Benchmark Results: Synthetics And Power Consumption
Our Sandra benchmark results are relatively predictable, given what we know about the CPUs in each machine. The memory-bandwidth test yields the biggest performance delta, with the CyberPower rig pulling way out in front thanks to its memory running at 1,600MHz compared to the 1,333 MT/s memory in the AVADirect box and Alienware’s relatively staid 1,066 MT/s memory.
Here’s an eye-popping chart. Even while sitting idle, CyberPower’s rig consumed more than 80 watts more than the Alienware; but that number nearly doubled when while running our Far Cry 2 benchmark.
6
Comments
Sponsored
Latest Desktops News
- 03/02 – Apple is Now World's Largest PC Company (If Tablets Are...
- 27/01 – Ex-Palm Chief Rubinstein Has Resigned From HP
- 25/01 – Apple Breaks Tech Records in Q1; 37 Million iPhones Sold
- 19/01 – Apple Recycling Programme Extended to UK, France, Germany
- 18/01 – Alienware Crams Gaming PC Into Console Form Factor
Latest Desktops reviews
- 10/10 – Nettop Round-Up: Four Tiny PCs, Benchmarked And Reviewed
- 05/10 – Building A Liquid-Cooled MicroATX Gaming Monster
- 27/09 – Intel vPro: Three Generations Of Remote Management
- 26/09 – System Builder Marathon, Sept. 2011: System Value Compared
- 22/09 – System Builder Marathon, Sept. 2011: $500 Gaming PC
Nice to see a comparison that doesn't put Alienware on top all the time. Yes, they have a fantastic reputation and some truly slick kit, but when it comes down to the meat and potatos of it, Alienware's insides always seem to be a little underpowered or overpriced (depending on your viewpoint).
Comparing an i7 920 to an i7 940 is a little redundant though considering they were both overclocked and therefore the clock speed difference was maintained.
What I'd like to see (and this is beyond the scope of this article) is a 920 or 940 clocked to 3.2GHz and run against a 965 Extreme at stock - goes the increased QPI transfer rate of the Extreme make any tangible difference in real-world conditions?
The question remains: who will be stupid enough to pay $1200+ for the upcoming i7 975 when you can overclock a $270 920 to 3.4+GHz anyways..
@starmate:
That's always been the question since overclocking has been viable and commonplace. And this is why I posed the question.
Both the 920 and the 940 have a QPI rate of 4.8GT/s so I won't bother with the 940 - I'll just get a 920 and overclock it to match the 940 and save about £200.
The same goes for the 965 and 975. As the 975 is Extreme too, I'd assume it also has a QPI rate of 6.4 GT/s like the 965. If so, get a 965 and overclock it to match the 975 and save about £200.
However, what I want to know is how much of a difference does the QPI rate make? Does an extra 1.6 GT/s make a difference in single-CPU set ups, and therefore is worth getting an Extreme for maximum performance? Or will we see benefit in dual-CPU set ups like Xeon boards and Skulltrail 2?
And is it just me, or does "Gigatransfers per second" sound like a total, 3am bullshit decision "yeah, we'll rate the QPI in Gigatransfers - that'll do!"
I agree. No point in the 940 or 965 if you're willing to overclock the 920. I've read that the 965 isn't at all stable beyond 4ghz, and the 920 has been known to overclock beyond this,so where is the dilemma? 920 oc'ed, 6 gigs of 1600 DDR3 and a decent vid card and heatsink and you are laughing.
That's the route I'm going matt77 unless X58 Skulltrail proves to be something significant, hence me whittering on about the Extreme's higher QPI rate.
Am I the only one who notices that despite all other articles disabling the PhysX acceleration for 3DMark this particular article didn't?
Omitted the Graphics Score from this test suite as well, whilst showing all the other parts of the score breakdown. I suspect that the reviewer was giving all he could to help Alienware not look uncompetitive.
I also dislike that the prices weren't mentioned. All under $2500 but BY HOW MUCH? Alienware probably cost the most.
I have heard that alienware take exception to bad reviews and stop handing out the benchmarking samples if they are made to look bad.