Slashdot Mirror
HP Finally Reveals The Alpha Marvel
brejc8 writes "HP have revealed the new range of AlphaServer systems. The new EV7 processors show very reasonable performance figures. Revealed by the inquirer the 1GHz versions have very similar SPEC scores as the 1GHz Itanium 2 (INT_2000 of 875 and FP_2000 of 1,500). This is very intersting after HP were rumoured to ensure that "...no Alpha benchmark will be released until the Itanium platform(s) is/are faster"."
Alpha Lives! But Who Will Market It?
I have such fond memories of my 21264 Alpha, its a shame that they are so expensive now though. I always wanted to get a quad-processor board and try to find oil or compile my kernel in 1 min.
HP will probably make sure that these boards and chips are not accessible to the non-commercial Alhpa lovers. So I will have to wait 10 years to get a cheap one off of Ebay.
dollar for dollar, x86 offerings will be much lower in price and support costs in the 2-4 processor setups. I think HP should team up with a company like Apple or Sun and start offering processors on the alpha platform that run the other company's software. Can anyone say OSX on EV7?
Why read the article when I can just make up a snap judgement?
It would be nice to see HP sell Alpha as standalone processors and with a chipset offering, like in x86, for AT and ATX mobos. Custom Made-in-Taiwan parts will augument the system to produce very high power to cost ratios, and might allow the Alphas survival against the Itanium, UltraSparc, PowerPC and others.
Has anyone seen the cheapest-ever duron+mobo combos from ECS where the processor is actually mounted without a holder, via solder onto the board to make the thing really cheap? I know I would buy an offering like that using Alpha. Sure I know stability and secure hardware are the main reasons people buy full servers in the first place, but not all applications demand stability and flexibility to match the power, and I havent seen offerings in this region outside of the Wintel arena.
"Give orange me give eat orange me eat orange give me eat orange give me you." -Nim Chimpsky
"foremost for Windows" ? When most benchmarks for Itanium2 systems are posted using HP-UX or Linux ?
HP is pushing IPF systems as the systems you want if you want flexibility in running Windows, Linux or high-end Unix (read: not Linux 2.4 or 2.4++).
Rotating register files and lots of the other features that Itanium has, aren't inherent in their ISA. There is nothing in EPIC that says you need rotating registers. These are just things that Intel thought would be really useful and people haven't started exploiting yet.
I think they had a good idea when they designed the ISA, but botched it a little bit on the cpu architecture. However, as compiler technology advances and software starts taking advantage of the "feature bloat" I think we will see a drastic improvement in Itanium performance.
Easy, lets count on how much time and money both intel and HP invested in Itanium? 10 years and 5 billion dollars!
Their VP's and stockbrokers demand a return on their investment and will get it one way or another. In the bussiness world their is no such thing as a "bad investment" sadly enough. They are very brutal to failure and will do everything to save face. Both CEO's of Intel and HP would be canned if they decided to not continue the itanium. Even though this might be the best approach in the long run.
The itanium is a bloated overclocked piece of crap. Its an engineering disaster and the only reason it performs mediocrely well is because it is majorly overclocked with a one pound heat sink and a 500 watt fan that would blow away any case less then 50 pounds. Its true. Its a monster and nearly impossible to program for in assembly. This also makes it perform not to well under Linux since gcc is not very optimized for it.
I agree that the alpha is a better chip and yes the EV7 is actually obsolete( intel canned the ev8). May it rest in peace. Stupid bussinessmen gota love em.
http://saveie6.com/
it's funny how intel says "epic is simple, no ooo complexity" but doesn't mention the all rediculous crap like rotating register files, etc, etc. afaict, ia64 is MORE complex than any risc chip
Rotating register files are part of the original RISC II architecture. The Itanium has some fundamentally good design features. In a standard superscalar chip, a missed branch results in a pipeline flush, which is a huge overhead. In Itanium, all instructions are predicated, so most branch-like structures cease to exist, and instructions which are speculatively executed can simply not be retired. This can lead to a significantly higher instruction throughput. The rotating register files concept is a very good one, as it allows functions to be called without having to write registers to memory (which is slow) or cache (which is not fast).
Perhaps with regard to compiler support, Intel will follow Apple's route (which, is by definition good, since Apple are doing it) and contribute code to gcc (In Apple's case to improve AltiVec support). After all, if Linux runs faster on an Itanium, it would only help Intel sell more chips, which is what the enjoy doing most.
The Alpha has some very nice features, but slating the Itanium architecture because the Itanium and Itanium II (both of which are really intended as proof-of-concept demos rather than commercial CPUs is ludicrous.
I am TheRaven on Soylent News
ooo, a dupe comment for a dupe story!
marvel was already in the works before the HpaQ merger, and it would really make little sense to take a chip all the way to fab w/o at least running SOME of them to try and recoup some cost.
Plus it will probably give Intel a good idea of which components of Marvel to rape for the next gen of the (t)Itanic.
addendum: Dec/Compaq admins/users were also promised at least one more alpha for binary-compatible upgrades as a means to stretch past/current investment in systems while they figure out their next step (i.e. "oh peachy, alpha is dead.. what the fsck do we do now?") Had HPaQ reversed that decision I would bet the suddenly-abandoned Alpha users would cross HP systems of their list of potential replacement (myself, I was looking to switch to p-series IBM boxen).
I was a very short-lived DecpaQ Tru64 admin, but have to admit I fell in lust for the OS and architechure. Our alphas ran superb for their age and the obscene obese demands our Oracle DBA inflicted upon them. Nary a whimper. I still think it's mildly criminal Compaq threw away the horsepower farm simply because they were too stupid to market the things properly.
-'fester
Sure it works great. Unless you care about POSIX sigcontext. In which case the SPARC/UltraSPARC Linux kernel sucks like a Hoover. And yes, the SPARC kernel developers know about it, and no they don't give a damned. It's the only version of the Linux kernel with a broken sigcontext implementation.
Oh, and GCC still doesn't support all of UltraSPARC's 64-bit instructions. And no, Linux distros for UltraSPARC don't come compiled 64-bit. And last I checked (a couple months ago) I still couldn't get glibc to compile in 64-bit mode on Linux/UltraSPARC. It just choked somewhere in the middle of the build due to broken code generated by GCC.
Linux kicks ass over Slowlaris and SunOS when run on 32-bit SPARC processors. But when it comes to 64-bit UltraSPARCs it simply bites. And nobody seems to care enough about the lack of performance or support to make it better.
But Linux/UltraSPARC does make a good web server or the like. An Ultra 5 doesn't cost too much more than a similarly rated PC and is essentially immune to all r00tkits because the script kiddies don't have tools for Linux/UltraSPARC. For Linux/x86 yes, and for Solaris/UltraSPARC, but not for Linux on an UltraSPARC. You just get a weird message in your logs and the program in question just dies quietly.