View Full Version : Should I buy RAM?

wolf walt
01 January 2005, 11:25 PM
HI, Ive got a P4/2,6GHz-System with 2x256 MB DDR400 RAM (dual channel). Would more RAM accelerate architectural vizualisation work using fg/gi considerably (no animation stuff yet). Thanks!

01 January 2005, 10:21 PM
of course, you must have 1 GB and you will get more performance

01 January 2005, 11:59 PM
i think that the proc. counts the most for fg/gi, but, sure, more ram is allways welcomed

01 January 2005, 12:09 AM
I would reccomend one other thing in addition to upgrading to 1Gb+ Ram... add a fast, small hard drive to your system, partition a chunk of it (about 3Gb) and use it as a Paging drive. This will make you system perform better overall and make any system paging substantially faster. You can use the other chunk of the second drive however you wish.

01 January 2005, 09:35 PM
how do I set part of my drive as a paging file ? Is it the drive my win os is installed ? or just a separate drive that doesn't have program files and os installed ? I've done it a few years ago and now I can't remember.

01 January 2005, 09:52 PM
I keep trying to get around to making a tutorial on optimizing a windows machine for graphic artists... I'll see if I can get to it next week.

Basically the page file is:

Control Panel> System> Advanced [TAB]> Performance (Settings)> Advanced [TAB]> Virtual Memory (Change)

Select the System Drive from the list, choose No Paging File and press Set... Then select the Drive/Partition you want to use (ideally the first partition on a second drive about 2 - 4.25Gb in size) and set a Custom Size of:

Min: 1.5 your total system Ram up to 2Gb
Max: 2.5 your total system Ram up to 4Gb

Press Set and then restart your machine... If I get a chance to do the tut (with other tweaks and goodies), I'll post a link in this thread.

01 January 2005, 01:43 AM
can't wait for the tut JDex, thanks for sharing :)

01 January 2005, 09:21 AM
I would reccomend one other thing in addition to upgrading to 1Gb+ Ram... add a fast, small hard drive to your system, partition a chunk of it (about 3Gb) and use it as a Paging drive. This will make you system perform better overall and make any system paging substantially faster. You can use the other chunk of the second drive however you wish.

why is having a seperate partition for the paging file important? Does this simply reduce fragmentation?

wolf walt
01 January 2005, 10:31 AM
Thanks a lot for your replies. So I see, I have to upgrade my system here...:D

Any other killertricks (or killermistake...) beside fineadjustment of accurancy and min/max radius-settings of fg/gi?

01 January 2005, 12:17 PM
I recommend to use about 1 gb ram. But this is for who doing some things like me. Small one shot renders...

Virtual Memory is slow than ram. My last choise is vm. but you will always need it. specially in raster applications like photoshop

01 January 2005, 03:12 AM
Depending on how much upgrading you are about to do, another option might be to add a RAID 0 configuration to your system. In short, this means two identical harddrives set up to share bits of files between the two drives (a technique called striping), resulting in what appears to be one drive to the operating system. But this one drive is twice the size of each of the physical drives, since it is the combination of both drives. The advantage of doing this is that this configuration is up to twice as fast as either drive by itself. This is because when files (data) is being moved to and from the drives, roughly half is coming from one drive, and half from the other. So if a drive can transfer 80 MB/s, then a RAID 0 configuration of two harddrives could deliver up to 160 MB/s. There are inefficiencies in certain cases, but when you are dealing with game data, video, rendering of large scenes, etc., you will get closer to double the data transfer rate of a single drive.

The implementation requires two identical harddrives, and the RAID hardware. The RAID hardware could be obtained as a PCI card, such as an add-in SCSI card that has RAID capability built-in. But considering the cost of such a card compared to the price of good motherboards, the best way is to buy a motherboard with the capability built in. You should be able to find a motherboard with RAID for $80-$120, maybe less depending on other features required. My current motherboard has Serial-ATA (SATA) RAID built-in, and I have two SATA 80 GB drives ($59) configured with RAID 0. SATA drives have upto 150MB/s transfer rate capability. I have the potential for almost 300MB/s transfer rates, but this depends on sub-systems on the motherboard, and the actual data being transferred. A given motherboard implementation may not give you full data transfer rates between your harddrive and your RAM. In any case, stuff loads really fast on my system.

Another reason to consider a new motherboard is the question of memory upgrades. If you are upgrading an older system, you may have to buy an older style of memory, which will be useless in the next computer you buy. If you did buy a new motherboard, you could then make sure that you get the latest, greatest memory-type, which is DDR2. Finally, if your current system is not making use of a 'dual-channel' memory configuration, this might be another reason to upgrade the motherboard. DDR memory can be used in pairs of identical sticks of memory on some motherboards in a dual-channel configuration. This can improve performance a bit more over other implementations.

01 January 2005, 08:03 AM
Just something about RAID:

you can read the whole thing, or jump to for conclusion.

In short, the conclusion is that speed increase while using RAID 0 is minimal. It's not worth it.

01 January 2005, 07:12 PM
In short, the conclusion is that speed increase while using RAID 0 is minimal. It's not worth it.

Read the 116 comments to the article and a different conclusion may be reached. Many knowledgeable readers had some serious problems with the benchmarks used and the claims made by the article. The first major clue that something is wrong with the benchmark setup or benchmark programs used is the fact that the pure harddisk performance test was giving much less than a 50% increase. A properly configured RAID 0 setup should have delivered closer to a 100% increase, since the test should simply measuring data throughput from two drives added together. Also, don't expect any great performance gain in a benchmark that simply measures common office tasks!! Any noticeable performance gain would only come into play during transfers of large amounts of data. The benchmarks he used are composite benchmarks that measure total system performance over the course of a battery of common tasks. In other words, the impact of disk performance on the end score is watered down by everything else, and the measurement of the tasks for which RAID 0 may be beneficial is watered down by all the tasks that it cannot benefit.

Also, he claims that the Mean Time Between Failure is halved in a RAID 0 setup. This is a common misconception based on the fact that the 'probability' of failure is doubled in any similar system. (i.e., if the probablility of failure in the first year is one in 10 million, then the probability of failure of a 2-disk RAID 0 array in the first year is one in 5 million.) This does not mean that the MTBF for each drive is magically halved because they are joined in a RAID array. (If I buy a million disks from Maxtor, the odds are good that one will fail in the first week or month, perhaps. But if I only buy two disks, the odds are very, very high that BOTH disks will NOT fail for the full term of the MTBF, which is five years for many harddrives.)

Also, this site has a high-end buyers' guide that actually recommends the Raptor drives in a RAID 0 configuration as the 'ultimate' machine. Which is right? The buyers' guide or the RAID 0 benchmark article?

Really, the whole debate is very similar to the issue of Hyperthreading and dual-processor multitasking. If you are surfing the net and emailing, these advanced features won't make a difference. You have to actually be doing high-end stuff with certain software before the benefits can be realized.

In short, if you are doing something with >10MB files, such as many CGI tasks, then you will see benefits. You probably won't notice any great benefit while doing common home or office computing tasks.

01 January 2005, 08:00 PM
One of two Seagate wide SCSI 320 drives that I'm using in a raid on my Mac just died last week after only three months in service. That was not a happy experience, but at least Other World Computing is replacing it for free.

01 January 2005, 09:59 PM
jaymackey: After reading some other articles on RAID and re-reading anandtech's article (and comments), I admit the article might not be that accurate and has some flaws. But more importantly, it's aimed at normal desktop enviroments. CG workstations are quite a different thing.

But on the original subject, RAM, I would suggest wolf walt to get at least 1Gig of it. You can't really see the benefits of more ram before you run out of it during rendering.

CGTalk Moderation
01 January 2006, 11:00 AM
This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.