AMD's K6-233MMX

Drivers: It looks like thier a new engine in town for the racing simulations with availability of the new AMD's K6-233MMX. The K6 appears to be better than the INTEL PENTIUM PRO with any measurements used. We can't wait here at the IVGA as we expect our K6-233MMX any day to complete our testing. But simply put we like the AMD K6 because: 1). 25% the cost of a INTEL PENTIUM PRO 200-MMX 2). Up to 4 Times as much cache and it starts at 512K, where the INTEL 512K as impossible to get.. 3). Currently supports MMX out of the box.. 4). Runs circles around the INTEL PENTIUM PRO. 5). With the VIA and AMD-640 Chip set will support 75, 83 and 100MHZ bus speeds 6). Supports the SDRAM. Its truly the first: 1). 300Mhz Processor 2). 100Mhz Bus 3). 100Mhz Memory Speed. Now for what a couple of other people are saying... Check out the link at the RACEROCK - AUTO-SIMULATION GATEWAY http://www.ivga.com/pjfnas2.htm Don Wilshe, IVGA =================================================================== AMD-K6(TM) MMX Processor Here's What They're Saying... "In side-by-side comparisons, it [the AMD-K6 processor] equaled and even bested equivalent Intel machines on our Winstone benchmark. What's more, the K6 is equipped from day one with MMX... Even better, the K6 exhibits none of the Pentium Pro's lackadaisical behavior running 16-bit code..." --Bill Machrone, PC Week. "It's going to be huge," said Toby Grace, president of Computer 1, a large seller of PC's and components in Austin. "The K6 is an awesome chip. Intel is shaking in its boots." --Austin American Statesman "...An Exceptionally Strong Comptetitor...." --Computer World "AMD's K6 Beats Intel...." --San Jose Mercury News "K6 Is World's Fastest x86 Chip" --Linley Gwennap, Microprocessor Report "...A serious competitive challenge to Intel..." --Los Angeles Times "..There's a ton of money to be saved by going with the K6...." --Nathan Brookwood, (Dataquest for c/net) "AMD is poised... to become the most significant alternate to Intel in 1997. At 8.8 million transistors, AMD's K6 is the most complex of the next-generation designs... With a compact, high-performance design, lots of fab capacity, strong sales and support, and multiple design teams, AMD should be in a strong position." --Michael Slater, Microprocessor Report ================================================================ Ok, Dave Sparks and Eric Busch what do you think of the K6????

>We can't wait here at the IVGA as we expect our K6-233MMX any day >to complete our testing. But simply put we like the AMD K6 because: I've been very interested in the release of the K6, but I'm going to reserve gushing over it until I know it'll run the DMA versions of Nascar 2 and Indycar 2. --

>Drivers: >It looks like thier a new engine in town for the racing simulations >with availability of the new AMD's K6-233MMX. The K6 appears to >be better than the INTEL PENTIUM PRO with any measurements used. I'm still waiting for the Merced to appear. I'm sure that you said that this was going to be available around now when you tested it ;-)) Cheers, Richard

Quoted text (11 lines)

> >Drivers: > >It looks like thier a new engine in town for the racing simulations > >with availability of the new AMD's K6-233MMX. The K6 appears to > >be better than the INTEL PENTIUM PRO with any measurements used. > I'm still waiting for the Merced to appear. I'm sure that you said that > this was going to be available around now when you tested it ;-)) > Cheers, > Richard

Richard: p7 600mhz is running fine. I hope intel releases them soon and maybe the AMD thing will give them a push.. Don Wilshe, IVGA The Problem with the P7 is you cant use sound cards or gamecards at that bus speed

( reduced ) i would like to add AMD K6 is also great for barbecue. Leave your steak on it while you play your favourite game. Steak should be ready in 30mins. ( tops )

>Let's use some common sense.. SH*T happens.. Manufacturers don't keep >up with the deadlines.. Look at Win97...now moved to 98 Common sense is probably best used in foresight rather than hindsigt. It was pretty obvious that Merced was a LONG way off when it was "tested". Cheers! John

Problem w/ p7 is that you can't use ISA soundcards or gamecards (or many modems for that matter) w/ them succesfully due to bus speed timing problems. The new generation of upcoming PCI 2.1 compliant cards of these types should fill in the gap by the time Intel decides that the P7 is ready, though. Micha

>p7 600mhz is running fine. I hope intel releases them soon and >maybe the AMD thing will give them a push.. Merced/Wilamette/P7 is still slated for a 1999 release, pretty much as it has always been. There are still two major releases (Deschutes and Katami) slated for release after Klamath/Pentium II and before Merced. Sorry, Don, but your insistence that you've tested this chip really makes me wonder about everything else you've said. Given the timescales I find it difficult to believe that anything other than a few early prototypes exist. And anyone getting their hands on those would not be so stupid as to break the terms of the inevitable NDA. I still don't know what you "tested" (and to be honest, I don't think you do either <g>), but a 600MHz P7? No way! I really would like to believe you on this one, but..... Cheers, Richard

Don, whoever wrote that is labouring under the misapprehension that is is a GOOD thing to be complex. Although both "Pentium II" and K6 are made using a CMOS process, there the similarity ends. - Intel uses a 0.28m CMOS process and AMD use a 0.35m process. This finer line geometry helps Intel reduce heat and more easily achieve higher clock speeds. - Intel's design uses a four layer process but AMD has used a FIVE layer process and a smaller die size. While this helps them minimise costs (more die per 200mm substrate), both those facts will again increase heat and be a problem for increasing clock speed. - K6 has a slower FPU than Intel. No big deal unless you're a Quake fan. - Initial reports are that MMX is slower than Intel's although based upon the same implementation (Cyrix developed their own for the M2, a lot of features of which will be in Intel's MMX2). The main problem though is speed. Although initial benchmarks show the K6 as being equal to or slightly ahead of Intel, the benchmarks don't address MMX effectively and so cannot give a true "like with like" comparison. Regardless of that, the point we come back to is that the one advantage the K6 has over Intel is pric. That's it, pure and simple. Now the price can be changed with a flick of Gordon Moore's pen, so where would that leave AMD? From the enormous profits of 386, 486 and Pentium, Intel easily have the finances for a prolonged price war, and what's more they have around twenty well established production facilities churning out top quality devices. AMD.....don't. they have neither the finances nor the fab space that Intel have, having cancelled a lot of investments during the market slump in 1996. Well, I don't know about them but I'm certainly*** on to see what Intel's riposte will be before declaring my undying love for all things AMD. Things could get pretty interesting between these two, and even more so when Cyrix join the party. I hope it will lead to lower prices and three equally competing products but somehow I can't see it. What AMD will most likely do is stop Intel charging the customary enormous premium for newly introduced MPUs, but as to whether they'll make any real market penetration is waaaaay up in the air at present. Cheers! John

John: Thanks for the input, But I thought TOM HARDWARE reported the FPU on the K6-200 faster than the INTEL.. I might be wrong as usaul.. Don Wilshe, IVGA

Quoted text (7 lines)

> John: > Thanks for the input, But I thought TOM HARDWARE > reported the FPU on the K6-200 faster than the > INTEL.. I might be wrong as usaul.. > Don Wilshe, IVGA

John Wallace: I am Sorry: I went back and read the article and the FPU is not as good.. Your right and thanks for input!!! Don Wilshe

(stuff deleted) Interesting post, John. Have you read the comparison on Tom's Hardware Page? I just skimmed over it (I have a passing interest in this, but there is just too much of a furor to get much in the way of straight facts), and I thought I saw him mention that the K6 beats Intel at MMX using Intels own MMX benchmark. I play Quake, but my main concern with the K6 will be its ability to run N2 and ICR2 in DMA modes. Heard anything on that? --

On Tue, 08 Apr 1997 15:17:00 GMT, "John Wallace" >- K6 has a slower FPU than Intel. No big deal unless you're a Quake fan. I really think that you're going to see the vast majority of 3D games using the FPU heavily, now that floating point math is faster than fixed point (integer) math, at least on the Pentium and PPro (which together represent the vast majority of the CPU market...) Discounting FPU performance as "only relevant to scientists" is anachronistic... ----Jim Sokoloff

Richard: All you have to do is call me.. You want to see one!!! Don

No thanks Don, Fab-10 is a lot closer. I'll ask Intel next time I speak to them ;-) Cheers! John

It's not as good, but at least for now it's a truckload cheaper so worth getting e***d about. Still, Intel quarterly price cuts are due soon, let's see what they do. AMD won't hit the market yet in volume so Intel will probably maintain their higher pricing. Expect fireworks when AMD get Fab-25 ramped and start shipping this puppy in volume. Cheers! John

Quoted text (13 lines)

> >John Wallace : I am Sorry: > >I went back and read the article and the FPU is not > >as good.. Your right and thanks for input!!! > It's not as good, but at least for now it's a truckload cheaper so > worth getting e***d about. Still, Intel quarterly price cuts are due > soon, let's see what they do. AMD won't hit the market yet in volume > so Intel will probably maintain their higher pricing. Expect fireworks > when AMD get Fab-25 ramped and start shipping this puppy in volume. > Cheers! > John

John: I am now into another performance group and they say take the hundreds of dollars in saving and buy a VOODOO based assist board and you will really have the best combo... What do you think... I am testing my first K6-233.. Looks like I can push it fine to K6-262 at 83mhz don wilshe

>Did you read the article about the PII from PC Week? They benchmarked >a PII 266Mhz, it WAS 14% faster than the K6 233!! BUT, here's the >catch, if you do some calculations with the clock it comes out to >%14!!!! Something to think about!! Make sense?? >In other words, it's just the clockspeed helping out the PII.. Then the real question is "Does anyone really care *why* their computer's CPU is faster?" I sure don't... They can make it faster by increasing the clock rating, by increasing parallelism, increasing cache, whatever. The bottom line to me (and I presume to over 99% of the desktop CPU market) is "how fast is the chip vs how much it costs?" The underlying technology, while interesting, is completely secondary. ---Jim

v's compatability, v's "what's around the corner", v's probable lifetime, v's required speed, v's disposable income, v's advertising hype, v's, v's..... There's an awful lot to take into account over purely speed and cost. On that premise (and the demand of games set apart) a 486-133 would power virtually every PC on earth. Cheers! John

>Don and John.. Here's some info.. for ya!!; ) Remember this is with >GLQUAKE w/ Voodoo... Oh yeah, goto the web site.. >>Just noticed new K6 test results in which K6 200 beats P5 200 in >>Quake. >>Both run 3*66 (P5 was overcloked). This test was made with 3dfx Voodoo >>enchanced system: Correct me if I'm wrong but 3x66 is NOT a 200Mhz system as claimed above. That aside, what the test appears to show is that when the burden of the slow FPU is taken from the K6 and passed to the Voodoo, it is to all intents and purposes the equal of a P5? There will undoubtedly still be some overhead from the FPU, but in any case I think Quake isn't a good test. No-one uses GLQuake for proper ***, you can't use it on QuakeWorld and using it on a LAN game will get you creamed against a decent player - it adds a fair bit to your latency, enough to make a difference. Anyway, time to go watch the closing holes at Augusta. Wonder if a European will do it again this year? ;) Cheers! John

<snip> Is Tiger Moving? Lol :-]

Quoted text (10 lines)

>>The bottom line to me (and I presume to over 99% of the desktop CPU >>market) is "how fast is the chip vs how much it costs?" >v's compatability, v's "what's around the corner", v's probable >lifetime, v's required speed, v's disposable income, v's advertising >hype, v's, v's..... >There's an awful lot to take into account over purely speed and cost. >On that premise (and the demand of games set apart) a 486-133 would >power virtually every PC on earth.

That's all true, but I was referring to a previous post that implied that speed gains through increasing the frequency rating (clock speed) where basically "cheating" :-) Side Note: IMO, the 486 has been so left behind in speed, that what it costs is almost irrelevant. ---Jim

>Correct me if I'm wrong but 3x66 is NOT a 200Mhz system as claimed >above. Well, since no-one else noticed I was on the loony juice the other night, I'll correct myself and point out that of course 66x3 _is_ 200Mhz (if calculated as the proper 66.6666Mhz). Gotta cut down that wine consumption...

>Correct me if I'm wrong but 3x66 is NOT a 200Mhz system as claimed 3x66 = 198, for all intents and purposes it is a 200 system. GLquake doesn't add to my latency over my ISDN line. I use GLquake to play netquake quite a bit. Of course I use QW more often now, but as soon as GLquakeworld is out, I will be using that also. Also, on news.3dfx.com there have been numerous posts about GLquake with the K6. From the posts there the K6 is _slower_ than an equal clocked Pentium in GLquake. The K6 also has much slower "wizmark" scores which is 3DFX's own benchmark program to test 3D cards. It is quite a bit slower in Wizmark actually. Wizmark uses "Glide" which is the native 3DFX API. Glide is heavily FPU based. You can figure the rest out on your own. rj -- sig? I don't need no stinkin sig

> 3x66 = 198, for all intents and purposes it is a 200 system. Actually 66.666Mhz x 3 is 199.998 - pretty much exactly 200Mhz. I corrected this in a previous post and apologised for my alocholic excesses! I'm not too sure what I was thinking when I wrote that, because I even remember double checking it at the time. Wierd... Yes, I think probably your ISDN line is too slow to be able to sense the difference through gameplay, although I guarantee you that it DOES make a difference. I play either over a LAN (four NICs directly connected) or over the internet using a firect SuperJANET connection, and GLQuake adds 33% to the latency. Now this seems like nothing when you're talking about a latency of 33ms compared to 44ms with GLQuake, but every time I play Gib_McNailz (a good friend and clan member) we are never more than a couple of frags apart. As soon as he switches GLQuake I whip his ***every time, easily beating him. Once or twice can be ascribed to luck or coincidence, but this happens time and time again. I'm running P-200MMX at 225Mhz so it's certainly not because the system can't handle it - the change is purely down to the addition of GLQuake. In deathmatch response time is infinitely more important than pretty graphics, far more so than in sims, and if GLQuake is perceptably slower than normal Quake then only losers will use it. So what are you saying? The K6's FPU performance versus Intel's is well known by now, this is no surprise news. However, remember that AMD and Intel both use the same crippled implementation of MMX, which restricts switching between MMX based instructions and FPU. With all of the hype about MMX and marketing possibilities of sticking MMX on a box, more developers will be pushed toward MMX than FPU. After all the arguing has been made, what is left is that the K6 is going to be sold on price, and price alone. Intel currently has some 85% of the WORLD microprocessor market, and because of that they are always going to be considered the standard for compatability, speed, desirability, whatever. Anything else is a second-best compromise and will be chosen based upon price/performance ratio. Comparing the performance of the K6 against the equivalent Intel CPU is meaningless if you don't consider price. Look at the K6 against the equally priced Intel offering and then come back and complain if it's slower. Cheers! John