with a 6x multi 'on-board' :banana
Image
Jim
K6'er Elite
Posts: 1745
Joined: Wed Jan 21, 2004 7:10 pm
Location: Toronto

Post by Jim »

Well then I was wrong about that. Apparently somebody did put up a 6x; but I had never seen one. Conclusion being, just cause you ain't seen it, doesn't mean it doesn't exist. Thanks for the lesson.
Superpuppy 3
K6-3+ 450 ACZ (6x100)
DFI K6BV3+/66 Rev B2 (2 Meg) w/ 2x28mm Chipset Fans
2x256 Meg PC 133 Hynix SDRAM
1x 20G Maxtor (7200)
2x 80G Maxtor (7200) Ducted w/ 2x486 Fans Mount
52/24/52/16 LG CDR/RW/DVD
8/4/3/12/24/16/32 LG Super Multi
ATI 9000 aiw Radeon AGP
SB Audigy 1 MP3 Sound
CMD 649 IDE Controller
NEC USB 2 Card
User avatar
tazwegion
Veteran K6'er
Posts: 349
Joined: Sat Jun 21, 2003 11:13 am
Location: Victoria, Australia

Post by tazwegion »

No problem... to be honest it's a freakish m/board, the first time I came across it (in-law's computer) I wanted to OC it there and then, it's native K62-500 ran stable on the SiS chipset @ 550Mhz... needless to say I can hardly await it's return as I'm down a SS7 box due to the kids gaming system upgrade :roll:
Post Reply