Slashdot Mirror
AMD Athlon (K7) Ships
Sir-Techlot writes "AMDs wed site has a page saying that the Athlon will be shipping today (6/23/99). Tells a bit of info we already know also. "
← Back to Stories (view on slashdot.org)
← Back to Stories (view on slashdot.org)
Don't get me wrong, the K7 (Athlon) sounds like an incredible chip and is sure to make AMD some money, but is it enough to save the company? If you read AMD's press release today you know that they have larger than expected loses(200 million), so they need this chip be their cash cow. The problem is that they destroyed the market that they are shooting for with the K7(Computers between $1,299 & $2,599 USD) by selling those damned K6. The K6 (all the other low end chips) created a huge sub $1000 computer market. Why is the average consumer going to spend $2,000 for a machine when they(in their eyes) can get a compuer that does the same for under $600 from eMachines? They need to make a ton of money with this chip and I don't think the market is really there. They tried using the Christensen effect to beat Intel and it bit them in the ass. And I think Intel is looking more into the future by developing the advanced 0.13 micron technology and 300mm waffers(This will give them a cost saving of 30% alone). In the long term (3-5) Intel will win because AMD is spending all their resources (and then some) to beat Intel in the near term with their K7. Just my .02 cents. :)
So does anyone know if anyone is paying attention to this on the Linux kernel development to get support for SMP working with the EV6 multiprocessor technologies? As I understand it, on the Intel architecture, only Intel's MP spec is supported. Does the Alpha port support this MP spec and would it be easy to port over to Intel arch?
The answer's probably more complex than you'd like. For integer/pointer operations, the K7's supposed to be faster than the PIII at same clock rate. For floating-point it's supposed to be a little worse. On the *other* hand, the K7 will use PC133 SDRAMs and a 200MHz L3 cache, while the PIII will remain with the 100MHz memory bus for a while yet, and the K7 uses a memory bus twice as wide (128 bits) as the PIII's, *and* the K7 has larger caches and a larger TLB on top of it. So if you're looking at the system as a whole, the K7-based *systems* should outperform PIII-based *systems* hands-down. Even floatingpoint-intensive algorithms bang on the memory bus, and any weakness in the K7's floating point unit will be more than offset by the lower memory latency and higher memory bandwidth.
-
-- Guges --
-
Err...a damned good one.
First off, there's the 128k of L1 (4x that of every Intel P6, including the Xeon).
But, of course, what you're implying here is that the K7 can't run with the Xeon because the Xeon has up to 2MB of L2-cache--very important for, well, server type stuff (big databases, etc.)--and, most ominously, it runs at the same speed as the CPU (for comparison, the PII/PIII has 512k running at half speed; the Celeron has only 128k but it's full speed...and they're about as fast at equal clock speeds, depending on what you're running).
Well, the K7's that shipped today only have 512k of L2, and that only at half clock speed...
However, AMD plans on shipping K7's with both half and full speed L2-caches, ranging in size from the aforementioned 512k all the way up to an absopositively humongous 8MB.
Wow.
Of course, it's gonna cost AMD a chunk of change to fab a chip with 8MB of full-speed L2-cache ( Wow ), but if you think they're gonna be tacking on the same exorbitant markups that Intel sticks on the Xeon (a PIII Xeon w/2MB L2 is $3400 at pricewatch)...well, I sorta doubt it.
But, you may stammer, the Xeon is better for SMP?
Hardly. The Xeon, while technically capable of up to 8-way SMP, only sells in 1, 2, and 4-way configurations so far (I think). 8-way K7's should be up and selling quite soon, and I've heard talk of a 16-way chipset. Oh, and if you're wondering why no one's bothered to make an 8-way Xeon box, it's because the communication between the chips in a multichip Xeon server has to go through the 100Mhz system bus; not only is 100Mhz not so fast, but adding chips means less bus bandwidth for everything else. Long story short, you don't get anywhere near linear performance gains by adding Xeons, and in fact an 8-way box might not be any faster than a 4-way. Luckily, the K7 fixes that little problem by providing a dedicated path for the chips to communicate to each other.
So unless AMD has problems fabbing such massive caches, it looks as though the Xeon's in a bit of trouble...
With all of the pre-release (disappointing) K7 benchmarks floating out there I can hardly wait to get my hands on a K7 to test it thoroughly... AMD, are you listening?
From the information released so far about its internal architecture it should be a scorcher, but only extensive testing will tell the real story.
--------- Webmaster, http://www.cpureview.com and
Well, SMP motherboards tend to be noticably more expensive than UP boards. Many programs don't make good use of the second (third, fourth, etc...) processor. Some people will buy the 600 just to have the latest and greatest. And, of course, most people run Win9x "for the games", and it just doesn't support SMP.
If I can find a reasonably priced dual board, that 500x2 option sounds real nice for a Linux box, though...
And all the reliable speculation (as opposed to the "D00D! INT3L R00lZ AMD'Z A$$" stuff (and the converse)) I've seen indicates that this chip is going to be the "fastest x86 ever" that they bill it as...
And, yeah, the Cyrices had their problems. My 6x86M2-233 will run neck-in-neck with a Celeron 333a doing kernel compiles (though some of that's probably just faster disks and stuff in the Cyrix box), but it gets its butt kicked at Quake and the like. Cyrix's FPU has always sucked big-time. They made a bet that office applications were going to be more important than games in the future, and they lost. They are good machines for a programmer who spends more time compiling than fragging, though.
Another thing to keep in mind when comparing clock speeds is that the Cyrices are not clocked as fast as the number stamped on them seems to indicate... my "233" is a 200MHz processor, and your "150+" was, I believe, a 120MHz chip, so comparing a 150+ to a 120 is a fair comparison... comparing it to a 150MHz Pentium isn't, really. And there's no way to really compare on price... the most expensive Cyrices are so much cheaper than the cheapest Intels...
Ok, I have couple little tidbits to share. First of all, those prices are really closer to what us resellers will be getting them at. Expect $25-50 more from a legit store. You can also exspect that price or probably less from the stores on pricewatch, because they are stolen or not manafactured of they dont deliver. I am not going to tell you where to buy your chips, but just look at the places on pricewatch. A 15 day warrantee? That is obviously not an authorized chip, all AMD chips come with a 3year manafactures warantee, same with Intel. Its no wonder they are cheaper then what us legit resellers can get from our distributors. So when you go out to buy your next chip think about whether you want to save a couple bucks or spend a little more to get a high quality and warenteed product.
Also, comparing K7's to P3's is not realistic. The real comparison is closer to Xeons to K7. And the K7 is faster then any current Intel chip. The pricing is also cheap, considering that it was rumored to cost as much as Xeons do, in the thousands. So be happy.
Anyway, thats all. Good Day.
Jeff Knox
If I ever quit my present job and work on microprocessor circuits I'll do my best to kill names like Pentium or Celeron or Athlon and stuff. They all sound either very similar or just aren't very beefy sounding. They sound like they're picked by focus groups. I hate focus groups.
Instead I'll pick names that are more in your face, much like 60's and 70's sports cars. These chips are powerful and they suck up as much power as old sports cars suck up gas. They deserve names with guts. Names like the Vindicator, maybe with performance specs right in the title, the way sports cars used to proudly display their number of cylinders and engine displacement. The Vindicator 128 1 gig.
Forget dancing femme-boys in gold lame bunny suits, open up a early 70's Popular Mechanics sometime. Thats how advertising should be. Scantily clad girls draped across the vehicle. Chips are small, there won't be any draping across them. Maybe a gorgeous blonde in a string bikini. The camera pans in to her feet and slowly makes its way up her body. Caressing every last inch of her perfect legs and thighs. Just above the bikini line is the chip. Slowly the camera zooms and focuses in on the chip and its name. The Vindicator 128 1 gig. Fade out. No text, no boring announcers. Maybe the wa-wa guitar track from a porno flick. The viewer knows if he buys one of these babies he's getting some!
Ah well, maybe not. The management didn't think adding tesla coils, jacobs ladders and a full height lava lamp to our big servers was a good idea. Even after I explained the concept of retro-computing.
So the K7 is faster then the Xeon at SPECint, and much faster for SPECfp. When multi-processor chipsets come out it will in all likelyhood be far faster then the MP Xeon (thanks to DECs point to point "bus" design). The K7 can eventually beat the Xeon's L2 cache size and match the speed (note that the SPEC numbers are better with a worse L2 cache, I assume due to the far better L1 cache, and a more agressave OOO engine).
So far things are better for AMD then I recall them ever being. However each time AMD had a CPU that could threaten Intel they have had problems making enough of them. Will things really be better this time?
From reading AMDs press release there are a few clues. They will not be selling their CPUs at a fixed discount from the similar Intel ones. They listed a bunch of reasons, but I susspect a big one is they will adjust price to reflect yeild. They also keep talking about the K7 gunning for the high-end market (after all the K6 will not be discontinued for quite some time!). The profit per chip is far higher there, fewer chips will have to be sold to earn the big bucks.
The downside of corse is many K7 systems won't be signifigantly cheeper then the Intel Xenon systems (or so I assume). We may even see the MP K7 systems costing more then the Xenon, after all it will in all likelyhood be a much faster beast.
The upside is Intel may be forced to cut Xenon prices, and in any even when Intel starts beating the K7 there will be more price cuts (and AMD's new fab plant may help them manage volume shipments for once)
Oh yeah? What sort of cache does the K7 have?
(There's a reason that the old Pro-200s can out benchmark a standard Pentium III. Of course, it depends if you are running Quake or databases)
--
Business. Numbers. Money. People. Computer World.
IBM's attempt to take on Microsoft - OS/2
IBM's attempt to take on Intel - PowerPersonal/PReP.
'Nuff Said.
--
Business. Numbers. Money. People. Computer World.
Everyone seems to be speculating that K7 motherboards will accept Alphas without much problem. But there's actually a big problem - the Alpha uses a completely different BIOS/Firmware than a standard x86. (Doesn't the Alpha firmware have an 8086 emulator built-in in order to initialize ISA cards?)
I don't know what the cost or technical difficulties of a duel/swappable BIOS would be, but imagine that there's enough trouble there so that a you-pick Alpha/K7 board is not going to become a commodity item.
During the Phoenix-is-putting-ads-in-the-boot-sequence discussion, a guy from Dell's BIoS division posted, and his point was that creating a BIOS for a specific board is not an easy task. Considering that Alpha is essentially a workstation/server OS, the market could and probably will continue to bear the extra couple hundred bucks.
--
Business. Numbers. Money. People. Computer World.
I guess my point was not that it was difficult/impossible, just unlikely in the low-end motherboard market.
Merged firmware would be great though - I would love an x86 PC with a real boot firmware setup. Once you start adding lots of controllers and drives, the over-extended PC BIOS starts showing it's age.
--
Business. Numbers. Money. People. Computer World.
Check out the prices, they aren't half bad.
600mhz - $699
550mhz - $479
500mhz - $324
Also, from browsing the AMD website, I came across this K7 FAQ. It's actually pretty interesting and gives some new information.
Check out the Athlon logo. There is the number 7 hidden in it. Seems like an AMD artist wanted to send a message.
I am not an idiot. Please use my name to email me.
"That's right, I'm quoting myself."
-Upsilon
If you play games a lot, don't buy K7 now.. Why?
Because there is no a tested motherboard and chipset for it, you are not guaranteed that your prefered 3D video card will work without crashes.
Don't get me wrong. I talk from experience. When the first super 7 (socket) motherboards appeared they were as unstable as hell at doing 3D stuff. They supported AGP but once you pushed them hard the system would just freeze or your game would not run. It took about a year of waiting until decent Super 7 motherboards apperared that would actually work with AGP 3D accellerators without your game/system freezing. Even right now, Super7 motherboards still have issues with my latest and fastest 3D cards. Even at faster CPU clocks speeds
AMD would systems would get lower fps than Celeron systems while both using the same TNT video card.
But if you don't play a lot I guess it is a nice number cruncher for server or compiling kernels...
AMD's FAQ says:
Does this mean that standard cases won't provide enough cooling? Or does "enclosure" mean something else?
YES! YES! YES! YES! YES! YIPEEE!!!
Ok. I'm better now. I've heard that there will be some delay on the dual K7 boards....so, I guess I'll just have to wait to get me a Dual-Athlon...or, should that be a Bi-Athlon?
Can they hook it directly into my brain like Intel did with Homer Simpson? I'd like that!
Definitely an x86 class machine. Because of this I doubt it would work in an alpha motherboard especially since it is using a "Slot A" format for the processor. (Aren't Alpha's still traditional chips?) EV6 is refering to the bus protocol that the processor uses; the rest of the motherboard probably still has to look like an x86 motherboard.
(I'm not entirely sure about this, so someone please correct me if you know for sure that it will run on an alpha motherboard, but looking at the differences between the architectures it would seem unlikely that it would just work.)
-- NetBSD - Better for your uptime than Viagra