Slashdot Mirror
IBM To Produce Copper Alphas For Compaq
LinuxGeek writes: "IBM will be producing copper interconnect Alpha CPUs. The samples are already running 1.2GHz. Hopefully they can make them cheap and plentiful."
Similarly, sokoban
writes:
"News.com is reporting that IBM is planning to manufacture Copper Alpha
CPU's. Now that is a fast piece of equipment. Here is the
link." And nobody ever got fired for buying ... err, Alphas. Soon, 32-bit will seem so quaint, eh?
Not a thing would be my guess. Intel hasn't been anywhere near the top of the technology curve for many years. Nevertheless, they are immensely profitable. Ask yourself: what would you do? I would keep on hype, hype, hyping it up selling to gullible, brain-dead consumers until the consumers either kill themselves off in fits of idiocy or wise up. And I'd laugh all the way to the bank.
Careful with the POWER versus PowerPC, as they're not the same thing. The POWER architecture is really fast, even at low clocks. I remember that SpecFP used to have a 200MHz POWER3 chip in second place after the 500MHz Alpha 21264, ahead of the PA-RISC and everyone else. I think they've at least doubled the clocks by now, and I think they're in the POWER4 generation as well. They had a 64-bit version, the RS-II 64, and I believe they've also moved that ahead to the RS-III 64.
But I've never been able to really keep track of IBM's processors, it just seems tough to get info and keep it straight. Especially with the POWER versus PowerPC mixup.
-Paul Komarek
Otoh, they support linux but not Tru64 (no Tru64 PALcode)...
"My opinions are my own, and I've got *lots* of them!"
The (current) 21264's are aggressively out-of-order speculative-execution CPU's, and on typical "random" codes will execute more than twice as many instructions per clock tick as the '164 does.
"My opinions are my own, and I've got *lots* of them!"
"My opinions are my own, and I've got *lots* of them!"
If you're not part of the solution, you're part of the precipitate.
Consider:
Notably, Netscape is not readily available. (I know that the Digital UNIX version can be hacked into submission, but it is not an easy rpm -i netscape.rpm away...)
Linux may not be the rats-nest of spectacularly nonportable software that Windows NT is, but getting things like KDE, GNOME, GIMP, and such has been a fairly slow process.
I've tried running the Diamond Rio package, rio on my Alpha box; it gets quite confused, probably due to some 32-bit-ism in the "driver."
More critical is the paucity of graphics cards supported by XFree86 on Alpha. It is probably similar for other 64 bit platforms; you're restricted to whatever there are "open" drivers for, and there are some cards (Cirrus comes to mind) that have architectures that are distinctly unfriendly to 64 bit operations.
All of this adds up to Alpha not being a particularly consumer-friendly platform. People that "just want it to work" will find it somewhat challenging in this regard.
This is presently the case for Linux/Alpha; it was the case for Windows NT/Alpha.
The same will likely be true for IA-64, for a goodly while, by the way. The pains that Alpha has had may help IA-64 porting efforts to be less painful, but what "cramps" persist are likely to be irritants to adopters of IA-64.
In short, I don't think this was a likely niche, except amongst "daft hobbyists," which doesn't represent all that "nice" a niche...
If you're not part of the solution, you're part of the precipitate.
Heh, consider this:
:-(
[marcus@earth marcus]$ uptime
10:44pm up 331 days, 9:55, 6 users, load average: 0.14, 0.05, 0.01
[marcus@earth marcus]$ uname -a
Linux earth 2.2.5 #3 Sun May 30 13:10:30 CDT 1999 alpha unknown
That's an OLD alpha axppci33(266MHz, 256K cache, 64MB RAM, 6GB of disks) that does qmail, 2 apache servers(one being a devel/demo of java servlets), leafnode, ftpd, local DNS, PPP on three dial-in lines plus firewalling and masquerade for 6 boxes on the local LAN and the dial-ins 24/7 and obviously working on that 365 part. Its plenty fast as there is no problem saturating the 10Mb ethernet with apache(raw files) or ftpd.
Total cost of about $550.00.
Since the ADSL line is about to arrive, I'm going to have to take it down to add another NIC.
No, it does not run X, Netscape, nor any "office" software. Consumer platform? No, it's flawless just as it is.
Good judgement comes from experience, and experience comes from bad judgement.
- W. Wriston, former Citibank CEO
AMD has some SPECfp-95 numbers here. They show a 800Mhz athlon at 26.6 (P-III shown at 19.8), 900Mhz at 27.5, and 1000Mhz 29.4 They use to have some SPECint numbers as well, showing about a 15% lead over the P-III, but that was at a clock speed where the Athlon had the half speed L2 cache, and it was when the P-III also had a half speed cache (P-IIIs with the smaller, faster, more associtave cache L2 are a bit faster on many benchmarks then the older P-IIIs). So I don't know how the current P-III stands in int to the current K7. Probbaly not as good or AMD would publish those numbers.
The orignal poster wanted to know if the Alpha was competitave with lower clock speeds. The answers shown here say the 21264 is. I want to point out that the picture was somewhat diffrent with the 21164 and 21064. They are strictly in-order machines, and the 21064 has very strict instruction grouping rules.
The 21064 (frequently) does less work per clock then a P-III, P-II, and even frequently a PPro. Of corse the 21064 was new a very long time ago (8 years?). The 21064 did somewhat less work per clock then most of it's contempary RISCs. But it was clocked much much faster. (i.e. 133Mhz 21064 vs the 50Mhz TI SuperSPARC, definitly much faster then the 386s and 486s of it's day as well)
The 21164 did more work per clock then the 21064. It's contemparary RISCs were doing much more work per clock. Far more out of order dispatch and specultave execution in the other RISCs, and in the PPro which Intel brought out during the 21164's lifetime (I think). However the 21264 once again had a dramitaic clock speed advantage (2x or more in some cases - 3x to 4x over the Intels for a while).
The 21264 is a diffrent kind of Alpha. The first one to do a lot of out of order and specultave execution. Clock for clock it gets far more work done then the 21164. Which is good because it no longer outclocks many of it's foes. It has even lost the top slot in SPECint which the Alpha has held more or less since it's introduction (not at all times, but never lost it for more then a few months that I recall -- then again I didn't check every month!). Even the SPECfp it shows some weeknesses it hasn't in a long time (I recall it having a firmer grip on #1 here, losing when HP announced new HPPA machines, and to some of the RS/6000s, but allways regaining it fairly soon after).
Seeing a 21264 with a really fast clock will definitly be an impressave thing. It will strain it's 8M L2 cache at 1.2Mhz for sure. Probbably blow right past it. I wonder if they will have a much larget L2, or if they will use a big L3?
I also look forward to the 21364. I don't think Compaq has said a lot about it, other then SMT. Simultanius Multi-Threading, which looks similar to SMP on a chip, but it lets the CPU assign functional units on a per-cycle basis to acheve maximum throughput. I wonder what else the 21364 will do. I wonder if it will go back to a simpler in-order less-specultave execution model (using SMT to tolarate the memory latency).
I heard the 21364 taped out in December as planned. I don't know how the test samples were, or what the rest of the schedule is.
P.S. dispite never having owned an Alpha, I'm definitly a member of the "Alpha is fast cult". A pity it has allways been cheeper to ray trace using more CPUs of something else (the K7 last I did my estimates).
I'm a musician, and I would dearly love to have one of these things to set up a realtime synthesis/audio processing engine for jamming through MIDI. I consider that to be a "home use."
I used to do quite a bit with POVRAY and 3D modelling. It would be neat to have a machine that could render a frame in a few seconds (instead of 20-30 minutes). That would really cut down on the time required to develop a scene.
You could compile the kernel really fast! ;-)
It turns out that arithmetic operators cost half as many transistors/operator if the bit order goes from lowest-order bit to highest-order bit, as in an Intel chip, than they do if the bit order is highest-order bit first.
Having just finished my final year in Computer Engineering learning chip design, I can say with confidence that this isn't true. The implementation is the same no matter which order the bits are in (just use a mirror image of the layout).
You might also want to visit the EGCS site, which mentions Alpha updates to GCC as recently as March. GCC does support the newer EV6 architecture. The problem with Linux is more that large portions of it are hand-coded for the GCC assembler -- and optimizing kernel code doesn't do much since so little time is spent in the kernel anyway. ;-)
So IBM is fabbing alphas at 1.2 GHz, meanwhile the POWER line, which is their baby, seems to be stuck at 500 MHz or so (last time I checked). What gives?
BTW, an Alpha I purchased a few years back is STILL faster at floating point than the latest from Intel or AMD (OK, the Athlon is close).
Octane rating is how fuel performs on a standard test engine. If the sample fuel performs as good as a mixture of 95% Octane and 5% Septane then it is given an Octane rating of 95. If it performs better than 100% pure Octane a math formula is used to assign an Octane rating of 100 plus (Mainly used for racing and aviation fuel) Octane is used as the base because it has the best anti-knock properties of any of the hydrocarbons in gasoline. Septane is used as the other part of the mix because it has the worst anti-knock properties.
Quemadmodum gladius neminem occidit, occidentis telum est
And since FX32! (Intel x86 emulator) was outlawed (because dynamic recompilation violated software EULA's becuase it automatically reversed engineered the object code), I don't expect the Alpha to ever the viable in the end-consumer marketplace.
Uhhh...I don't know what you're talking about. Two things:
AFAIK (I worked on the VC compiler for Alpha) FX32 was never "outlawed". In fact, it was integrated into the Alpha version of Win2000.
Second, FX32 is a moot point since Compaq dumped NT as an Alpha OS. Even if FX32 was "outlawed", it wouldn't be why Alpha will never be a consumer platform. It would be because Windows is a dead OS, as far as the Alpha is concerned (boy that's fun to say).
Having said that, I don't think Compaq has any inclination or incentive to price/market the Alpha toward end users.
--GnrcMan--
Well, regardless of that, I worked with the team that makes the proprietary Compaq compiler and it truely is amazing.
But you are right, EGCS is making inroads as far as Alpha optimization goes(working on the Alpha compiler is my pet project(haven't submitted anything yet) now that I'm no longer working on VC for Alpha, which used the same optimizer, "GEM"), but GEM is the only way to go right now.
--GnrcMan--
You mean iso -octane (octane number 100) and n-heptane (octane number 0). There is no compound called septane. And for diesel engines, the corresponding numbers are called cetane numbers.
I can throw myself at the ground, and miss.
www.alphalinux.org
The link is a vendor list, alphalinux.org is a nice place to look for info. Also check comp.os.linux.alpha .
You won't be able to get a new alpha for less than $3500US, but used ones are available for cheap, sometimes free.
Ace's has some benchmarks (under linux, no less) of the different compilers. They prove your point quite well.
-
__
Comment submitted. There will be a delay before you understand what you posted.
Pfft, you're not even trying. Ask someone who's compiled the 2.0 series kernel on an the minimum spec Amiga (That Linux will run on) Mmmm, Motorola 68030, at 25Mhz. Now thats slow.
Syllable : It's an Operating System
Are you using your Alpha box to make money? (Click here then scroll down to the tree.) Or are you just running the distributed.net client?
Will I retire or break 10K?
Man.. this is great! First off, hat's off to the Alpha. I've used several of these systems, and all kidding aside, a single processor 533 with 512Meg will happily run 6 quake servers! That's a serious beating.
I'm glad to see some of Intel's technology rolling over into the other processors. Hardware companies should take cues from the open source movement. If you see something you like, use it in your product and come out with something even better. I've gotten worn out from the Gigahertz wars, but some seriously beefy Alphas are always welcome in my place.
I do hope that some companies will take notice and start porting their games (like HalfLife) to the Alpha platform. They'd make totally superior gaming machines since the FPU is actually worth something as compared to Intel.
1.2GHz Alpha... 4MB cache... Drool!!! When can I get one?
-What have you contributed lately?
Yeah, a fast alpha is not what every home user needs. However, there are a few justifiable reasons for such huge amounts of power. First of all, speech recognition is really moving forward much faster than the processing power is. IBM released their ViaVoice recently, and it runs well on a G4 450, but by giving it a consideralby faster system, accuracy and response could be vastly improved. Also, Desktop video production is becoming a reality with IEEE 1394 camcorders coming down in price. Any DV program would be much nicer on a fast system with lots of ram. Also, there is the trickle-down, racing improves the breed mentality behind faster processors. If the high-end systems move up to these ultra-fast Alphas, then theoreically, the previous high-end will become the mid-level consumer systems and the prices will drop across the board. The hope is not that everyone has a 1.2 Ghz alpha by the end of the year, but that Copper chips and such trickle down to the masses.
09 F9 11 02 9D 74 E3 5B D8 41 56 C5 63 56 88 C0 is the magic number.
Oh come on, the answer is obvious!
... but that's getting OT) halves in speed every 18 months. If it weren't for faster chips, we'd all be going comatose waiting for our goddamn spreadsheets to recalculate! ;p
We have Moore's law, which tells us that chips double in speed every 18 months. We also have Gate's law, which tells us that software from most proprietary companies (especially one who's name is an accurate description of Mr. Gates' reproductive organ
You'll probably be able to find a EV56 machine for around $2000.
I recently purchased a DEC Alpha DPW 433au (EV56 21164a/Miata mobo) from egghead.com for US$1,199.   They recently had them avail. for US$1,049 (I just checked and I don't see it anymore... they only had 4 left and they may have been snapped up).
What is impossible to find is a place that will sell you just a motherboard or just an Alpha.
Amen!   Although I've seen more places that will sell the mobo and parts like E.L.I or other places that will sell you "inexpensive" (in quotes - although considering, the prices aren't too bad) systems, like DCG, but other than those, it has been a pain trying to find an upgrade processor for less than I paid for the whole system!
I still like my little alpha though...   heh heh.
-- Win2k: "It's not so much that it's only 65,000 bugs, it's just that they stopped at 65,535 to prevent an overflow."
Did you check Compaq and Samsung's marketing company's website? Alpha Processor Inc. They have a list of system vendor's. You'll probably be able to find a EV56 machine for around $2000.
What is impossible to find is a place that will sell you just a motherboard or just an Alpha. If I could go out and buy my own EV6 and mb and build the rest of the machine around it, I would have done so a long time ago.
Bill
Does anyone have any comparisons of Alpha vs other processors at equivalent clocks? I remember comparing processors 3 years ago, when the Alpha was getting comparatively high clock rates. When you really looked at a ratio of Work/MHz (of course, the definition of "work" can vary tremendously), the Alpha wasn't very "efficient" in terms of how much it did in a clock. Of course, the was the entire intention behind RISC. Lots of clocks by reducing how much was done each clock.
But now with the Alpha unable to sustain a big clock rate advantage, does the Alpha get any sort of real performance advantage when compared to x86, or particularly other less "reduced" architectures like HP's, Sun's or IBM's?
--
Sometimes it's best to just let stupid people be stupid.
Right. Just the other day I was mulling over perhaps getting a medium to a low-range Alpha. And I reached a conclusion that the high price is the real problem with Alphas.
I think it's mostly has to do with the fact that Compaq, it seems, doesn't really want to market this hardware to consumers. Here I was, seriously considering one, but...
For starters, there doesn't appear to be a lot of places selling those things out there. I've tried searching the 'Net, I didn't find anything. I couldn't believe it when I searched shop.yahoo.com, which basically aggregates hundreds (if not thousands) of merchants' e-commerce shops under one roof, and came up dry.
After I went to Ground Zero -- Compaq -- I came away with a discreet impression that they really don't want to talk to a mere peon like me. Their web site, apparently, is really put together more like a business-to-business (or government) type of a deal.
That's really a shame. I think that I'm not the only one who might be interested in looking into alternatives to the x86 platforms. If Compaq only invested a small portion of that R&D that went into this newfangled chip into, instead, investing in marketing the Alpha to consumers, they might find themselves a nice niche amongst consumers too.
According to this EETimes article, high-frequency induction is a growing concern. It cites an academic paper stating "copper is much more problematic than aluminum when it comes to inductance...."
SAN DIEGO, Calif. -- The perils of chip design at 0.18 micron and below mandate new research in a number of areas, according to presenters at the recent International Symposium on Physical Design (ISPD) 2000 here. Calls were made for breakthroughs in areas such as power analysis and estimation, inductance modeling, analog layout, incremental CAD and process variability.
Beyond academic papers, this year's ISPD included a keynote speech by Aart de Geus, chairman and chief executive officer of Synopsys Inc. (Mountain View, Calif.) In his keynote speech, de Geus called for better estimation and analysis tools, and for attention to new problems such as inductance.
"Finally, after 25 years, inductance is coming back," de Geus said. "With high current, inductance does matter." But it won't be a big issue for most designers, he added, until feature sizes drop below 0.18 micron.
A paper given by Li-Fu Chang of Frequency Technology Inc. (Santa Clara, Calif.) outlined that company's research in inductance modeling of on-chip copper interconnects. Copper is much more problematic than aluminum when it comes to inductance, the paper notes.
Chang said the paper presents a "full chip" inductance modeling architecture. "Inductance modeling is much harder work than Rs and Cs [resistance and capacitance]," he noted.
How much do you want to bet that when I try to specify copper Alphas, that dimwit teenager in Academic Purchasing (who thinks he's 3l337 because he can remember the HardOCP URL 2 times out of three) will get me a stack'o'coppermines instead?
Oh, and think of all the lovely conversations you'll overhear waiting in line with that last Y power connector for your Beowulf as the wannabe Mr. Know-it-alls ("I order all my systems on-line!") tie up the Saturday salesclerks with impossibly self-contradictory orders.
_____________
If you can go to bed, knowing you did a valuable thing today, you're very lucky. If you can't... it's not bedtime
Well, it would seem that the future is in Copper interconects. I'm just wondering what is Intel going to do about it. Last time I read up on it they refuzed to license the technology from IBM.
Anything I missed? Is Intel going to go Copper, or are they just going to keep on naming their processors so that they sound like it?
2 letters : vi.
Treatment, not tyranny. End the drug war and free our American POWs.
See my user info for links.
Yes, these things will totally kick ass, but if you use gcc and glibm it would be like running 80 octane gasoline in a Corvette. This is a brand new architecture that the gnu development is years behind on. Compaq has compilers, libraries and prebuilt apps that run _much_ faster. I don't think anything has been done to gcc since the EV5 architecture.
The Compaq tools are mostly available for free, but you can't compile kernels with them. We're trying to catalog what can actually be compiled with them but most Linux system source assumes that you are using gcc. Sad, really. I'm sure that things would run much better when using a compiler that actually knows how many execution units the processor has...
So, on another note, anyone know of anyplace (on the web or otherwise) that I can find out exactly what IS up with BIG_ENDIAN and LITTLE_ENDIAN? I'm not going to be using a BIG_ENDIAN machine anytime soon, but I've always been interested.
Do you want the literary reference from which the names are taken? (Swift, _Gulliver's Travels_, Book 1, "The Voyage to Lilliput". There were two tribes of Lilliputians who were fighting a war over whether one should eat one's breakfast egg from the big end (the Big-Endians) or the little end (the Little-Endians). I never can remember which tribe Gulliver was allied with, so don't ask me.) Or do you want the technical explanation of why Intel puts the bytes in a word "backwards", so that you can't read the characters in a UCS-2 string from the bytes? (It turns out that arithmetic operators cost half as many transistors/operator if the bit order goes from lowest-order bit to highest-order bit, as in an Intel chip, than they do if the bit order is highest-order bit first. It doesn't make all that much of a difference any more, since the transistor count on the ALU is only a tiny fraction of the cost of the chip, but in the old days, it mattered very much indeed. I can't ever remember which tribe I'm allied with, so don't ask me.)