Time To Choose Some RAM! :)

[ladybastilla]

Hello

I'm getting ready to build my next PC and one of the choices I know I'll need to make is how much (and what speed of) RAM I want to buy.

Here is what I use my PC for:

-- Scrivener / Microsoft Office
-- Photoshop Elements
-- Windows Movie Maker
-- MMOs (World of Warcraft, LOTRO, FFXIV, etc.)
-- Skype
-- iTunes
-- Google Chrome

I have heard that for video editing in a modern build, you should purchase 16gb RAM. What I'm a little confused about is who that advice is aimed at. Do my 5 to 10 YouTube videos each month mean I should be going with 16gb in my next build? Or is that not at the level this advice is aimed at, meaning I should stick with 8gb?

Do I have the money in theory? Yes.

However, there are many aspects to building a PC and one of the biggest dangers is buying stuff one does not need. If 16gb will make a difference I'll gladly do it. I just want to make sure.

Lastly, do I want 1600 or 1866? I don't think I need anything higher then that, but feedback would be helpful.

Have a great day and thanks for your time.

 

[ksham]

There's no harm in getting 16GB of RAM for the extra room in case you need it. And 1600 should be fine.

 

[Legohouse]

Hello,

As mentioned in the above post 16GB should be more than enough for these applications and yeah there is a performance difference between 8GB and 16GB. Considering the speed, 1600MHz should be alright with Intel CPU but if you are using AMD then 1866MHz is best bet. Moreover, in general regardless of the processor you use 16GB DDR3-1866MHz CL8 is better than 16GB DDR3-1600MHz CL7. But both configuration are good. As far as memory populating is concerned I think it would best to go with 2 x 8GB rather 4 x 4GB unless you want to make use of the quad channel memory architecture if your motherboard supports it. There are many websites that can provide decent memory, I will list some I prefer, namely - Crucial.com, newegg.com, memorystock.com etc. Good luck with the upgrade.

 

[PyjamasCat]

For basic editing, you wouldn't notice a difference between those two speeds, it's when you are asking RAM to stream large files the frequency will matter. But even then, you probably won't see any difference that benchmarks will show.

 

[Legohouse]

For basic editing, you wouldn't notice a difference between those two speeds, it's when you are asking RAM to stream large files the frequency will matter. But even then, you probably won't see any difference that benchmarks will show.

Yes, I agree with that.

 

[Tradesman1]

Just to clarify - 1866 and AMD are good with their higher end FX CPUs, if going lower and AMD then 1600 would be better. With Intel I generally start at 1866 for non K model CPUs, with the K models whether it's the new Haswell or the Sangy Bridge or Ivy Bridge again 1866 is good, though I then tend to move to 2133 or better if the budget allows for it

 

[ksham]

The cost increase for 1866 and especially 2133 is not worth it. SB's sweet spot is 1600. IB's sweet spot is 1866. You really won't get much performance benefit going to 1866 or higher. Unless the CPU is an APU. Most apps are also not too sensitive to RAM speeds.

 

[Tradesman1]

On Ksham's comment, I beg to disagree. After a combined in excess of 160 SB and IB builds, I've found, if you want to call it that 1866, to be a 'sweet spot', my approach with clients is to set a budget and basically work from that, which is why I specified the K model CPUs as different. If building based on the K models, most all are looking for high end, so I would go for the highest end components you can put in the system..starting with the CPU, the GPU, an SSD opposed to platter drives etc. Then when you get to the DRAM go for the best combo of freq/CL that the budget allows for the amount of DRAM desired. If building high end - every performance edge you can put in adds to the overall finished product.

 

[ksham]

@Tradesman1: what does that have to do with my post? You didn't say anything about what you disagree on any why you disagree. The only "performance edge" you may see is in benchmarks. Even when I'm building a computer for someone else, I don't use up the entire budget if I don't need to. Nitpicking on upping RAM speeds is pointless to overall performance.

 

[Tradesman1]

I disagree that the base 1600 DRAM is the sweet spot, and further disagree that going higher, whether it be 1866, 2133, 2800, 3000 or whatever is not worthwhile. You say you won't see much 'performance benefit', but the fact is there will be performance benefit, and if building a high performance system, don't you want to get all the performance benefit you can? I do, and so do all my clients, which is why we go through a budgeting process and look into all the components in conjunction with what they will be doing with the system and what they MIGHT end up doing with the system

 

[ladybastilla]

Thanks for all the feedback, guys. How did it skip my mind not to include what I'd be using this with? Oops. My bad.

PROCESSOR: AMD X6 FX-6300 (95W) Six-Core Socket AM3+, 3.5GHz CPU, 8Mb Cache, 32nm (FD6300WMHKBOX)

http://www.canadacomputers.com/product_info.php?cPath=4_64&item_id=053298

MOTHERBOARD: ASUS M5A97 R2.0 Socket AM3+ AMD 970/SB950 Chipset
- Dual Channel DDR3 2133(O.C.) MHz, 2x PCI-Express x16
- GLAN, 6x SATA 6.0Gb/s, 2x USB 3.0, 6x USB 2.0
- ATX

http://www.canadacomputers.com/product_info.php?cPath=26_335&item_id=051996

GRAPHICS: Sapphire (11199-20-20G) Radeon HD 7870 XT WITH BOOST 2GB GDDR5
- 925 MHz Clock, 6000 MHz Memory
- PCI-Express 3.0, Dual-Link DVI, Double Mini DisplayPort, HDMI

http://www.canadacomputers.com/product_info.php?cPath=43_1200_557_558&item_id=057211

***

So I take it that means I want 16gb of 1866? Brand recommendations are welcome / appreciated too along with the answer to this. If I have more then one option, I can hopefully get a sale since this is a little more (on the RAM and the GPU) then I'd originally thought of getting. But I'm pretty sure I'll be thanking myself (and all of you) down the road despite the slightly higher cost this summer.

Also, FWIW, my goal is to get the best, most practical performance that I can for my money. If something (for instance, adding an SSD or doubling my RAM) is going to make my actual life (as in--my experience with my PC, which I use for hours every single day) better, I'll do it. On the other hand, if we're splitting hairs over benchmarks that's all nice and good, but I'm essentially an 'out of the box' user. I'm not an overclocker, I'm running one good video card, you get the idea. Hope that helps to clarify a little.

 

[ksham]

1. 1600 is the sweet spot for Sandy Bridge. Look up benchmarks or do them yourself.

2. Yes; I said not much performance benefit, which does not mean or imply no performance benefit. It is also not worthwhile because you're spending way too much money on so little. Those 2133+ RAM are not cheap. The only reason I can see you upping RAM costs is because you maxed out other components. Which means your client budgets are off the roof. So that's pointless. Otherwise, you should use the money to upgrade another component instead of RAM.

 

[Tradesman1]

Thank you for making my point for me - I don't set the budget, the client does. From that point we look for the best combo of components that will fit the current needs and probable future needs, then we work with those parameters to fill all the performance slots, and as I stated starting with CPU, GPU, SSD, etc DRAM comes towards the end (unless that's a particular parameter they want filled). As far as benchmarks, those are all well and good, and are exactly that benchmarks....a single program running a set defined objective...to me that really doesn't compare to a real life user out there multitasking, running numerous programs, VMs, data intensive tasks, large data sets, video, etc.

So no, higher end DRAM isn't cheap (and in fact some is downright ridiculous) and can go into thousands of dollars, but people want it, I have one guy that was wanting the Trident 2800/CL12 32GB set @ $1,200 and in checking at the Egg came to find they have already sold out of their initial shipments of these sets (and they just went on sale last Sunday). If a client wants something in particular that's going to bottleneck the system in some way, I let them know and explain why, then (and it happens fairly often) we may have to start from scratch and to stay within their budget may have to go to a lower end system or increase their budget, but I can honestly say, DRAM seldom plays a part in any of that.

 

[ksham]

@Trademan1: completely agree with you there. But there's a difference between recommending it to someone and abiding by your client's wants and wishes. Even though 2133 sounds good on paper, I wouldn't recommend it to someone. But if someone must have it, then there is no two ways about it.

 

[Tradesman1]

A good thing about builders getting together and talking, can always agree to disagree I recommend it on given systems (generally K models) because I can actually see the difference when I switch from say 1600 to 2133 to 2400 and higher, granted there's not a ton of people out there that really use a system like I generally do (numerous bowser windows, applications and often VMs with other OSs in them), but also as I said, if they want performance, I try and add every little bit I can, at times it can be robbing Peter to pay Paul, i.e. someone wants a 1,000 or 1,200 PSU, but doesn't need it, might suggest dropping that for a slightly smaller one, and instead of going with a single 240 GB SSD they want, suggest getting a pair of 120GB SSDs and put them in RAID 0 (big boost!!)

 

[PyjamasCat]

I have just tested loading Minecraft HD texture pack, (512x512 R3D CRAFT, Shaders Mod with SEUS, 5.1GB RAM used in total) with 8Gb Corsair Veangence Dual Channel. Data from SSD to RAM I assume, because all my minecraft stuff is there.

Test 1 RAM Settings: 7 7 8 21 1600MHz, 30 seconds to load to main menu.
Test 2 RAM Settings: 9 10 10 30 2000MHz, 23 seconds to load to main menu.

Other things that may be relevant,
CPU: 3570k @ 5.0GHz
Mobo: GA-Z77X-D3H
OS: Win 8
GPU: Gigabyte 670 Windforce X3

It was about 7 seconds faster to load. I watched the RAM usage in Task Manager, the 1600MHz stalled twice and longer than the two stalls in the 2000MHz. (By "stall", I mean did not increase amount being used, just sat at the same level of usage.) How would I test random usage in a real world application? Opening a list of files of different types?

 

[PyjamasCat]

A good thing about builders getting together and talking, can always agree to disagree I recommend it on given systems (generally K models) because I can actually see the difference when I switch from say 1600 to 2133 to 2400 and higher, granted there's not a ton of people out there that really use a system like I generally do (numerous bowser windows, applications and often VMs with other OSs in them), but also as I said, if they want performance, I try and add every little bit I can, at times it can be robbing Peter to pay Paul, i.e. someone wants a 1,000 or 1,200 PSU, but doesn't need it, might suggest dropping that for a slightly smaller one, and instead of going with a single 240 GB SSD they want, suggest getting a pair of 120GB SSDs and put them in RAID 0 (big boost!!)

Two SSDs in RAID 0?! That is insane.
Would you see any difference in performance at all? SSDs are very fast on their own.

 

[Tradesman1]

I'd open a few diffenet apps, maybe run a virus check or malwarebytes, if you do any database work indexing a good size data file, rendering video, even opening numerous browser windows

 

[ksham]

@Tradesman1: still won't justify a RAID 0.

 

[Tradesman1]

Define justify. I've been putting SSDs into RAID 0 almost since they were available and the speed gains are very noticeable (when set up properly and used properly)...Most people don't have any idea how to employ a SSD let alone a RAID (of any kind)...All to often I get systems in for upgrades, repairs, etc and people are so proud of their SSD they put is - and that's the only drive they have - which is NOT how to use a SSD on a desktop, and especially not a SSD RAID.

If your are leaning to price, negligible 2 60GB close to a 120, 2 120s, close to a 240. Setting one up can take a little while if you want best performance as choosing block size can entail testing, though once you have done a few, you can go from experience and set one up in minutes.

I've prob got well over 100 clients using SSDs RAIDs and when they want new systems that's right there at the top of their lists. Put some one on a system with a single, then on an equivalent with a RAID and have them do the same on both, it's very noticeable, unless they simply do one thing at a time, if you multi-task and start using all the cores in the CPU it shows -Which should be WHY one buys a multi-core CPU...if just going to game or do email, then no, no gain

 

[ksham]

Unless you are constantly opening large applications, there's little gain to almost no gain. Why not just get more RAM and keep all the apps open instead of constantly closing them and re-opening them? And in small applications, the load times do not improve noticeably between a SSD and a SSD RAID 0.

 

[Tradesman1]

Am heading out to a job site, will reply in the early PM, but yes, there is a large diff.