Jump to content

CPU - best bang for my buck?


Recommended Posts

Having some niggles trying to find the most suitable processor for a somewhat low range budget.

 

Looking around the £250 mark but if I can get one that's significantly cheaper but only a few % slower that would help.

 

I started a discussion that wandered into the realm of system updgrade (Well, a new system) and then the subject of CPUs came up. Someone said that Xeons are faster at rendering.

 

So my question is, I found a Xeon that's a tiny bit slower but significantly cheaper than the i7 3770 that I had picked out, so will it be faster at rendering than the i7?

 

Here is the comparison (cpubenchmark.net):

 

My current cpu.

 

The Xeon E3-1230 and it's price here.

 

i7 3770k bench and it's price here.

 

I'm using a laptop currently so also wondering what the speed difference will be like because the score will only go from around 6500 to 9500 which is like 30% but an entirely new system will cost a grand. So want to make sure I get as much increase as possible.

Link to comment
Share on other sites

If a quad core @3,2GHz is enough for you, if you don't want to overclock and if you don't need the internal GPU of the 3770 (K) the E3-1230 could be an interesting alternative. But i would at least look for an E3-1230 V2 which is the ivy bridge based successor. Or, if you can wait a few weeks this one will also be obsolete and becomes replaced by the Haswell based E3-1230 V3. Both are roughly 6-8% faster that its predesessor, so the V3 should be nearly 15% faster than a sandy bridge based one.

But looking at this list it also seems to be priced $25 higher than the V2.

Image1.png

 

 

edit: i just saw that you had already selected the V2 and only the PassMark link shows the old one... so you can forget this sandy bridge part. It would be only a question of ivy bridge (V2) now or Haswell (V3) later.

Edited by numerobis
Link to comment
Share on other sites

If you have a model with iGPU you don't need to install an external GPU. For example for a render node or an office pc where the iGPU is enough.

But i understood that you want to build a workstation. So in this case the iGPU will not help you because you need a separate GPU to get a good 3D performance. Here is the Haswell iGPU compared to some low end video cards:

http://www.techpowerup.com/reviews/Intel/Core_i7_4770K_Haswell_GPU/20.html

 

And a comparison of the cards in this chart with the "real" mid and high end cards - search for GT 440 or HD 7750...

http://www.tomshardware.com/charts/2012-vga-gpgpu/19-Tom-s-Hardware-Index-B-Performance,2976.html

 

 

Concerning your question about the Xeons being faster than the i7... they are based on the same chips with some additional features like ECC memory. So no, Xeons are not faster than i7 when they are clocked the same. Xeons are only "faster" for rendering because the are available with 8 and soon with 10 and 12 cores while the i7 still stay at 6 cores.

You can get them without iGPU (all i3/i5/i7 have iGPUs) and there are no 'K'-Xeons so Xeons can't be overclocked.

Edited by numerobis
Link to comment
Share on other sites

I'm also wondering what the speed difference going from a haswell i7 laptop to a desktop i7 will be?

 

http://uk.hardware.info/reviews/4597/2/48-desktop-and-66-mobile-processors-tested-in-cinebench-115-results

http://www.notebookcheck.net/Intel-Core-i7-4900MQ-Notebook-Processor.86104.0.html

http://www.notebookcheck.net/Intel-Core-i7-4930MX-Notebook-Processor.86103.0.html

 

But it is not recommended to use a notebook for heavy rendering tasks. They will run very hot...

Edited by numerobis
Link to comment
Share on other sites

Xeons are not faster @ rendering.

If you are comparing similar architectures or architectures with a known IPC difference %, you can guestimate the ballpark of performance without trying to decipher sythetic benchmarks and their silly ratings.

 

2670QM - 2.2GHz x 8 threads = 17.6GHz = 100% of current

3770K - 3.5GHz x 8 threads + 7% overhead (ivy bridge IPC gains over sandy) = 30GHz. - 159% raw adv, about 170% factoring in IPC increase.

4770K - 3.5GHz x 8 threads + 7% + 5% overhead (Haswell over Ivy bridge over sandy) = 30GHz. - raw adv, about 178% factoring in IPC increase

3930K - 3.2GHz x 12 threads = 38.4GHz. - 218% raw adv, no IPC increase

2x Xeon E5 2690 - 2.9GHz x 32 threads = 92.8Ghz - 527% raw adv, no IPC increase

 

Many dislike this "core/MHz" aggregate comparisons, but I find them passable when you know what you are comparing.

 

Also, all the synthetics you are using for comparison are a mix of "stuff" that people think are important, and then weigh a final score based on what they think is "more" important than the other.

 

Cinebench is a much more normalized way to compare CPU performance in multithreaded apps.

In this page you can see some scores:

And some more from anandtech

Cinebench 11.5 - Multithreaded scores

i7 2820QM (2.3 GHz x 8T) - 5.32 (that would be 5% faster than yours in theory)

i7 3770K (3.5 GHz x 8T) - 7.61

i7 4770K (3.5 GHz x 8T) - 8.1

i7 3930K (3.2GHz x12T) - 10.19

2x E52690 (2.9GHz x 32) - 24.17

 

The above are within 8-9% of what the aggregate process predicts btw for the single CPUs, and within 11% for the 2P system (scalling is not perfect).

 

When we say "Xeon is not better", we are refering to absolute performance, factoring in overclocking, something not possible with most Xeons, while possible with most i7s. A 3930K overclocked to 4.8~5GHz, does 13.5~14 @ Cinebench 11.5. That is better than any 8-core Xeon does, and single core performance is vastly better also. Xeons are relevant for 2P/4P applications. 1P Xeon values are simply not there for a rendering workstation imho.

 

So, in a "closer to real life" benching scenario, the difference is "there"...while your synthetics give you a 7.5 ov 7.7 or some other "weighted" ratio which means nothing.

 

You already have a fast i7 laptop, so getting a desktop quad @ stock speeds won't be that impressive for general tasks.

 

Still, a 3770K or 4770K should be notably faster than a 27xxQM when rendering, and this is what the above figures are showing, as rendering uses all of the available threads, and gains are added up. Single core performance differences of your 27xxQM vs. a desktop CPU is not nearly as impressive.

 

Another gain, is that ivy bridge and haswell turboboost much more aggressively and can do it for more than one cores. Ofc this is temperature based: the cooler the CPU "feels", the more and more cores it turboboosts.

 

Laptops, and especially the MBP are running hot. I don't think that's an issue personally, and I did use laptops rendering exclusively through school etc, but turboboost will just not happen when the machine is hot, and there is nothing you can do about it.

 

Blowers / base coolers make you feel better, the laptop case feels cooler, but know what? The cooling system of a MBP's CPU/GPU assembly is more or less thermally isolated from the case. And you feel the case getting cooler, not the CPU/GPU. What heats up the case is the hot air moved around internally / blown out from the back, not direct contanct of the CPU heatsink to the case. Putting a blower underneath - where there are no intakes - does NOTHING more than cooling the case. The components will be running just as hot, you just will feel it less. The situation is different with laptops having intakes you can increase the supply of fresh air to. You just have to place the fan strategically to do so...

Edited by dtolios
Link to comment
Share on other sites

You are the man Dimitris, that is one comprehensive answer right there!

 

So you're saying that, judging by the bench results, the processors themselves appear fairly close. But in terms of rendering, the faster i7 processors start to pull ahead more noticeably? as in 170% faster in the case of the 3770k?

 

I was concerned about overclocking due to needing to implement the right cooling measures. Which invalidates warranties etc. But maybe one could get the stock speed and overclock at a later date.

 

I am very much in need of more ram though, and a better GPU. So adding all those together is a factor. Currently I don't like to work in a scene with more than 3million polys.

Link to comment
Share on other sites

Gotta love the internet for answers from people who know their stuff. Very well argued answers from numerobis and Dimitris. I would like a server PC that's on 24/7 anyway, so I'll probably get a macbook, £800-900 custom PC and use the PC as a remote rendering machine if I'm not working at home.

 

I'm pretty sure I'll get a 4gb gtx670 and i7-4770, What mobo would you recommend? Perhaps even something that might work with hackintosh/mavericks in the future?

Link to comment
Share on other sites

So you're saying that, judging by the bench results, the processors themselves appear fairly close. But in terms of rendering, the faster i7 processors start to pull ahead more noticeably? as in 170% faster in the case of the 3770k?

 

 

Well, synthetic benchmarks weigh the average the way their authors saw fit. Rarely a great value to follow them imho. Doesn't hurt to see synthetic tests, but results can definitely be bended one way or the other.

I was concerned about overclocking due to needing to implement the right cooling measures. Which invalidates warranties etc. But maybe one could get the stock speed and overclock at a later date.

 

Overclocking doesn't invalidate any warranty.

3770K and all "K" intel CPUs are designated as "unlocked" processors. Exactly for those that wish to overclock them. It is a big marketing game for intel (and AMD). Intel even sells extended "overclocking" warranties to cover pretty much anything you can do to the chip other than physically breaking it. Not needed in any way really.

 

Same goes for using non-factory cooling. No restrictions.

Overclocking 1155 & 1150 CPUs is VERY easy. I.e. have a decent motherboard, set multiplier to desired level i.e. default = 35 for 35x100, and go through a couple of forums to figure out how much you should bump the Vcore. Literally 20seconds in the BIOS if you don't know your ways.

 

4.3GHz is relatively easy for 4770K, 4.3~4.5 for 3770K. 1GHz is a pretty decent increase in speed!

 

Delliding (that's the warranty breaker cause you physically alter the CPU), might get you closer to 4.8~5GHz. At these speeds, a 4770K might be "threatening" to a stock 3930K.

 

I am very much in need of more ram though, and a better GPU. So adding all those together is a factor. Currently I don't like to work in a scene with more than 3million polys.
To get a good GPU in a laptop, you will pay more than $1000 (or the equiv. in pounds). Ram is relatively cheap.
Link to comment
Share on other sites

I would like a server PC that's on 24/7 anyway, so I'll probably get a macbook, £800-900 custom PC and use the PC as a remote rendering machine if I'm not working at home.

 

I'm pretty sure I'll get a 4gb gtx670 and i7-4770, What mobo would you recommend? Perhaps even something that might work with hackintosh/mavericks in the future?

 

Why would you get a 670 for a remote server? Esp. a hackintosh (so distrubuted VRay RT GPU is not an option)?

CPU power is all you need. Just use the integrated graphics for w/e you need to hook a monitor to it.

Link to comment
Share on other sites

It'd be remote only when I'm not at home - which is about 50% of the time. When I'm home, I'd use it as my main workstation. Why the 670? Relatively low price (£200 for 4gb) and should be good enough for boosting After Effects, 3ds max viewport and (hopefully?) Vray RT. Would the 670 be good enough for VrayRT GPU rendering or will I get a better result with good old CPU rendering on the i7?

Link to comment
Share on other sites

670 is def. good bang for the back for any CUDA oriented boosting.

A 7950 would be far better than any GTX in OpenCL Adobe CC (i.e. 40% better than Titan, for a card that retails $190 in the US now), but would do nothing (nearly) in VRay RT.

 

Give it some time, and you might get 770s in decent prices too. Don't know if the 4GB will drop as fast as the 2GB, but the latter is cheaper now than 670s were 1 year ago, and a 770 is slightly faster than a 680 (shares nearly 100% the hardware with a 680 really, but has memory clocked @ 1GHz higher, and a smarter boosting - i.e. self overclocking - BIOS).

 

For most stuff, CPU rendering is still "the way to go". But for testing lighting or if you can adapt your workflow to the RT limitations, GPUs a leagues ahead. And always much easier and cheaper to make/upgrade multi-gpu systems than multi-CPU ones.

Edited by dtolios
Link to comment
Share on other sites

Interesting. I had never considered ATI or AMD before but looking at a 7950 build with a fx8-8350, 990fx mobo and 16gb ram with case/psu comes to about £700. The gtx670/i7-4770 build I put together was a couple of hundred more. Gut instinct is telling me to go with Intel but maybe it's worth considering other options too

Link to comment
Share on other sites

If you have the money, the 4770K is the better CPU, tho the "bang for buck" of the 8350 is hard to beat.

But you can always have AMD GPU on an intel system, nothing wrong with that.

AMD Radeon 9xxx cards are around the corner, and US retailers are getting rid of all the 7xxx stock at pretty good prices.

79xx cards are compute monsters.

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...