[Ion]
WCG Team Assistant
- Joined
- Sep 15, 2009
- Messages
- 13,391 (2.38/day)
- Location
- Raleigh, North Carolina, United States
System Name | Niedersachsen / Ribe / Minsk |
---|---|
Processor | i3 3240 / i7-3520M / 4x Opteron 6376 @ 2.86GHz |
Motherboard | BIOSTAR H61M / HP Q77 / Supermicro H8QG7 |
Cooling | Stock / Stock / 4x 1U G34 |
Memory | 1x8GB / 2x4GB / 4x4GB |
Video Card(s) | GTX260 / Intel HD 4000 / nVidia GT310 |
Storage | 80GB Intel SSD / 256GB Intel SSD / 2x 60GB SSD (RAID1) |
Display(s) | Dell 3007 + HP 2245w / 12.1" 1366x768 / None |
Case | Antec NSK3480 / HP / Supermicro 1U |
Audio Device(s) | Onboard |
Power Supply | Enermax 500W / HP 130W / Supermicro Gold 1400W |
Keyboard | IBM Model M |
Software | Windows 7 (Niedersachsen/Ribe) / Linux Mint 17.2 (Minsk) |
There are some interesting things to note. The Intel Core i5-2500 and the Core i5-2500k are effectively the same processor on paper, but the 2500k has a higher average (see below). I'm assuming this is since the 2500k is commonly and easily overclocked. This proves that the data is indeed dirty, but also shows what "can be expected".
..and yes, I had to double check, but you are reading that correctly. For reasons I don't fully understand, the Core i5 apparently beats the Core i7 overall. Again, it was taken from this data.Code:Avg OS Proc Name 256.13 Win Intel Core i5 2500 3.30GHz 274.80 Win Intel Core i5 2500K 3.30GHz 190.90 Win Intel Core i7 2600 3.40GHz 197.87 Win Intel Core i7 2600K 3.40GHz
Edit of sudden realization: I'm starting to suspect this is all per core. The Core i7 is slower because hyper-threading ultimately slows down each work unit (by ~25-28%), but allows for twice the number of work units to be run at once. All work here is measured in CPU time per work unit. I'll update my chart with core count and take that in to consideration when reporting final performance.Code:C/Hr-faah C/Hr-fahv C/hr-cep2 Avg OS Full Name 313 238.9 216.5 256.13 Win Intel Core i5 2500 3.30GHz 238 178.5 156.2 190.9 Win Intel Core i7 2600 3.40GHz 76.04% 74.72% 72.15% 74.53%
This explains why the 4 core/4 thread i5-2500 appears to beat even the the 6 core/12 thread i7-3960X.
Higher performance and "free as in free beer", as the saying goes. Pretty simple to set up too, if you're going with Ubuntu. I hear there isn't much of any performance increase for AMD, but I'll check that out in a second.
Playing with those numbers, that's a 169.06% increase. That's insane!
..and I just realized I'm crunching numbers about crunching numbers.
http://lh4.ggpht.com/icJ8K8L1HwyK7x...VN8LBlv-SQEqa7ydD8tLqAyPkH7rdoEhLNhiVl87=s240
It's amazing--for dedicated cruncher rigs, I see no reason to use anything else. Only a couple exceptions--if I need to run the FAH GPU client or I need to use a WiFi card that Linux doesn't like, but otherwise it's unbeatable. Once I've decided for sure that the i3 is sufficient for gaming I really need to put Ubuntu on the 3930k--that should be at least another 5k PPD
Although I like Mint better
And yes, the difference for AMD systems seems to be smaller--clock-for-clock my Linux FX8350 only seems to do about 7% better than Norton's FX83x0s with Windows.
I'll probably be setting up the Atom system tomorrow some time--later than I wished, but it will have to do