2x i7 920 vs Dual Xeon E5520- advise please

[branskyj]

Hi all,

Did some digging around and found some helpful topics here and in other forums. There is already another thread here discussing 1x i7 920 vs 2x Xeon E5520 but that is not my dilemma. My current options are:

 

1. Buying two systems i7 920 2.66GHz each and adding them to my current two other systems as rendering nodes in a LAN network.

2. Buying a single system with a dual-CPU motherboard and two Xeons E5520 2.26GHz each on it (again it will be added to my current two other systems as rendering node through a LAN network).

 

I need to elaborate that I do not intend to overclock neither system nor I am interested in playing games. Price difference is not an issue (meaning I am satisfied with what 2x E5520 on a specialized motherboard cost at the moment but not willing to spend more on a better chip). The only interest I have is cutting down rendering times.

Which of the above options is better suited for that purpose? I checked some benchmarks (randomcontrol's frybench, 3dspeedmachine, cpubenchmark and other) but they compare 2x Xeon E5520 to 1x i7 920 and I do not know if adding a second i7 920 means doubling the existing score of a single one.

Is there anything special about having two Xeons E5520 on a single motherboard and running at 2.26GHz each that will overcome having two separate i7 920 2.66GHz each?

 

Any help is greatly appreciated.

[AJLynn]

Well, the dual CPU machine will have some advantages especially with rendering large single frames. It's better to have one machine do a render than to have two machines share it over distributed bucket rendering. (I'm assuming you're using something like mental ray or Vray that has that ability.) For animations, the two systems, being collectively a bit faster, would have the advantage. Then there's space and power consumption (the dual Xeon machine will use less power than the two i7's). But either solution is usable.

[branskyj]

Thank you for that post, AJLynn.

But why are you saying that a dual Xeon machine will handle large single frames better- is it because there is no network transfer involved between PCs or something in the architecture of Xeons makes them better in that matter? I am, as you suggested, using VRay.

Also on a different matter- if I have 4 machines (each of which is powerful enough to be a main workstation while the rest are used as rendering nodes), which one should I use for a workstation- the most powerfiul or the weakest of them? My concerns again are rendering times.

 

Thanks a lot for the help.

[DanGrover]

Thank you for that post, AJLynn.

But why are you saying that a dual Xeon machine will handle large single frames better- is it because there is no network transfer involved between PCs or something in the architecture of Xeons makes them better in that matter? I am, as you suggested, using VRay.

Also on a different matter- if I have 4 machines (each of which is powerful enough to be a main workstation while the rest are used as rendering nodes), which one should I use for a workstation- the most powerfiul or the weakest of them? My concerns again are rendering times.

 

Thanks a lot for the help.

 

I think this would depend entirely on what you're doing. If you're just modelling, use the slowest one. If you're testing lighting and shaders, it may be quicker in the long run to have the quicker feedback during testing of the faster CPU to get it onto the farm quicker.

[branskyj]

Thank you the advise, Dan. Will keep it in mind.

Cheers.

Edited March 9, 2010 by branskyj

[Slinger]

Julian, what kind of budget are you looking at? The new Intel i7 980x will be out in a week or two and it has 6 cores(12 logical) at 3.33ghz. There are many benchmarks out there of it's capabilities. The chip should run about $1000 US. Any socket 1366 motherboard will support it with a bios upgrade.

 

I am looking to build one myself.

 

Intel i7 980x -$1000

Gigabyte mobo- $300

GSkill 6gigs of 1600mhz ram- $200

CPU cooler- $80

 

I will reuse my Antec1200 case, Corsair HX750 PSu, and my eVGA GTX275 graphics card.

 

After selling my Gigabyte motherboard, Q9550, Tuniq Tower cooler, and 8gigs of GSkill Pis 1100mhz ram, I should be paying out of pocket a bit over $1000.

[AJLynn]

Chris, based on the Cinebench file you posted, expect a 73% increase in render speed plus whatever you get from OC'ing. This is a useful spreadsheet I've got here

[Slinger]

Thanks Andrew. I am looking forward to your results you find out.

[branskyj]

Julian, what kind of budget are you looking at? The new Intel i7 980x will be out in a week or two and it has 6 cores(12 logical) at 3.33ghz. There are many benchmarks out there of it's capabilities. The chip should run about $1000 US. Any socket 1366 motherboard will support it with a bios upgrade.

 

I am looking to build one myself.

 

Intel i7 980x -$1000

Gigabyte mobo- $300

GSkill 6gigs of 1600mhz ram- $200

CPU cooler- $80

 

I will reuse my Antec1200 case, Corsair HX750 PSu, and my eVGA GTX275 graphics card.

 

After selling my Gigabyte motherboard, Q9550, Tuniq Tower cooler, and 8gigs of GSkill Pis 1100mhz ram, I should be paying out of pocket a bit over $1000.

 

Sorry about my delay- I thought that was a dead thread and just stoped following it.

Thanks for the advise Slinger, but really the new i7 is way too expensive for me. I also checked some benchmarks with it, but in my opinion its performance do not justify the price tag (for now at least). I decided to go with two systems, i7 920 each and forget about the dual Xeon. I am already doing distributed rendering among my current three machines, I will just add the two new ones.

Cheers mate.