Slashdot Mirror
Yet Another K6 Series From AMD
EricFenderson writes: "AMD has released the K6-2+ and the K6-III+ for use in notebooks. It's made with a .18 micron process, has on-die cache, and has new power-saving technology called PowerNow. Reports say it's only being sold to companies integrating notebooks, but Tom's Hardware says it would be great as a system upgrade, and AMD should also be selling to retailers. Petition?" If you're nursing along a Socket 7 system especially, this series should be intriguing. The long-awaited notebook chip is noteworthy for having 32 power levels to choose from - maybe those all-day notebooks will really appear.
Since the Crusoe launch, the one of the most important things for notebooks is power consumption. While the Intel chips are hogs for power/waste heat, the Transmeta chips run cooler, albeit on a totally different archetecture.
Where are these new AMD chips compared to the above lot? Between the 2? The same as Intel?
From my POV, a laptop that can compile a piece of code in 5 minutes instead of 10 is of limited use if the battery dies after 30 seconds
These people look deep within my soul and assign me a number based on the order in which I joined. -- Homer Simpson
I completely concur with respect to the FUD that intel loves to spout...
I hope nobody forgot their ridiculous advertising scheme about "internet-power" chips. So idiotic a tale it's actually funny. (Microsoft FUD at least makes sense sometimes).
Or the hopeless Pentium 1GHz... anyone ever seen one? I've never seen one. The Gigahertz pentium is a paper monster.
Intel still has not clean up that disaster, and they now, to show their technical prowess, bring out a 1.13 GHz monster - in Slot1. I would be positively surprised if they can deliver this one, but even more so, I'ld be VERY surprised.
What also frightens me is the adoptation of SSE above 3dnow! - this paints the athlon in a needlessly bad light.
Then there is the absurd Rambus deal with intel. You should look at their "divine intervention" benchmarking data - anything less than divine intervention won't explain these results. Intel, for shady reasons, fully supports the Rambus humbug. It looks like silicon politics to me, and if I buy an intel system, I really can do without Rambus being rammed where the sun don't shine...
Someday, perhaps intel will have a real rival, but at the moment, I hope they get slaughtered, for I really don't like their "business model". They seem to think the consumer is lobotomized.
Until that attitude changes I am not buying any intel system.
The K5 and K6 units have pretty good integer performance, but the floating point blows. The numbering scheme isn't too difficult if one has followed the chip's life, but if you haven't, it doesn't really make much sense.
k5 = Socket 7 P54C pentium compatible - 16K instruction cache, 8k data cache
K6 = Socket 7 P55C(mmx) - 32K instruction cache, 20K predecode cache, 32k data cache
K6-2 = Super Socket 7 (mmx, 3dnow, 100MHz bus) - 32K instruction cache, 20k predecode cache, 32k data cache
k6-III = Super Socket 7 (mmx, 3dnow, 100MHz bus) - 32K instruction cache, 32K data cache, 256K on-chip full-speed L2 cache
Mobile K6-2+ = Super Socket 7 (mmx, 3dnow, 100MHz bus) - 32K instruction cache, 20K predecode cache, 32K data cache, 128K full-speed on-chip L2 cache
Mobile K6-III = Super Socket 7 (mmx, 3dnow, 100MHz bus) - 32K instruction, 20K predecode, 32K data, 256K on-chip full-speed L2 cache
The fastest the chip can go right now is 600MHz (6x100MHz).
The K6-III and the Mobile K6 series can all use a motherboard-based L3 cache of up to 2MB.
Hopefully this will shed some basic light on the K6 family.
An example of one of these boards is the A-Open AX59-Pro, which is one I've been using for quite some time. It's supposed to support processors up to 800 MHz (984 MHz, if you overclock the bus).
Actually, AMD's naming scheme follows Intel's, but is a bit more consistent. A quick run-down of roughly comparable chips:
AMD Intel
K5 ~ Pentium (aka P5)
K6 ~ Pentium Pro (aka P6)
K6-2 ~ Pentium 2 (second rev of P6)
K6-3 ~ Pentium 3 (third rev of P6)
K7 ~
K8 ~ Merced (first IA-64)
So, with AMD, the naming scheme is Kx, with x being the generation number of the core, like x86 used to be. The numbers following the "K6" line represent revs within a generation. Intel started that trend and AMD followed it.
Things are a bit complicated by the fact that Intel doesn't *have* a new core to compete with the Athlon. Intel hasn't made a new core in years. The IA-64 was supposed to be here by now, but it isn't. In the meantime, Intel has been trying to keep up by rev'ing it's Pentium Pro core, which dates back to ~94. While not inspiring, it's impressive that they've been able to squeeze out as much performance from the aging P6 core as they have.
Another complication is the fact that the Pentium 2 and 3 date back to the Pentium Pro design, *not* the "classic" Pentium design. Further, the Pentium Pro was a completely different core than the Pentium -- not just a Pentium with added frills. This confuses quite a few people.
The industry is complicated, so it takes some effort to follow. Naming is the least of worries when comparing processors.
hope I shed some light,
--Lenny
slotA.com's motherboard list has a dual Athlon motherboard listed. It's named "Tyan Dolphin" and should be released Q4 '00.
I don't know if this is based on information from Tyan though - I couldn't find any information on this on Tyans page.
--
Ner lbh sebz gur HFN? Gura lbh'ir whfg ivbyngrq gur QZPN!
Maybe this is also another step in laptop upgrades. My previous celeron laptop was like this. I could upgrade from a 333 to a 433 painlessly. Granted it's only 100mhz, but it is still an upgrade that put a little more life in my investment.
(I also have an all day not books made from plastic, metal and glass, its called a PalmV and I can take notes on it all day long).
Thad
Thad
The pentium was Intel's fith generation chip. It came after the 486 and is effectively a 586, but they decided to give it a real name for marketing reasons (and the fact that you can't trademark a number). But then they came out with the PentiumII, which is their sixth generation product. It should really have been called the Hexium, but I guess that name isn't very good from a marketing perspective ;-) . To add to the inconsistency, the PIII is still based on the same sixth generation core as the PII, with just a few minor revisions. Then they came out with the Coppermine (that was its codename), which was a bigger revision of the PII core than took place with the PIII, but it didn't even get a new name! It's called a PIII, just like the earlier PIIIs even though there's a significant difference. Now we have the PIV coming up, and that's a seventh generation design. So does increasing the number after "pentium" indicate a revision of a core or a new core? The answer: it depends on Intel's mood at the moment.
With AMD it's clear as day. The K5 is their fith generation chip. The K6 is their sixth generation chip. The K6-2 is a revision of the K6 core, and the K6-3 is another revision. These revisions roughly corespond to what Intel did with the PIII and Coppermine, respectively. The Athlon is AMD's seventh generation chip, and it was codenamed the K7, but they decided to give it a name (again, for marketing). The K6-2+ is a revision of the K6-2. K6-3 was taken, and calling it a K6-4 would be misleading because it really fits in between the K6-2 and K6-3. There's also the K6-3+ which does the same thing to the K6-3 as the K6-2+ did to the K6-2.
I am not an idiot. Please use my name to email me.
"That's right, I'm quoting myself."
-Upsilon
After the 386 there was the 486. After the 486, Intel changed their naming scheme ( I remember hearing something about not being able to trademark numbers or something) to PENTium. To illustrate the similarity between the chips AMD released the K5. PENT=5...
Pentium 2, or in other words ((Pentium 1) + 1), if the K5 is the same as the (Pentium 1) then the ((K5)+1) would be the K6. The Pentium III or the ((Pentium 1) + 2), would be the same as the (K5) + 2) or the K7.
Although they moved up to the K7, they continued production of K6 based chips, K6-2, and K6-3 are newer revisions of the K6 core. Just don't ask me why they chose the letter 'K'.
I don't understand your confusion.
LK
"Hi. This is my friend, Jack Shit, and you don't know him." - Lord Kano
Unfortunately, another problem that plagues the K6 series of CPU is that most versions of the CPU has the L2 cache on the motherboard, which means sluggish memory access performance depending on application.
Because the L2 cache on the Celeron runs at full CPU speed and is on the CPU die itself, the result is much improved memory performance. Now you know why AMD built the Duron with its on-die 64 KB L2 memory cache in addition to the 128 KB L1 memory cache.
Raymond in Mountain View, CA
Intel is out of this market - when was the last time they made a socket 7 processor? Pentium I. Their only viable processor for laptops is the Celeron, due to the PII/PIII large power consumption (even the low power models)
the battle for laptop/notebook processors is now between AMD, with its range of low-power processors, and Transmeta, with the Crusoe.
fross
Not proven? I still love how people still eat the Intel 'AMD is unproven and reliable FUD'. It's sad, they see though Microsoft's FUD to Linux and co, but Intel to AMD....
Prove, with links, NOW, that Intel's processors are more reliable and proven compared to AMD's. No, I don't care about overclockability, or the speed of the chip. I care about them being unstable and reliable.
Until you can prove otherwise, you are talking FUD. I hope you like advertising for Intel.
I work for an OEM, and one of the internal arguments against AMD is that they have no equivalent to Intel's Embedded Systems Group. Under this program, Intel designates certain products (CPUs, Chipsets, etc.) to be supported for at least five years.
Many folks designing embedded systems have platform-sensitive applications, and will not design products around platforms whose product lifespans are measured in months.
The extension of the K6 line brings its total life to nearly 3 years (and counting). This effectively removes the primary argument against AMD for embedded system development.
J.
Information wants to be Free. Useful Information will cost you.
I remember reading about the K6-2+ and the K6-III+ on The Register several months ago and yet it only just appears on Slashdot now!
Even in the pics on Toms Hardware the chips are marked "(c) 1999 AMD". It was my impression that K6-2+ chips had been sold in notebooks for several months now?
Don't get me wrong, I think the K6-x series of chips are great, low-cost, solutions (Hell, I run two at home!). I suppose it does just go to show that Socket 7 isn't dead after all.
I wonder how much Intel regret pulling out of Socket 7 with the Pentium 233MMX three years ago?
It's weird - at 233Mhz Intel declared Socket 7 a dead-dog and moved on to the abhorrent Slot 1 platform, yet AMD, Cyrix, IBM, IDT and Rise all managed to come up with processors perfectly (or imperfectly, in the cases of Rise, IDT and a lesser extent Cyrix/IBM) suited to the low-cost budget PC market which are still holding their ground today against Intel's Socket 370 line up.
--
jambo
system.admin.without.a.clue
-- js.
A quick search on Pricewatch:
AMD K6-2 550: $76
AMD K6-2 533: $66
AMD K6-2 450: $57
Intel Celeron 566: $98
Intel Celeron 533 PPGA: $92
Intel Celeron 500 PPGA: $91
Intel Celeron 466 PPGA: $82
Intel Celeron 433 PPGA: $75
And actually, the prices of the K6-2 chips have gone up in the past 3 monthes (I think because they stopped making them). I got a K6-2 500 in May for $42 +S&H.
But my point is to get equal price to a K6-2 550, you need to drop down to a Celeron 433. Now for the overclocked, that's probably not a big deal. Also for the power hungry, K6-2 is a joke. But if you're looking to make a very cheap system, K6-2 is probably your best choice. Makes a great linux/FreeBSD server.
The K6-2+ series (which this article is about) have 128K of full speed L2 cache on the chip. They also support L3 cache on the motherboard, so their cache performance is excellent.
.18 micron process, almost 100% of these chips will overclock to 600 Mhz. And all you have to do is set a couple of jumpers on the motherboard. For cheap overclocking, these chips are great.
Also, since the K6-2+ is manufactured with a
The only problem I see is availability. I've only ever seen two of these chips, and I see a lot of AMD processors. If you have a Socket 7 motherboard, and you see one of these chips, I'd highly recommend getting one.
--- "So THAT's what an invisible barrier looks like!" - Time Bandits
The k6-2+ (500, 533, 550) was introduced in June. Tom's Hardware published their test in mid July. These new chips have several voltage/speed settings so you can choose the preformance/power consumption tradeoff that makes sense to you. One of the power settings is adaptable.
At low power they only use 3 Watts (Celron 500's use 28 Watts). At full speed they use 16 Watts and are roughly the same speed as a similarly clocked Celeron.
These new chips have the same amount of L2 Cache as a Celeron and are therefore faster than the old K6-2's.
Old Socket 7 chips use less board real estate than newer chip formats (the mother boards can be smaller and cheaper)
I'd love to have one of these for an old super7 system - $100 to make it relatively up to date.
All generalizations are false, including this one. Mark Twain
The K6 competed with the P55 series Pentium chips but (IIRC) lack MMX/3Dnow support and SMP capability. This was a socket 7 chip.
The K5 replaced the P54 series chips (Pre-MMX, Single voltage Pentium 75-166) and was a socket 5 chip. This was the last AMD chip to be marketed using the old confusing "PR" rating system.
AMD does not have a monopoly on confusing chip names. There are actually at least three distinct Intel chips that bore the name "Pentium" (this does even consider the "mobile pentium" line):
The AMD K6 series of chips generally outperform their Intel counterparts at the same clock speed in all aspects except floating point. Considering that an AMD chip typically cost about the same as an Intel chip two steps slower (or even a bit less), the small loss in FPU performance is not significant. Looking at one current ad, I see that a K6-2/550 is going for $99, and a Celeron 533 is $129. The K6-2 performs better for 30% less.
The situation is even more in AMD's favor when you look at the Athalon. The Athalon fixed the K6's worst flaw - poor FPU performance. Clock for clock, Athalon beats (or at least matches) P-III across the board. Considering that that the current price for a PIII-600 is $315 and an Athalon 600 is $175, there is no comparison. An Athalon 700 is going $239, the same price as a PIII-500.
"The axiom 'An honest man has nothing to fear from the police'
Why is it that the proponents of "one nation under God" are so eager to get rid of "liberty and justice for all"?
Which version of chip are we meant to take this as? Intel have sensibly numbered pentiums 1,2,3,4,... but AMD seem to insist on going for 5, 6, 7, 6-2, 6-3,....
So is a K6 meant to be equivalent to a Pentium VI, or is a K6-III meant to be equivalent to a Pentium III? How are we meant to make sensible comparisons with such confusing numbers?
Some say Socket 7 is an old technology yes this is true, but does that make bad no.
Just because AMD plays to people who like to keep their good old socket 7 motherboards
does make them stupid. They know alot of people do not like change and/or
Do not like the big boys (Intel) so this is the crowd they are playing to.
I call that marketing.
I would like to see the first laptop with
smp enabeled. Who cares about battery lifetime...
I want a dual Athlon notebook!
that would be fun;)
I think the issue right now is that the K6 series of CPU's is already obselete even compared against the Celeron.
The reason is simple: weak FPU performance. I wouldn't recommend the K6 series if you're doing any programs that requires lots of CPU and FPU processing power, programs such as CAD/CAM and image-processing software.
Besides, with the price of Celeron CPU's being so cheap and the arrival of the AMD Duron (which has none of the slow FPU problems that plague the K6 series), you'd be better off with today's CPU's anyway.
Raymond in Mountain View, CA
Trading at only 10x expected earnings of $6 or $7 this year AMD is a huge bargain basement buy. Just its flash memory business alone, growing at 70 to 100% a year, is worth its stock price.
I don't like these K6 chips. As a stockholder we don't make hardly any money from selling them... They will be only 5 to 10% of production by 1Q2001. Mobile Athlons are coming.
Maybe we should paper launch them today. Let's announce a 1.5 ghz mobile Athlon, shipping in "limited" quantities (say 9 or 10) today!
Cheers,
Fox
Not that it matters, at all, but I have a VA-503+ w/a K6-III 400 and a TNT2. Aside from some problems with 2x AGP (switching down to 1X fixed the problem and I never noticed the difference anyway), I have had no problems with it and am still using it a year later. I would say that I am very happy with it, given the price (cheap compared to what was available from Intel at the time).
My rules are:
Never spend more than $150 on a processor.
Never spend more than $100 on a motherboard.
Never spend more than $200 on a hard drive.
Never spend more than $150 on a video card.
Never spend more than $30 on a sound card.
Never spend more than $175 on memory.
Splurge on the monitor.
I follow these rules every year and a half or so when upgrading, and have never been disappointed.
(although, I only had to splurge on a monitor once, which was the Iiyama Vision Master Pro 17 that I bought for $800 in 1996, and am still using today)
i wish there was a chart for the K6-III+. these are the chips that will be the celeron notebook killers. well, i guess there's the durons.
of course, i'd only get one if it came in a vaio or affordable thinkpad.
"you get hit and your head goes ping" --rocky horror picture show