Early Wednesday morning 1 December, our colleagues at The Inquirer ran a story entitled “HP expected to drop Trucluster from HP-UX” on their web site. According to the story, unnamed sources claimed that HP soon will announce that TruCluster technology is unlikely to be incorporated into HP-UX. More here.
The Truth About Tru64 UNIX and HP-UX
About The Author
Eugenia Loli
Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker.
Follow me on Twitter @EugeniaLoli
Here is the confirmation: http://biz.yahoo.com/bw/041202/25225_1.html
It simply isn’t going to happen. It’s a wonderful offering that HP cannot achieve performance or feature parity with.
The ROI on Itanium has been abysmal, one of the worst in history. Intel and HP have together invested nearly $10 billion and over 12 years developing the processor. According to the latest Gartner reports, Itanium processor sales are estimated to have reached barely 15% of the total investment, with an additional $3 billion slated for Itanium reserach over the next five years.
And sadly, it is this at the hands of this quite measurably failed investment that Tru64 and Alpha are being strangled. It’s certainly idealism at its worst… while the IA64 ISA provides the most potential for future optimization, on both the sides of processor implementors and compiler designers, the future potential of the Alpha ISA is still vast as well. In the mid-90s, following the release of the 21264 and before Compaq bought DEC, the Alpha processor engineers were dreaming of a 21364 much like the Sun Niagra, with a massive number of cores (12, I believe?) on a single CPU all connected via a cache coherent crossbar architecture. Meanwhile, the first dual core Itanium, the monsterous 1.7 billion transistor Monticeto, won’t even debut until next year.
What has happened to Alpha nowadays? Intel chopped its branch predictor design, the best of any processor at the time, and attempting to modify it to the P4. As incredible as it was, it was still not accurate enough to make its gargantuan and overly cumbersome 28-stage pipeline design efficient.
Despite HP’s attempts to crush it as they stand by their monumental investment failure, and despite being hacked apart for use in markedly inferior processor designs, the Alpha refuses to die. I just hope the Alpha remains strong enough that HP will begin to reconsider their failed investments.
Unfortunately, I think their attitude at this point is that they’re in far too deep with Itanium to write it off as a failure. I guess the best we can hope for is to simply wait another 10 years for the core and compiler designs to catch up with the remarkably overdesigned Itanium ISA.
…you made a fortune selling volume parts: why don’t you do the same now? sell budget-quality Itaniums to hobbiest. You already have Windows. Hobbiest will start picking this up, but soon everyone will. Superior FP -> superior gaming no?
Throw on a cheap x86 chip to help with the emulation like Sun does. Sell it at a loss till market builds and then profit. If I could buy a cheap Intanium I could. As it is I’m looking at used UltraSPARCs.
Does anyone have a breakdown of Alpha revenues/profits vs. the other architectures?
I understood that the only bit of Tru64 that was going to be put into HP-UX was the clustering stuff. The reason that HP gave was “It works better than our clustering does”. The rest of the story is not really newsworthy.
IMHO, HP is up a gum tree as it has really be its server business on Itanic and there is no way they can retrieve any ROI on their investment in it. As an HP Shareholder(ex Compaq) I have repetedly tried to get some answers about this from senior management. The HP CEO should be sacked.
…you made a fortune selling volume parts: why don’t you do the same now? sell budget-quality Itaniums to hobbiest. You already have Windows. Hobbiest will start picking this up, but soon everyone will. Superior FP -> superior gaming no?
Throw on a cheap x86 chip to help with the emulation like Sun does. Sell it at a loss till market builds and then profit. If I could buy a cheap Intanium I could. As it is I’m looking at used UltraSPARCs
SUN created volume for their UltraSPARC by selling affordable workstations, servers and so forth. That volume allows them to lower the cost. Intel had the opportunity 5 years ago to get a IHV network working, creating third party motherboards, but Intel didn’t. They told potential OEM vendors like me to f*ckoff; and sorry, people like *ME* make up 1/2 the PC’s sold; we’re the small white box vendor that small to medium businesses turn to for their hardware – not the Dells, HPs or IBMs, but the small vendor at the end of the road who is willing to give his or her customers the personal service of going the extra yards.
Itanium is a monumental failure, and Opteron for all intensive purposes is the success story that Intel would love to emulate, the problem is, the only people willing hand over cold hard cash for their EMT64 equiped are Dell whores and Intel sycophants that have a vested interest in buying Intel products.
For the rest of us, those of us who have a thing called a “budget” and “accountability”; Opteron is the way to go, and as the bandwidth of hypertransport increases, and the demand for systems with more processors, so will Opteron will scale higher.
Besides marketing answers, I’m really interested in true answer on how HP can properly support enterprise when it does not own the: CPU, OS (RedHat, Windows), middleware (e.g. apps server, web server), Volume Manager, Clustering, Database Server? HP owns none of this thing.
I have several Itanium boxes running and they were ALL a pain in the ass to get from HP. Some shipped without parts, some without the OS, and most of the came in much later than promised. They are all excellent boxes, but HOT as hell. I would not want one as a workstation.
These boxes are in a different class than your usual P4, and the boxes that they are replacing are all PA-RISC. My other servers are all Intel Xeons and they are not only much cheaper, but the performance is more than acceptable – if there is a problem I can easily (and cheaply) upgrade, or throw more systems at the problem.
I am running Red Hat and HPUX on these boxes.
Sun has been pushing the lack of committment to HP-UX et al from HP for several months, and regardless of the technical merit of HP choosing to not do this, it’s still a black mark on HP and HP-UX.
Intel has a CPU with lower TCO than opterons called the Xeon with extended memory technology.
I’m sorry but I use both opterons and xeons and I think the Xeon is much better regardless if Intel is a horrible evil monpolostic company. I Hate them but damn, I think the Xeon is better.
Intel has a CPU with lower TCO than opterons called the Xeon with extended memory technology.
I’m sorry but I use both opterons and xeons and I think the Xeon is much better regardless if Intel is a horrible evil monpolostic company. I Hate them but damn, I think the Xeon is better.
Oh, pulease, what evidence do you have to back up the statement that a sole CPU can drastically change the over all total cost of ownership? hell, thats like saying, that a whole car is made different merely by changing the colour of the knobs on the stereo.
Yes, until recently, AMD offerings were crappy, hell, I remember back in the old days, there was actually merit to the “Intel Inside” programme; you knew when you bought Intel, you got quality; because that was a fact, the rest were cheap, crappy copies, and Intel produced the superior chip.
Fast forward a few years, Opteron has been released; it is cheaper, faster and better still, has the blessing of the big players in the industry. IBM, SUN and HP. When it boils down to it, the top tier have said overwhelmingly that AMD is the way to go.
Are you running these in a dual or quad processor configuration? Opterons seem to cook the XEONs on database queries and I/O and seem to do better under heavy server loads. Also, from what I have read 8P Opterons will totally blow away the XEONs because of their efficient hypertransport design. Maybe you are doing rendering on the XEONs or running some kind of encoding/decoding application? Oh, did I mention that our Opterons were cheaper than the XEON alternative? It seems like Intels answer is to try and optimize prefetch and memory access and just throw more L2 cache at the problem…