quick RAM question

[Stephen Hughes]

Hi Guys,

 

Quick question for some of you experts:

 

Would 64GB RAM @ 2133Mhz be a lot better than 32GB RAM @ 3000Mhz for max and vray?

[numerobis]

I think this is a question that only you can answer for yourself. How much RAM do you need for your projects?

If you don't use more than 32GB RAM then buying more than 32GB wil give you... nothing.

But concerning the 2133MHz vs 3000MHz question... a difference in rendering speed is almost non-existent. And the faster RAM can be even worse https://rog.asus.com/forum/showthread.php?53207-Cinebench-score-and-differant-Ram-Speeds

I would go for 2133MHz or 2400MHz RAM. Everything higher is a waste of money.

[Dimitris Tolios]

Even if you don't need 64GB, I would not see any benefit going for more than DDR4-2400 speeds. You will barely see any benefit going to 2400 over 2133.

 

Current generation of CPUs cannot really utilize those speeds, and if overclocking interests you the least*, you should know that CPU overclocking doesn't work well with RAM overclocking with the current generation of Haswell-E CPUs, and you are hindering the max atainable CPU clock for a given vCore setting by making the CPU trying to boost both itself and the RAM.

 

Needless to say, that tangible benefits from overclocking a CPU are there and clearly measurable, while tangible benefits from faster than 2133 RAM are simply not there, especially for rendering tasks. You can get the 2400 cause those are not that more expensive, so that you get the satisfaction of "not going for the cheaperst option", but I would have zero remorse actually going for the cheapest option.

 

Remember, Haswell-E for s2011-3 & DDR4 is not that much different than Haswell for s1150 & DDR3. And for s1150, 1866 speeds were already "good enough, and you won't see much of a difference going 2133 or more".

 

DDR4 is an excuse for companies to charge more for something that has no benefits for the average consumer over DDR3, simply because RAM hasn't been the bottleneck in our systems for quite a few years.

I see no justification going for "boutique" memory vendors like Corsair, that charge ridiculously more for a product they just re-badge (corsair doesn't make memory) and test for overclocking, that outside of memory benches offers nothing but looks.

 

 

*you cannot say "no" to this question really, cause DDR4-3000 is not manufactured yet, you just pay for a set that is "certified" to overclock to 3000, and you have to go in the BIOS and manually put in the settings for the "desired" speed, and pump the default voltage higher = what you do for overclocking anyways

[deadkaiser]

lol that's a long response where a simple no would have sufficed...... So yer No unless you use more than 32GB when rendering if you want to check this open resource monitor when rendering a scene (ctl shift exit - Performance tab)

If you are using more than 32GB of ram you probably are using tons of complex geometry like 50 high detail trees eg in that case you real aught to get multiscatter or forest pack instead of buying more ram..... So its not hard to use more than 32GB But its rarely needed!

 

On another note does anyone even have any proof that Vray render times is improved by ram Mhz instead of timing? I know for some programmers timing is more important....

[numerobis]

another comparison (2133-3000MHz)

 

http://www.tomshardware.com/reviews/corsair-vengeance-ddr4-2666-overclocking,4011-3.html

[Francisco Penaloza]

lol that's a long response where a simple no would have sufficed......

 

I think you should have a little more respect for someone who took the time to give an educated answer.

It is always better to know the why, rather just follow random opinions, like yes or not.

Edited February 1, 2015 by fco3d

[deadkaiser]

another comparison (2133-3000MHz)

 

http://www.tomshardware.com/reviews/corsair-vengeance-ddr4-2666-overclocking,4011-3.html

 

Interesting, but it still does not show MHZ v timing for actual Vray rendering which what I was really interested in.... but assume the difference is negligible.....

 

on another note I love how PC and hypersensitive Americans are lol chill

[Dimitris Tolios]

There are few exchanges of large quantities of data between the CPU & RAM when rendering for timings or speed to matter that much.

 

Also, timings are "relative" to the frequency. Having larger/looser/slower timings (= in layman terms delay between cycles), but much higher frequencies (= more cycles per second), might get to lower real latency.

 

e.g. a DDR3 2133 CAS 11-11 module has lower real latency vs 1866MHz CAS 9-11 , or 1600MHz CAS 8-9 and 1333MHz 7-8.

Sure, tighter timings are nice, but for the same or very similar speeds, and thus DDR4 despite having notably slower timings than DDR3 "on paper", makes up for it with higher speeds.

 

Forgive my rant, I just have to built up my case & argument. If that exposes a mistake in my thinking process, the better:

I get to learn something new if someone points it out.

 

Also, don't rush into conclusions:

Afaik there is no one (North) American born & raised answering in this thread so far, for your generalizations to have validity

[Stephen Hughes]

Thanks everyone (especially Dimitris) for your replies,

 

I do like the analysis of why the higher clock speeds are not really worth paying for. Helps me understand the reasoning in choosing something i know little about. And it just saved me £200...so Forest pack Pro can be seen as a wee bonus!

[deadkaiser]

There are few exchanges of large quantities of data between the CPU & RAM when rendering for timings or speed to matter that much.

 

Also, timings are "relative" to the frequency. Having larger/looser/slower timings (= in layman terms delay between cycles), but much higher frequencies (= more cycles per second), might get to lower real latency.

 

Also, don't rush into conclusions:

Afaik there is no one (North) American born & raised answering in this thread so far, for your generalizations to have validity

 

O it was a general statement like “relative” timing to frequency but let’s leave that before I offend someone!! Consider it all tongue in cheek.

So if that’s all the case should we be recommending Virtual memory as a viable option to those people with only 8gb/16gb systems?? I’ve never use Virtual memory just intrigued….

 

OP Glad to see you have found new things to spend you cash on ...... i personally would have recommended beer but i guess Forest pack is a nice bonus! love those forest colours!

[xjapanese]

I appreciate all the information given here still I would like to repeat the question :

 

What are the aboslute practical dimension of RAM speed while using 3ds max and vray or other engine?

1333 to 2600Mhz ? Render times are not affected in any way so that doésn't need to be discussed..

but does that mean the same when other practical operations are performed..like loading copying na other basic operation in 3ds max..

..and if the conclusion is that it does not have any relevant difference ( even to be discussed) for 3ds max..is it the same with AE..

 

Thanks

[deadkaiser]

ram Mhz really has a rather small impact on most high performance program's whether that is PC games v 3dsmax or photoshop. I'm sure there are exceptions however if you were to replace 1333Mhz with 2600Mhz i seriously doubt you or anyone could tell the difference.

 

So in summery people really need to stop worrying about ram and start thinking about OCing there CPU's when it comes to rendering.

 

The only really reason you need more ram is if you are using all of your current ram during renders.

As for AE rendering i bet it's the same however I don't have any proof.

[Dimitris Tolios]

Price differences easily warrant going for DDR3 1866. DDR3 1600 or 1333 are not that cheaper, while the performance diff from 1333 to 1866 is considerable in benchmarks. No, most likely you won't see it with naked eye etc.

 

Past 1866, returns are greatly diminished. The only differences I've seen myself and online going for faster RAM, is when you pump very fast GPUs through games, and the difference is NOT notable with naked eye. Imagine having a system that with DDR3 1600 will be pumping 160 fps @ ultra/all-out quality, and with DDR4 2400 will be doing 165-170 fps. i.e. you had something blazing fast, and you just go marginally faster.

 

This kinds of improvements in viewport performance are not linear. e.g. if you had a 10% improvement from 120 fps to 132 fps, doesn't mean you will get 11 fps when you had 10 fps. The RAM speed is not the limiting factor in most operations. When RAM speed becomes a limiting factor, 90+% of the time you will be dealing with an already overclocked, top of the line CPU and a top of the line GPU (or more GPUs in SLI). i.e. the top of the line RAM will be the icing over a very big and expensive cake. By itself it does nothing.

 

That's for viewports.

 

For rendering, lets just say it doesn't really matter. As mentioned above, a CPU with more cores @ same speed or same cores but higher clocks (overclocked or not) & DDR1333/1600 will beat handily a more basic CPU with ultra-fast RAM, so your money should better go the CPU side, way before you just thinking RAM.

[Nikolaos M]

The truth is that rendering times are very slightly affected by RAM speed (I saw a 1-2% increase in Cinebench R15 score moving from Corsair Vengeance 1600, C9 to G.Skill Ares 2400, C11, but someone could argue that this is only a C4D benchmark), but the cost difference between these two RAM speed categories is so small that make me think, why not? Other everyday tasks profit more, because they are more memory intensive, like zipping-unzipping files etc. So, my conclusion is that there is little to gain, performance wise, and not much to lose, price wise. Considering all that, I would go for the faster RAM.