Jump to content

i7 vs Xeon


Recommended Posts

I am currently looking at adding to my render farm. I currently have i7 processors in my machine which run at 3.5 ghz. The Xeons appear to be a lot lower in GHZ. Are they any quicker then the i7.

 

For single cpu applications, the Xeons have little value.

Some quad core, Sandy-e Xeons are not that expensive, but the only advantage those have over i5s are Hyperthreading, and a larger L3 cache.

Unfortunately the latter does not affect performance enough to make a difference, given that Xeon chips run at a massively slower clock speed.

 

The Xeon E5-2620 is on paper an interesting offering, retailing around $415 in the US (before VAT). 6 cores (12 threads), massive 15MB cache...could be a 3930K at 2/3 the price, but no...limited a 2GHz (2.5 turbo boost), it is outperformed even by 2nd gen stock speed Sandy Bridge 2600K. In Cinebench multithraded benchmark the E5-2620 might get around 6.3 while the 2600K ($300) gets around 6.8. The 3930K ($560) being technically 1.5x the 2600K, does score about 50% above the 1155 LGA CPUs. The newer 3770K ($340) outperforms the 2600K slightly due to improved architecture, while consuming less Watts. It would be the CPU of choice for a render node for me, unless you want to cut costs, when I would go for a i5/3570 (around $200, quad core, no HT). If you don't plan on overclocking (which widens the gap with the Xeons even more), you could always go for the non-K versions, and save $20-30 on each chip.

 

So, value-wise, single Xeons cannot beat i7s, unless you start talking multi-CPU systems, where Xeons and Opterons play their own, pretty expensive due to MoBo selection game.

Link to comment
Share on other sites

Thanks for such a comprehensive response Demitris. The I7s I have in my current render farm I built myself. I'm debating if I should build my additional machines with i7s or purchase render machines. The render machines such as the Boxx Render Pro all use Xeons. But I was surprised at how low the GHZ were in the Xeons. In my limited hardware knowledge I all was thought when it came to rendering the higher the GHZ the faster rendering time.

Link to comment
Share on other sites

There are many factors that affect the performance of a CPU.

The micro architecture differences on hardware level, the software optimization etc. Of course when you have comparable architectures, the more cycles per second you operate at, the better. It has been some time since the "GHz" are irrelevant talk, when the lower clocked Athlon 64s were whipping the floor with the P4 architecture, which was since then abandoned.

Fortunately or unfortunately, modern Xeons and i7s share the same micro architecture in more places than they don't. IOPS are not the same Hz by Hz, but the "mainstream" i7s have such a clock advantage, that the Xeons simply fall short when they go head to head and ofc $ for $.

 

If you want the maximum performance out of a single box, Xeons have the advantage of scalable performance gains: you can add 2 or more processors (tho more than dual slot mobos are not easy to find for Xeons), and ofc you nearly double the performance. Do that with an already fast chip, like the fast E5 xeons who will set you back more than $/€1000, and yes - you get a fast data crunching beast.

 

+Intel+Xeon+E5-2620+%40+2.00GHz"]See how dual E5-2620 fair in a standardized CPU benchmark vs. 2 gen Sandy i7 & i5s

 

It boils down to space and cost.

 

Boxx asks for $3300 for that slick little RenderPro box, featuring 2x six cores and 8GB ECC 1600 Ram. For the same amount of money you can setup 5-6 single CPU i7 nodes with the same Ram, counting mAtx mobos, build in GPU, Win 7 64 Pro OS, a small 2.5" HDD (64GB class SSDs are probably easily within budget too) etc. And yes, I do believe that this lil farm it will be 3x the performance of the singe Boxx machine, yet bulkier, will draw more power though the wall (at least 3x-4x...), requires you to build and maintain it etc...

 

For me it would be an easy choice: getting 1-2x i7s, trying to "optimize" my workflow around them, and then start adding if an when I need a few more. Others will stand by their "need" for ECC peace of mind", the size restrictions, the "1st class" support, or whatever they judge the most important in a factory built workstation or render node.

Edited by dtolios
Link to comment
Share on other sites

We are talking single CPU render nodes here - not servers - and as far as value for money goes, unfortunately I cannot agree with you.

 

Unlike the i7 which comes in a handful of versions, there are more than 10x SB-E Xeon models, ranging from $200 to more than $2000 for the CPU alone. Ofc the $1500+ Xeons are "better" than i7s for renderings, but as an over-generalization (yet far from a lie), you could guestimate that you won't get better performance with a Xeon that doesn't cost 2x the price the i7 you are trying to compare with, and even then you ofc won't get nearly 2x the performance.

Link to comment
Share on other sites

Guest calumreid

I think i agree with Dimitris. Im certainley not as fluent in all the cpu knowledge as Dimitris is but from what im reading it sounds like get an i7 if you are just getting a single cpu, but if you want to build a pc with say 2 cpu's then you should get 2 xeons?

Link to comment
Share on other sites

I'm pretty sure Supermicro make dual xeon mobo's.

 

Supermicro and Tyan are the well known ones, popular in server and supercomputer cluster applications, as well as workstations.

EVGA and Asus occasionally produce more enthusiast oriented boards for Xeons.

Amazon.co.uk might be your source, otherwise search for your "local" dealers contact information in the manufacturer's site.

 

Edit: would not do a separate post or thread for that, but it is not a bad idea to test the possibility in your intentions for using AMD bulldozer chips in your rendering nodes. Single thread performance is far from class leading, tho ofc moden CPUs are so fast that most sites use synthetic benchmarks to determine between super-fast CPUs, where fanboys from each side have argued that in many ways are biased, embracing one architecture over the other "that lil bit" more.

 

In pure rendering scenarios / rendering benchmarks, like Cinebench multithreaded test, the FX 8120 beats the i5 CPUs. Not a bad feat. for a chip that gives you 8 threads for around $150. Motherboards for the AMD chips are also slightly cheaper, making up for a pretty appealing package, since multithreaded rendering will be pretty much all those will do: why not do it with the most efficient way for your money? 8120 Overclocks great (just like all I/S-bridges), so the balance does not change.

Edited by dtolios
Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
 Share

×
×
  • Create New...