i7 or XEON

[Michael J. Brown]

I haven't found (scene render) benchmarking data for these two processors, so I am hoping someone might be able to give me some useful info here.

 

All other things being equal, which of these two setups will render quickest. And will the difference be substantial or just negligible?

 

single i7 980X 3.33GHz or dual Xeon X5680 3.33GHz

 

 

Thanks.

[AJLynn]

On contemporary render engines, comparing processors from the same family, all you need to do is total the amount of GHz to get relative render speeds. All 3400, 3500, 5500 and 5600 series Xeons are from the same family as i7's, and the X5680 is a 6 core as is the i7-980X. So the i7 box has a total of 20 GHz and the dual X5680 is 40 GHz, so the dual Xeon is approximately twice as fast at rendering as the 980X.

[ihabkal]

or even maybe more than twice faster with vray as it scales up quite well. But you need a looot of ram to feed all the extra cores and keep them at 100% speed.

[Michael J. Brown]

or even maybe more than twice faster with vray as it scales up quite well. But you need a looot of ram to feed all the extra cores and keep them at 100% speed.

 

More than 12GB RAM?

[AJLynn]

12GB should be enough. Just get a config that keeps half the slots free and later on you could upgrade to 24GB if needed.

[Tim Nelson]

If you've got the cash, buy a dual xeon system. And if you get an i7 make sure it has ram capacity greater than 8gb.

[Michael J. Brown]

If you've got the cash, buy a dual xeon system. And if you get an i7 make sure it has ram capacity greater than 8gb.

 

 

Well, actually, that's exactly what I'm in the process of doing right now. Dual Xeon X5680 3.33GHz with 24GB ram.

Edited December 27, 2010 by renderhaus

[Jason Matthews]

I have a few questions. Please pardon the ignorance.

 

First the i7:

Are you supposed to only calculate the cores and not the threads? If so, then a typical i7 has 4 cores and 8 threads then each thread is running at 1/2 the speed of the core. If this is true, then other than price it actually makes no sense in buying an i7 configuration if you can get an equal or better single chip xeon system.

 

How does the extreme i7 differ from the standard i7 other than the bus speed? Does the bus speed have that much of an impact on the render times? If so how would you enter that into the equation of figuring out pure render speed?

 

Second the Xeon:

Is there going to be a substantial difference going from say an L5530(2.4ghz) to say an E5620(2.4ghz)? The only difference I see is the bus speed. Again, how does the bus speed fit into the equation?

 

Thanks,

[AJLynn]

In an efficiently multithreaded renderer, on an i7 with HT, HT adds maybe 15%, plus or minus, to render speed. I assume it's been left on and the speed boost gets wrapped into the overall speed score. This works for Xeons and i7s. A Xeon and an i7 with the same specs will give you the same performance. (So I could just as easily say that it never makes sense to buy a single CPU Xeon when there's a comparable i7 for less money.)

 

"Extreme edition" Intel chips (same as AMD "Black" versions) have unlocked multipliers. The multiplier is the number by which the memory's base speed is multiplied to get the CPU speed. In most CPUs this is hardwired, so to OC the CPU you OC the RAM. In an unlocked chip this number can be adjusted, so you can OC the CPU independently of the RAM, or, since 1600 RAM is so common but the CPUs only call for 1333 RAM, you can OC the RAM without OCing the CPU. Anyway, being able to do both the RAM and the CPU independently lets you max out each without either being constrained by the other. Enthusiastic OCers like that.

 

From what I've seen the most important factor by far in render time is CPU power. The effect of RAM is not as important here as in gaming.

 

As for the Xeons, the 5600 is an update to the 5500 with a newer 32nm fab process (5500 is 45nm) which is nice, and they did improve the cache and a faster uncore speed, which is also nice. The L series Xeons use less power. (We're talking about 20W here, which is good but not a dealbreaker.) The performance differences are not substantial. (Actually, in benchmarks I've seen, the differences between E55, E56 and X56 Xeons with the same cores and clock speed have been within the margin of error - so I can't truthfully say that any of those render faster than the others.)

[Jason Matthews]

Thank you very much for the breakdown. The reason for all of the questions is that I am trying to put together a "benchmark" for our office CAD systems, but I would also like them to double as render nodes during off hours. The i7s seem to be the best bank of the buck in terms of a CAD/Revit stations and as secondary render nodes. I also like the fact that having two workstations with i7s (one with a good video card and one with an onboard video card) I can send the rendering to one of the stations with the lesser video card and continue working on my other station for relatively the same price as one dual Xeon station.

[AJLynn]

Yeah, I'd do something along the lines of i7-800 series with FirePro V5800 for the CAD/Revit boxes and for the rendering, 6-cores if you can budget them.

[Jason Matthews]

Will definitly look into that.