anon (anon@anon.com) on 2/25/09 wrote:
---------------------------
>I don't know. I've seen IPF sales breakdowns from a couple of years ago with UNIX
>at about 50%. You think that if HP sells a superdome with primarily HP-UX and one
>Windows partition, that gets chalked up as a Windows sale? That would be interesting.


No, I'm suggesting that it gets chalked up as both a Unix sale and a Windows sale. I don't know that, but I don't see how anything else could possibly be true if Unix sales were ever reported at 50%.

I'll bet until Intel started selling 64 bit x86, there was a fair bit of experimentation around Windows on IPF, to see if it scaled better for Exchange or SQL Server, that sort of thing. I'm sure there were some implementations that went that way, and probably some implemented back then that are still running that way. But I'll bet most of that stopped the day Intel started selling 64 bit x86 servers, and completely halted once they started selling quad core CPUs.

When it comes time to upgrade, do you think any of these existing IPF on Windows systems will stick with IPF when they can buy a 24 core x86 server that will cost much less and avoid potential headaches with Microsoft updates coming out more slowly (I don't know this is the case, but I'd be shocked if they release IPF and x86 updates at the same time) Along with obviously much less choice for third party add-ons, both of which would be pretty important for Exchange or SQL Server. Or even IIS, if anyone would be insane enough to have ever thought it made sense to run that on IPF.

You seem to think that Poulson has a chance of reinvigorating IPF. Tukwila will be the real test. Its quad core, built in 65nm. If it can't reach the performance levels the Core 2 Quad Xeon hit in 65nm, I give Poulson no chance of beating what x86 will achieve in 32nm.

The only thing helping IPF is that AMD has fallen behind and Intel hasn't had to push their CPUs to their full potential the last couple years. Given how far under TDP and how much overclocking headroom Intel's products have lately, they are leaving a lot of performance on the floor that might make the race at 32nm look closer than it really is. But despite that I still think x86 will win, on pure performance, and totally blow IPF out of the water on price/performance. So at best it might steal some Sun customers if they get tired of waiting for Rock or it flops, but they'll never take back any x86 customers.