Sun's Zippy New Chips 246
Mark the Revelator writes: "Reuters has a story about Sun unveiling it's latest and greatest UltraSparcIII chips. The new chips are being made by TI and are the first UltraSparcs to use copper instead of aluminum for transistor connections. Although they're supposed to compete with Intel's Itanium chips, they only run at 900MHz ... for now."
Sun's MAJC (Score:3, Interesting)
Benchmarks anyone ? (Score:1)
Intel D850GB motherboard Pentium 4 1.7 GHz: SPECint2000 = 586
All from http://www.specbench.org.
My old manager had a great reason for buying Sun; no one will ever put M$ windows on it!
Re:Benchmarks anyone ? (Score:2)
Links [cnet.com]
(Preview looked good... what's going on there?)
Zippy Processors? (Score:1)
Here's [zippythepinhead.com] a real Zippy processor.
TI? (Score:1)
Does this mean that I'll be able to run the "Space Rocks" game I programmed on my TI-85 graphing calculator?
Ha ha, very funny (Score:4, Insightful)
They make all of Sun's UltraSparc chips, and also manufacture other, more esoteric things - like dual core chips (DSP and ARM, known as OMAP).
All in all, TI is much, much more than calculators.
a completely arbitrary comment (Score:1)
I know its just anectdotal, but hey, some of the best stuff is!
Nice to see they aren't caving (Score:2)
Don't get me wrong, I love my Athlons, but I used to work in an HP based shop with PA-RISC all around. I'll never forget when the K-Class and N Class servers first came into our data center with the latest PA-RISC beasts - they were so fast it was scary (this was like 3 years ago)
Re:Nice to see they aren't caving (Score:2)
Re:Nice to see they aren't caving (Score:1)
However, the HP visualize workstations tend to be a balanced solution - better than SGI and Sun, at least in the recent past.
this could have been cool.... (Score:1)
missing the point as usual (Score:2, Insightful)
You get (IMHO) the best OS you can run on a server (*incomparably* more reliable than Linux in my experience).
You get a better build quality than with PC class gear. (not so with the low-end Ultras I know, but have you tried carrying an E450 around lately?) I've worked with Sun boxes (mail hubs, NIS servers etc.) that haven't been swtiched off in 8 or 10 years.
You get excellent support - hardware or software. It costs, but it's worth it.
You get as much SMP as you could want.
You get insane amounts of addressable RAM and faster bus speeds.
In short kids, you get *proper computers* running a *proper OS*.
MHz to MHZ (Score:5, Informative)
Just because the MHz on the Sun equipment (900MHz) is lower than the current Pentium (1.5MHz), don't be fooled into thinking the Intel hardware is better. What matters after all, is throughput and pumping that data. Check your specs [spec.org]!
Check this 4 CPU Intel vs the 1 CPU Sun considering plain speed...
CINT2000: Intel Corporation Intel D850GB motherboard(1.5 GHz, Pentium 4 processor) - 536 524
CFP2000: Intel Corporation Intel D850GB motherboard(1.5 GHz, Pentium 4 processor) - 558 549
CINT2000: Sun Microsystems Sun Blade 1000 Model 1900 - 467 438
CFP2000: Sun Microsystems Sun Blade 1000 Model 1900 - 482 427
CINT2000: Advanced Micro Devices Tyan Thunder K7 Motherboard, 1.2GHz Athlon MP Processor - 522 495
CFP2000: Advanced Micro Devices Tyan Thunder K7 Motherboard, 1.2GHz Athlon MP Processor - 481 433
Throughput on the Sun with 2 CPU, but strangely enough, none for any Intel hardware. Throw a 2 CPU AMD in there, though...
CINT2000 rate: Sun Microsystems Sun Blade 1000 Model 2900 - 10.7 9.97
CFP2000 rate: Sun Microsystems Sun Blade 1000 Model 2900 - 10.2 9.09
CINT2000 rate: Advanced Micro Devic Tyan Thunder K7 Motherboard, 1.2GHz 2CPU - 10.8 11.1
CFP2000 rate: Advanced Micro Devic Tyan Thunder K7 Motherboard, 1.2GHz 2CPU - 8.30 9.14
Re:MHz to MHZ (Score:1)
Re:MHz to MHZ (Score:2)
However, you often need to consider the need for scalability, reallyreally good uptime, support for old machines a few years down the track (and parts availability), and so on. Even the way that performance degrades as load increases can be a differentiating factor. The extra $$$ you spend on Sun gear certainly does get you more, but the question is whether that is worth paying for in your situation...
rr
Re:MHz to MHZ (Score:2)
Re:MHz to MHZ (Score:2)
Suppose that a CPU fails in your precious server: on Intel hardware you have to take the server down, replace the CPU and reboot.
Certain models of Sun hardware have CPU hot-plugging: a faulty CPU gets detected by the OS and is taken offline. A techie comes, replaces the CPU without turning the computer off, and then the new CPU is taken online without breaking a sweat.
Some goes for RAM.
Ever tried changing a broken RAM stick on Intel wihtout powering off? Don't do that, you can't.
The PC platform only has one strong point: it (used to be) is simple and widespread, and those factors mean cheap. Other architectures are not as simple or not as widespread, thus are more expensive. But they are consistent, they are scalable, and they are reliable. All things that the PC architecture is not.
Re:MHz to MHZ (Score:4, Informative)
>>>>>>>>>
I believe it says Pentium 4 as in the "Pentium 4," not 4 Pentium CPUs
CINT2000: Intel Corporation Intel D850GB motherboard(1.5 GHz, Pentium 4 processor) - 536 524
CFP2000: Intel Corporation Intel D850GB motherboard(1.5 GHz, Pentium 4 processor) - 558 549
CINT2000: Sun Microsystems Sun Blade 1000 Model 1900 - 467 438
CFP2000: Sun Microsystems Sun Blade 1000 Model 1900 - 482 427
CINT2000: Advanced Micro Devices Tyan Thunder K7 Motherboard, 1.2GHz Athlon MP Processor - 522 495
CFP2000: Advanced Micro Devices Tyan Thunder K7 Motherboard, 1.2GHz Athlon MP Processor - 481 433
>>>>>>>>>>
So you just proved that the P4 chop-shops the UltraSparc in SPEC...
Throughput on the Sun with 2 CPU, but strangely enough, none for any Intel hardware. Throw a 2 CPU AMD in there, though...
>>>>>>>>>
Again, P4 doesn't do SMP, but Athlon does.
CINT2000 rate: Sun Microsystems Sun Blade 1000 Model 2900 - 10.7 9.97
CFP2000 rate: Sun Microsystems Sun Blade 1000 Model 2900 - 10.2 9.09
CINT2000 rate: Advanced Micro Devic Tyan Thunder K7 Motherboard, 1.2GHz 2CPU - 10.8 11.1
CFP2000 rate: Advanced Micro Devic Tyan Thunder K7 Motherboard, 1.2GHz 2CPU - 8.30 9.14
>>>>>>>>
So, the dual CPU athlon beats the UltraSparc in SPEC as well.
Avoid showing data that refutes you claims...
Only 900 mhz, but... (Score:1, Informative)
RISC processor performance fallacy (Score:2, Interesting)
> and have always toasted any CISC or CISC-to-RISC
> processor of a much higher processor rating.
That is misleading and, in fact, bordering on the level of a total lie. The benefits of RISC architectures are not performance. They're simplicity. This simplicity, in the past, sometimes had the benefit of increasing performance, but higher performance is not a rule in and of itself.
Saying that "pure RISC processors
In specfp_2000, the lowest frequency Pentium 4 scores a 516, while the highest frequency UltraSPARC III scores a 482. The slowest Pentium 4 is 7.1% faster than the fastest USPARCIII.
In specint_2000, the slowest P4 gets a 490, while the fastest Sun processor gets a 467. Here, the wimpiest current generation Intel processor is 4.9% faster than the best thing Sun can offer.
These above factors keep in mind that the Sun chips are *specifically* architected to achieve the highest performance possible, pretty much regardless of cost. They are full-on server chips. The Intel Pentium 4 series are designed with cost factors in mind. The Pentium 4 cannot be a three thousand dollar behemoth due to its target market (actually, the 750MHz USPARCIII processor module costs about $7k on Sun's website). So the USPARCIII can have the benefit of loads of added performance enhancing features while the Netburst (P4) architecture has to cut corners at every step.
the UltraSPARC III outperforms the Pentium 4 on a clock for clock basis. Of course, the original Pentium outperforms the Pentium 4 on a clock for clock basis on many benchmarks, too. This means nothing. It is merely reflective of different design strategies. I can easily point out the fact that the Pentium 4 offers higher performance per watt or higher performance per number of integer ALUs. But, like "performance per megahertz", those are also stupid measurements.
There is nothing out there which would cause me to believe that an x86 processor made with the design strategy of the UltraSPARC III ("we're gonna sell this for thousands of dollars, so throw in the kitchen sink, too!") would not outperform the UltraSPARC III at like frequency. Well, except if the the fp instructions on the USIII are three operand, but that's a special case. ^_^
-JC
http://www.jc-news.com/
I can remember thinking... (Score:2)
Re:I can remember thinking... (Score:2)
And later I got a 300 baud modem. Double RaH.
Later,
daWiz
The Wizard's Realm - Telegard 2.0 - 686-0235
Clock Speed is a marketing tool (Score:1)
Here's the formula for computing a metric that allows you to compare two different chips regardless of architecture when running a specific program:
Effective Speed( Program p ) = Instructions in p * Avg # of cycles to complete 1 instruction * seconds per cycleIn other words seconds per cycle is the inverse of clock speed.
What can be gained from this equation:
Stop drooling about clock speeds, it is nearly meaningless, and is only a marketing tool. If users thought feature size was cool, we'd be having this argument about how "0.18 micron feature size" is meaningless when trumpeted out of context.
Start getting interested in SPECint and SPECfp metrics. Why don't chip makers start focusing on those metrics?
When Moore's law starts failing, someday, we'll see far more innovation in chip designs that don't relate to feature size or clock speed. There's a whole unrealized sea of optimization that could happen to speed up current designs right now. We all beat up Microsoft for being a monopoly, how about Intel :-).
Cool. (Score:2, Funny)
With the same switch, world peace will come, britney spears will have some talent and just maybe, code redworm will stop being a "threat". Ok ok.. I went overboard witht he britney spears thing, sorry ;)
Britney (Score:1)
It's just that you usually the only place you see that kind of talent is in porno movies...
Megahurtz! (Score:2, Interesting)
Intel machines in the EDA market (Re:Megahurtz!) (Score:1)
It's just sad that many EDA companies tried offering their products on Linux; and had to pull them back because of low demand.
Re:Megahurtz! (Score:1)
What? (Score:3, Insightful)
Just becaue it runs "only" at 900mhz doesn't mean anything compared to an Itanium running at a higher clock speed. There are many more factures like pipelines, cache, and over all archetecture. A 900mhz sparc could beat an Itanium at a higher clock speed just like Athlons and PIIIs can beat P4s in certain benchmarks while running at lower clockspeeds. (not saying it will or will not, but you can't discount one processor based only on clock.)
Does anybody understand anything about processors? (Score:3, Insightful)
The Itanium achieved some truely awesome SPEC-FP scores that made Sun look pretty bad. At FP, Itanium whales.
Itanium suffers from the same problems as the Pentium 4, in some ways, in that you can't ever branch. If you can find code that does this, and doesn't have many NOPs, the Itanium will perform very well. That doesn't describe much general-purpose code in the real world.
So, the crux of this is that Itaniums are faster at some things, just like the Pentium 4 is faster at some things. The risk is that these Intel processor applications are becomming highly specialized, and better general-purpose processors are available.
RISC/CISC (Score:2, Offtopic)
Re:RISC/CISC (Score:1)
No, Itanium is VLIW. It has some benefits of RISC. (Score:1)
While it may be up for some debate, CISC processors are known primarily for opcodes of variable length (some instructions are one word, some two, some three, etc.). They are also known for an overly-rich instruction set and a smaller number of registers.
RISC processors are known for uniform length of opcodes, a library of instructions that has been tuned to optimize compiler output, no direct operations on memory other than load/store, and a larger number of registers (sometimes allocated in clumps like the Sparc register windows).
VLIW (which Intel calls Epic) is also known for uniform length of opcodes, but gathers several opcodes into a bundle and executes the bundle all at once. The Itanium executes 3 instructions at a time in this way. The Itanium's main weakness is that it cannot execute these "bundles" out-of-order, and it relies upon compile-time analysis for most of its optimization when the least is known about the executible code.
AFAIK, both the Pentium 3/4 and the Athlon still have out-of-order RISC processors at the core, and translation units that move x86 code into the native code. Cyrix was the last vendor to make a real x86-CISC processor, but it couldn't scale much beyond 400MHz, so VIA killed it.
The original Pentium had something similar to the VLIW idea in that it had two parallel execution units for executing two instructions at once (super-scalar), but the second execution unit got switched off if there was a dependency between the group of two instructions.
Re:RISC/CISC (Score:5, Insightful)
The Itanium is a branch off of a different tree, Very Long Instruction Word, which is a branch off of RISC. VLIW let's a compiler pack multiple commands to multiple execution units into a single long word. The idea is to use very RISCy commands to keep a superscalar set of execution units more fully utilized. Great idea, if your compiler can do it.
Re:RISC/CISC (Score:1, Interesting)
Re:RISC/CISC (Score:1)
Re:RISC/CISC (Score:3, Interesting)
Is this a good idea though? I mean, using one of today's compilers, ported to a IA64/Itanium architecture, a compiled program might run very slowly, since today's compilers probably let a bit of the optimization (within reason) up to the CPU. This would also mean that it may be a little while until some quality IA64 compilers are released. Or am I misinformed?
Re:RISC/CISC (Score:1)
Already done. [intel.com]
They even have a compiler plugin for VC++ that generates fast binaries. Check this [aceshardware.com] out.
Re:RISC/CISC (Score:2)
Re:RISC/CISC (Score:2, Interesting)
From my point of view, the Iantium shares a lot of similarities with the iAPX432 of 20 years ago. Both are new architectures that purport to emulate the previous technology, but from all reports at least, don't do that emulation very well. Both rely on software technology that exists only in the laboratory, if there.
Only time will tell if both share the same marketing fate.
Re:RISC/CISC (Score:1)
The Intel is more of a mixed design---some instructions are complex, but the pipeline on Intel chips is long. As with the Alphas, very little is done per clock cycle, so they've been able to push up the clock speed. Intel has lots of $$$, so it hasn't been difficult for them to achieve very high clock speed levels.
Btw, does anyone know what the current Alpha clock speed is?
The CISC philosophy, which Sun adopts, is to have fewer clocks, but to get more done each clock. CISC designs don't need such a long pipeline and hence aren't as susceptible to branch mis-predictions. However, the higher Mhz processoros usually still win since they have long pipelines which allow many smaller operations to be done simultaneously. It might take 20 clocks to finish an instruction, but an instruction completes almost every clock. (Comp. Sci. professors like to use a "laundry" analogy.) Branches ("if," "while," "goto" in code) can be deadly, but only if the processor guesses the wrong branch to take. Fortunately for Intel and others that adopt the long-pipeline philosophy, simple algorithms can predict the correct branch with very high accuracy (99.9%+). Hence, long-pipeline chips tend to get stuff done faster than CISC-minded models.
Jason
Re:The Ultimate IBM POWER2 (Score:1)
My webserver, in fact. (though it's done other things as well)
Traffic Jam (Score:5, Insightful)
Part of Sun's success is how well they address the bus/throughput issue, as opposed to 'other' computer architectures. And that's why JUST comparing MHz is like comparing apples and oranges.
Or perhaps a better anology is comparing a Formula 1 Racing car stuck in down-town NYC Traffic, versus a 6 cylinder Honda Accord on flat, wide-open highway in Montanta, during the daytime when the weather is perfect.
Re:Traffic Jam (Score:1)
Give the racer and the pickup each a load to haul. The racing car will pass the pickup several times hauling the load, but the truck driver will be twiddling his thumbs waiting for the other guy to finish.
While Sun boxen are only decent at CPU power, that was not the central design goal. Pushing data is.
Re:F1 vs. Honda in chick appeal (Score:3, Funny)
On the other hand, a sparc runs the software I want to run, and the software I earn tons of money from. So of course, having tons of money gets higher quality chicks better than any car
the AC
[not a politically correct post since I'm in a country which has outlawed 'Merkin correctness]
Re:F1 vs. Honda in chick appeal (Score:1)
Re:F1 vs. Honda in chick appeal (Score:1)
Re:Traffic Jam (Score:3, Insightful)
The Intel chip is like a sports car.
The UltraSparc is a Tractor Trailer
They both have the same horse power (in theory). The intel's nice and zippy, assuming that you don't have alot of weight. The UltraSparc may not be as zippy, but it can handle a heck-of-a-lotta more load. Its sorta like the difference in gear ratios. Now, would you like to pull a tractor trailer with a Lamborghini? No. Would you like to drag race the Tractor (stock of course - I've seen the races on speedvision (GRIN))? No. It depends on your application. DB's are a heavy load. Just think about it.
Sun Blade 1000's run them. (Score:2, Interesting)
It blows any of the pentium-based machines we have here out of the water.
-fialar
Re:Sun Blade 1000's run them. (Score:1)
Re:Sun Blade 1000's run them. (Score:2, Informative)
uh, more bits doesn't mean a faster cpu. in fact, it means pointers are twice as long, which means they take up twice the cache and twice the memory bandwidth. the fact that most 64bit cpus are faster than more 32bit cpus has absolutely nothing to do with them being 64bit.
megahertz schmegahertz (Score:2)
i'll have to remember this when the next G4 vs Pentium4 debate comes up...
UltraSPARC chips Vs Intel (Score:3, Informative)
UltraSPARC 450s do things about the same time as Pentium 900s, etc.
These should be screamers. Don't be fooled by the number attached to the chip.
DanH
Re:UltraSPARC chips Vs Intel (Score:1)
Re:UltraSPARC chips Vs Intel (Score:5, Informative)
Where the real advantages come in is with things like memory architectures (eg, memory interleaving) and bus speeds (where the system bandwidth is more than an x86 solution) which is relevant in databases. Added to that, you can scale these up much more (the E6800 can have 24 900MHz CPU's, for instance; Fujitsu have recently released a 128 CPU system based on their USII clone at 500+MHz).
If you want a measure of raw CPU performance, check www.spec.org; currently, the fastest single CPU systems are Intel P4's (although some alphas come damn close). The Sun 280R doesn't come close to that, although it is faster than its clock speed would suggest...
Re:UltraSPARC chips Vs Intel (Score:2)
My comment was on clock speed and how most people only look at the CPU speed and decide that is how fast the machine works without taking into account the different architectures as a whole into account, as it seems the writer ofthe article saw '900 Mhz' and used the word 'only' in the write up.
DanH
Stop the slaughter (Score:4, Funny)
How in a civilised society can we sit back and let this apocolypse happen? I say its time to end this now. Boycott processors. Save the instructions
Re:Stop the slaughter (Score:5, Funny)
Re:Stop the slaughter (Score:5, Funny)
Re:Stop the slaughter (Score:1)
Re:Stop the slaughter (Score:1)
Bingo (Score:1)
Re:Stop the slaughter (Score:3, Interesting)
Actually, this will happen. Current processors are designed so that with each cycle they load and decode the instructions they're going to execute (and, of course, the data the instruction is going to work on). When the instruction is completed, it's thrown away. This is highly inefficient for loops, because the same code is loaded again and again. Think about audio or video decoders -- the same decoding instrcutions are reloaded all the time.
Future processors will, at least partly, be reconfigurable, that is they will load a set of instructions and save it, and then have to load the data only. This is supposed to be the optimum between a hardware-only implementation (fastest, but can't change when, for example, encodings change) and current "software-only" implementations (most flexible, but processors must (re)load instructions in each cycle).
Take a look at, for example, PACT [pactcorp.com] if you're interested in this technology, they're one of the companies developing such processors.
Re:Stop the slaughter (Score:1)
You just described a cache, and they have been used for years.
MHz not only measure of speed (Score:3, Interesting)
Let's start with the whole RISC vs. CISC thing. Everyone knows that RISC is more efficient; the only thing that has kept CISC alive this long is backwards compatibility with the Wintel juggernaut. You develop a lean, efficient instruction set, then you write compiler back ends that take advantage of it.
Also keep in mind that Sun's motherboard designs are true performers. The path between the CPU, memory, and bus are designed to move data around in ways that just aren't possible with Intel.
Did you know that SPARC is more or less an "open" CPU design? It was designed to be a multi-vendor instruction set, one that would be 'common' without having one vendor calling all the shots. Read www.sparc.org [sparc.org] for more details.
Re:MHz not only measure of speed (Score:4, Insightful)
Really? By what measure? CISC is generally much more efficient with respect to code size, an important consideration in embedded systems.
I'll assume you were talking about the performance domain. Be careful with your categorizations. There are no "pure" RISC or CISC designs anymore. O-O-O superscalar architectures have pretty much killed any simplicity in so-called RISC designs. Now it's true that uniform instructions make O-O-O much easier. But vector processing and multimedia operations don't really qualify as RISC in the classic sense.
Sun has made some obvious mistakes in the past: fixed-size register windows and delay slots come to mind. Like Intel/HP they have in the past made the mistake of thinking that the compiler can do more than it really can (at least at this point). Parallelism is hard enough to extract at run time. It's much more difficult at compile time. Some of this has to do with maintaining the separate compilation model and speed/memory complexity issues (many compiler algorithms are NP-complete).
And of course, all CPU vendors except Intel/HP have made the mistake of having an inadequate number of general-purpose registers. Ironic, eh? :)
That's not to say the compiler can't do more. It can do a lot. Unfortunately, CPU vendors have not provided the necessary hardware to make this possible. In the future you will see a style much more similar to IA-64: the hardware and compiler conspiring together to extract parallelism, save power, etc.
Here's something to think about: the original intent of RISC was to allow simpler pipeline stages and higher clock speeds. So why does a CPU implementing a CISC-ish ISA have a 50% higher clock rate than a RISC-ish ISA implementation? Deeper pipeline, sure. But don't let labels fool you. There is much more going on in the architecture world.
I do agree with you on the scalability issues of SPARC systems. That's their bread and butter.
Different Architecture (Score:3, Informative)
Re:Different Architecture (Score:1)
Based on what data? Are they faster all around... in every type of computation? I doubt it. I'm sure they are much faster at some types of instructions, and I'd take a Sun over an Intel machine any day, but the performance depends on what types of programs you're running (and how many people are going to be running them at the same time.)
Re:Different Architecture (Score:2)
>>>>>>>>>
I wouldn't go that far. It sound like trolling to me.
It is register starved, has a crappy "grown" CISC instruction set that's a bitch to optimise,
>>>>>
I don't know about GCC, but Intel does a damn good job at it. (Check out Intel C++ 5.0)
has a wierd memory map that, again, is a bitch to optimise, tends to have crappy I/O throughput, cache coherency issues (related to aforementioned instrcution set and memory map),
>>>>>>
What "weird" memory map? I've actually read the whole x86 system designer docs, and the only wierdness is that I/O devices are mapped in at 640K instead of high up where they should be. That and the 4GB virtual address space (think big mmaps) but you can hardly fault it for being 32 bit.
highish latency on task switching
>>>>>>
Only if you use the standard TSS system (which nobody does). Plus, it has lots of optimizations (like global PTEs and delayed saving of the FP/SSE registers) that make task-switching quite speedy. Plus, from the (admittedly old) lmbench benchmarks, Linux 1.3 on x86 context switches faster than any other OS on any other platform. (www.bitmover.com/lmbench, I think). Plus, QNX manages to get. 4.5 microsecond task switches on Pentium-class processors, so I doubt x86 task switching is THAT slow.
(well... that depends on how enamoured you are of memory protection, of course - see AmigaOS/AROS on x86 for near-minimal latenncy, with no mem-protection...) and is JUST CRAP.
>>>>>>>>>
QNX does memory protection (as does Linux
Basically, for a given MHz rating, the x86 will always come out slowest, no matter what test you do. Except running x86 binaries, of course...
>>>>>>
Actually, the P4 runs SSE ops *really* fast.
Thanks to Open Source, I can now happily use Linux on PPC and ARM boxes, and there's finally some nice, cheap PPC MoBos that aren't apple coming onto the market.
>>>>>>>
So you want to pay more to get less functionality?
Seriously, x86 doesn't suck as much as you think it does. Intel and AMD have invested a huge amount of money in improving the architecture, and even if it is often slower in instructions per clock, the sheer clockspeed of x86 chips often make them outperform their competitors. Plus, they have great price/performance ratio.
Re:Different Architecture (Score:2)
Well, I've been happily using SUCKY computers since 1983. There's nothing more satisfying than using a really good kludge. Running a direct descendant of the world's first microprocessor (the Intel 4004) at thousands of times the original performance level is an awesome kludge. Throw in all of the goodies availible for x86 machines (OS's, apps, hardware) and its been like spending 18 years in a candy store.
You can stay up there in your ivory tower and watch the rest of us having fun down here in the fresh air.
Re:Different Architecture (Score:2, Informative)
Re:Different Architecture (Score:5, Informative)
I mean, yeah, they're totally different. And they're faster clock-per-clock (with added benefit to FP stuff).
But a 1.4GHz Athlon blows away a 7-800MHz UltraII for most kinds of computation. A 1 GHz Athlon seems to be about (42, 29) on the (retired) SPECint95/SPECfp95. A 450Mhz Ultra-II (not Ultra-IIi, I'm looking at results for an SPARCstation Ultra-60) gets about (20, 27). That's a bit faster int clock-per-clock, and a lot faster FP. Note that for practical stuff (databases, web, whatever) int is more important. Of course benchmarks are hard to interpret, but this gives you an idea. All the SPEC benchmarks are available at www.specbench.org. Of course there are no Ultra-III results, but I'm guessing it's not going to be 2x as fast as the best x86s (at least I'll wait to see the results before I believe it).
You use a Sun because you want an architecture that will scale smoothly up to 64-way (I *guarantee* that will be faster than any single x86 machine).
Actually if you want to both go fast at the low end and scale well, you can buy an RS/6000 -- IBMs Power3 and Power4 chips are absurdly fast and scale very well (and actually focus on memory bandwidth for database performance). But a bottom-of-the-line Sun is a lot cheaper than the cheapest RS/6000.
Full disclosure: I work for IBM (in software) and I've seen a good bit of internal stuff about IBM chips, esp. the upcoming Power4. Most of that information has now been published in MicroProcessor Review and is now publicly available, I think you'll find it if you poke around...
(even more amusing full disclosure: I'm a huge fan of old Sun stuff, their machines are beautifully engineered. i use a couple old 32bit sparcs for all kinds of things)
Re:Different Architecture (Score:2)
Or if you want a machine that was designed and built by engineers, not by a 14-year-old in his basement (or the Dell equivalent thereof).
I'm not a huge fan of Solaris, but Sun hardware is good stuff. I think SGI hardware is better stuff, or used to be before SGI decided it would be fun to implode.
You buy Sun if you want scalability and reliability. You buy Intel if money is an issue.
Re:Different Architecture (Score:2)
I think the UltraSPARC III is about 2-3 years late to the table.
The old RISC hardware chips are dying off slowly since they can't afford the build the kind of expensive Fabs that are needed to compete with Intel and AMD.
Except IBM, of course. I did read something about the Power4 [ibm.com] (see also this pdf [ibm.com] ) and its emphasis on maximizing memory BW - some of the figures sounded awesome. I was really looking forward to being able to pick and choose between Power4 and 21364(pdf) [sigda.org], but Compaq seems to have throttled the Alpha. As if the IA-64 is better!
For scalability, though, I have to wonder if rack-mounted Alpha4's connected with high-speed interfaces like Myrinet could provide an alternative to hardware like a Sun ES10000. I haven't tried Scyld or MOSIX to know if they make using such clusters a "smoothly scalable" solution. The big Sun SMP machines are nice, but they're also expensive and the aging UltraSPARC II processors are nothing to write home about anymore.
Re: Late (Score:2)
Back then they were way ahead of the game, since Intel/AMD were doing about a quarter their current clock speed, while Sun was doing about half.
I wonder what happened
Re:Different Architecture (Score:2)
Although I agree that the Alpha and the MIPS that SGI uses are going out. Also there is, or at least was, a planned kill date for the PA-RISC chips that HP uses. Those have evidence but the IBM Power series and the Sun Ultra series still seem to sell very well. Companies still need very large machines to take care of business, and the big players still use RISC architecture, architecture that scales.
The IA-64 has had major problems, and is several years late. Sun has pulled its support of the IA-64 chips even after they proved that Solaris 8 runs without a hitch on them because they realize that there is nothing to worry about right now. There are huge performance problems, and being a brand new architecture, it will have to go through several generations before the bugs are worked out. Meanwhile the Ultra chips have a reputation of having the fewest number of bugs upon release.
The big SMP machines may start to make more of a come back. The idea of having applications on the net and having bare bones machines in peoples homes is the directions that I see all of the major players going. The best machines to handle large loads are not x86 style machines but the "slow" work horses that today are all RISC.
Aging Ultra II processors may be nothing to write home about but they work, and work, and work. The chip is and more importantlyand more importantlystable, and reliable, and so is the OS. People can depend on their ability to handle large loads, even if they are not exactly the fastest things out there.
Rack mounted Alpha's could most likely handle the load of the Sun ES10000, but there are several drawbacks. First you have to deal with Alpha's, which, althought are amazing machines, have issues of lack of support and development, and frankly are very expensive. I am currently working in a High Energy Physics lab with huge numbers of relatively new Alphas and trying to find software that is not a rev or two behind the other UNICes is difficult. On top of that the Ultra 60 that was purchased was a much better deal with better support. Alpha clusters are powerful but a
multiprocessor Sun box will do a more effecient job because the processors are designed for it.
Slight nit to pick... (Score:5, Insightful)
As you implied, SMP performance is extremely important to people who buy Sun.
In this case, you wouldn't care much how an individual processor performed; you are most concerned with the performace of, say, a 32-way system and it's ability to quickly shuttle data between processors, memory, and disk.
Our beloved Athlon only scales to 2-way, and it's SMP architecture is now being entirely redesigned with the NUMA hypertransport.
Sun probably suffers in raw MHz and SPEC scores because they put so much effort into the SMP aspects.
And, of course, Sun outsells some (arguably) better technology (Power, Alpha) because they are much more open and their service organization is superior.
Now with deeper pipelines and higher operating tem (Score:2)
And it doesn't enable the Internet either.
I'm a little surprised a technical web site would fall for the pure marketing hype. Next we're going to have an article complaining that the Ultra SPARC IIIs run only at 900 MHz and can't play the new space cadet game [freeyellow.com]. That is a fun game it wasnt free but it was worth the money.
Thank you (Score:2)
Re:Different Architecture (Score:2)
Re:Different Architecture (Score:1)
But then again, your boss really likes you and you know you can con him into buying you hardware necessary do get your work done
Re:Different Architecture (Score:2)
Actually, it's using UltraSparc IIe chips. It's basically a USIIi with the bus interface from a III (based on conversation with a Sun engineer). It was done as an experiment to get the front side bus right, probably for the IIIi, and they figured out that it was better than the IIi, used it in the Blade 100, the X1, and IIRC the T1.
Re:Different Architecture (Score:2, Interesting)
-blah
Imagine a beowulf cluster of these (Score:2)
Re:Imagine a beowulf cluster of these (Score:1)
Comment removed (Score:4, Informative)
Re: (Score:2)
Source of Bloat not Source (Score:3, Insightful)
Windows is bloated because MS piles feature onto feature. The features don't work together, so there's a lot of implementation redundancy. If something goes wrong, a kludgy fix is added, making things worse. Everything gets totally redesigned every 6 months, so there's a lot of backward-compatibility support -- more implementation redundancy.
Re:Source of Bloat not Source (Score:2)
Re:Source of Bloat not Source (Score:2)
Sun's, to name one - it does an excellent job of not only standard optimizations, but also understanding what optimizations work to take full advantage of the UltraSPARC architecture, including the UPA and the other switching interconnects inside Sun boxes. (Remember, folks, Intel & AMD are archaic in this regard, REAL computer companies like Sun and IBM gave up buses in favor of switches for in-box interconnects years ago. That's the reason middling Sun boxes crank the I/O equivalent of the mainframes of a couple of years ago, more than Intel boxes can even dream of. (I know, I'm currently trying to get high performance I/O on an Intel platform (for software reasons), and it's darn near impossible - the I/O architecture is just garbage. If I could use a Suns or RS/6000s for this app, I could save a lot of money, and more grief.
There's a reason people are willing to part with all that money for big Unix boxes: It's that they're simply capable of doing (and thus cheaper for) the seriously heavy duty jobs. And no, Beowulf clusters don't help I/O, before anyone suggests that...
Re: (Score:2)
Other factors in Sparc's favor (Score:3, Informative)
Apples and Oranges (Score:2, Insightful)
Duh!
Re:Apples and Oranges (Score:1)
Re:Only 900mhz? (Score:1)
Cheap hardware (Score:1)
Some people think a machine with 256MB of SDRAM and two 30GB IDE drives make for a good server. Never mind considerations of ECC memory or mirrored disk. Let's not even talk about using overclocking in a server!
I've seen some pretty good reliability from properly architected Intel boxes. As good as a real Unix server? Not yet, but they're getting ever closer.
Re:Size (Score:2)
Re:Size (Score:2)
Re:Yawn! (Score:1)
Re:Yawn! (Score:2)
Runner up goes to Black Lab Linux. [blacklablinux.com]