Do anyone know how the new Haswell performs compared ti Ivy Bridge? It seems from most tests that they are not so different but I hav not found a good fpu comparison

Nothing specific and probably read some of the same things as you. However, given a recent BSOD on my general-purpose box, I'm seriously looking into a Quad-core version of the new beasties -- and accessories required for the care and feeding of said beastie. (Prime95 found nothing and I'm loathe to pursue more extensive testing unless it happens again. It's my oldest box, built around a Q6600, reliable except for a HDD failure years ago, which resulted in upgrade from XP_x64 to Vista_x64, which it still runs. Still, it's long in the tooth...)

My prejudice toward the new Haswell line, vs. Ivy Bridge, lies in power consumption and a few other enhancements covered in some of the articles. Note my use of the word "prejudice," which is what it is at this point of my timeline --> "prior judgement."

If you get one before me, please let us know how it compares. (I have only one Ivy Bridge box and that because of a Q9550 failure.) I'll do the same but it will probably be at least a month until I make the leap. (My hangup is OS: I don't want to return to Linux, nor do I want to stay with Vista, nor use the kiddie-toy GUI in Win8 - and Win8.1 is well into the future. Decisions, decisions...)

Best of luck, Steinar.

---

"We have met the enemy and he is us." -- Pogo
Greetings from coastal Washington state, the scenic US Pacific Northwest.

I heard that Haswell gives 8% higher performance on a climate model benchmark than Ivy Bridge. That wasn't a Hadley model though, so it might be different for CPDN.

---

I'm a volunteer and my views are my own.
News and Announcements and FAQ

Do anyone know how the new Haswell performs compared ti Ivy Bridge? It seems from most tests that they are not so different but I hav not found a good fpu comparison

Steinar,

As noted in my earlier response, a Haswell CPU was somewhere in my future. The future is now (if I might be forgiven that cliche).

Ivy Bridge i5-3550, 3.33GHz (stock), W7_x64 SP1: 0.74 s/TS with four HadCM3N running (one at 0.73 s/TS).
Haswell i5-4670, 3.34GHz (stock), W8(yuk!)_x64: ~0.66 s/TS also with four HadCM3N running (most of the time - boinc is still clearing the last few WCG tasks, which means the last CPDN task suffers frequent suspensions)

First task (running with WCG Clean Energy project tasks) for four hours ran at 0.61 s/TS. Times slowed, progressively, as WCG work was displaced by CPDN tasks, one at a time; currently: 0.65/0.65/0.69/0.66 s/TS. (The 0.69 s/TS task is the third of four added; I suspect the anomaly has more to do with the task's parameter list than machine factors.) Oldest task now has 141h58m and is 75.64% complete.

[Edit] Another comparison: I have three Q9550 machines doing nothing but running (mostly) CPDN. (One has a motion-detector attached to watch for earthquakes --> but the project has very low CPU use; all the software does is hang around waiting for the earth to shake and watch the clock so it can call-home to Stanford U. periodically). Those machines require 395 to 410 hours per HadCM3N task (I used to think they were fast)!

---

"We have met the enemy and he is us." -- Pogo
Greetings from coastal Washington state, the scenic US Pacific Northwest.