4GB of DDR2-800 or 8GB of DDR2-533

[gentrinity]

Im getting a E6600 and probably an MSI Platinum X, its a good price but its still debatable, dont want any Gigabyte Mobos cause they have compatibility issues with the Mushkin RAM i want.

The MAIN Apps I plan on using are Photoshop, Illustrator, Flash, Visual Studio. Secondary apps will include Premiere, After Effects and either Maya or Lightwave.

Heres the situation;

I can get 8GB of RAM (4x2GB of Mushkin EM2 DDR2-533)
or
I can get 4GB of RAM (2x2GB of Corsair XMS2 DDR2-800)

Obviously going to run either Vista or Pro 64bit

The research I have done;
From benchmarks I have seen, DDR2-800 will most benefit if I overclock the FSB to around 400MHz, there in lies a problem in which I value stability over performance, but I would like to know from people with overclocking experience whether 400MHz FSB can be just as stable as stock.

All those apps that I plan on running are RAM intensive apps, and I plan on multitasking them quite a bit, hence my belief that 8GBs will do wonders. Photoshop alone can use up 3GBs, so obviously 8GBs can be usefull when using 3 to 4 of those apps running, as well as various IE and AV apps in the background.

My main goal is to avoid virtual memory and scratch disks which will keep the performance up when im multitasking.

I have seen benchmarks where DDR2-800 is faster for some apps, but I have yet to find Adobe specific benchmarks, most are synthetic and it doesnt help me reach an educated conclusion.

I have also seen a video in Microsoft where SoftImage uses an 8GB PC to do some crazy rendering performance, so thats also a plus.

Price:
4GB = $600
8GB = $1,120

Since the price per stick is cheaper at 8GBs, its actually not too bad of an investment, especially if it will improve performance considerably when multitasking.

So I ask, which of these two options is better? Will the speed prove to make the performance close to 8GB, or should I stick with DDR2-533 and not worry about any overclocking.


Side-note:
I plan on moving to a Quadro when I get into more intensive video and 3D work, as you all know from the benchmarks on this site, I will see 100% boost for 3D rendering, 80% boost for premiere rendering, and 25% boost for photoshop. Thats a good investment in my opinion. I also plan on moving from a X1900 XT 256MB to a Quadro 1500 when the time is right. Just want to throw this info out there just in case it affect the review of this topics question.

 

[Mondoman]

It's going to take 6 months to get Vista released and apps stable on it. Although Adobe is pretty good about taking advantage of max hardware in their software, they might take even longer than Microsoft to make use of the extra RAM.
Since RAM prices have recently peaked and are dropping, why pay a premium price for lots of RAM today when you will be able to buy it substantially cheaper in 3-6 months' time? If you must buy some now, get 1GB for $100 or so and plan on replacing it later.
If someone is holding a gun to your head to buy now, you will obviously be much better off with 8GB than 4GB once Adobe gets its act together.
PS - If money doesn't matter much, ignore the above and just get 8GB of DDR2-800.

 

[gentrinity]

Forget Vista, im happy with 64bit for now, and even though adobe products arent native 64bit, they still run a bit faster, mainly due to the better memory management.

As far as the pricing goes, the problem that presents itself is this. I wont get the 8GB right away, im only going to get 4GB, but if I eventually want to get the 8GB, then I need to buy the Mushkin brand, so money isnt such a huge concern, cause by the time I will make the upgrade to 8GB will proably be in 3 to 6 months.

Do you feel that the actual speed of the RAM in no way can overtake the sheer capacity of the 8GB?

 

[rwaritsdario]

The boost in performance is minimun from 533 to 800 without OCing.
Running 533 youll match the FSB speed and the world will be happy like that.
Go for the 8GB youll 'need' it for what youll be doing.

 

[gentrinity]

The boost in performance is minimun from 533 to 800 without OCing.

That I already knew, but thanks anyway What Im asking is if overclocking to around 400MHz or so, in order to get 1:1 with 800 MHz, will help reduce the performance hit of going from 8GB to 4GB. I know 8GBs will be awesome, but its also expensive, and if faster RAM can help compensate for the lesser amount of RAM, then I would give it more consideration, considering its only around $50 extra to get faster RAM.

Heres why I ask:

Benchmarks at X-bit Labs show that moving from 533MHz to 800MHz if you overclock will give you around 30% increase in RAM bandwith in MB/s. Thats alot, but I dont know what that translates to in real world scenarios.

I plan on reading this article to get a better understanding of RAM bandwidth, ill post my conclusion as soon as I read it. I skimed through it and I think itll help solve this debate.

 

[Mondoman]

...What Im asking is if overclocking to around 400MHz or so, in order to get 1:1 with 800 MHz, will help reduce the performance hit of going from 8GB to 4GB. ...

You're comparing apples with oranges. OCing will increase performance on CPU-limited tasks (because internal CPU speed will increase in lockstep with FSB increase). 8GB vs 4GB RAM will reduce performance lost due to Win swapping info from physical RAM to hard disk (virtual memory). However, if an app isn't coded to take advantage of more than 2GB of total memory (real + virtual), this latter performance increase will likely mainly be evident as faster switching from app to app, not faster running of a given app. Current apps are stuck with the 2GB limit.

 

[FLA94FD]

Ram speed affect performance like a hyped air filter on a car 3%-5%.

I have no stability problems with a 30% OC's E6600 or 3.1Ghz. I have had it as high 3.3Ghz with 1.375 voltage on air with no issues.

I would go with more ram if you can live with 64bit.

 

[gentrinity]

OCing will increase performance on CPU-limited tasks (because internal CPU speed will increase in lockstep with FSB increase).

Yeah, but you cant get the higher FSB speed without OCing, which is what would make DDR2-800 perform considerably faster. So you get one if you do the other.

this latter performance increase will likely mainly be evident as faster switching from app to app, not faster running of a given app.

Thanks for that tip.

Current apps are stuck with the 2GB limit.

Photoshop can use 3GB on 64bit OS But yeah, youre right, until they become native 64bit apps, most apps will be stuck around that limit, but hopefully the guy at Adobe wasnt lying when he said that by x-mas, we should see the Adobe apps in native 64bit. Heres hoping.

Ram speed affect performance like a hyped air filter on a car 3%-5%.

I have no stability problems with a 30% OC's E6600 or 3.1Ghz. I have had it as high 3.3Ghz with 1.375 voltage on air with no issues.

I would go with more ram if you can live with 64bit.

Thats great to hear on your OCing performance.

I wish to do a little more research on this subject, cause it would be a bit incredulous of me to just go on youre analogy. When I see a 30% increase in bandwidth with synthetic benchmarks, i would like to get real world benchmarks to see exactly what they translate to.

I havent tried 64bit, but from all of the problems ive heard from people, im going to avoid quite a few of them. I dont have any peripherals other than a USB External HD, and the AV I use is Avast, so no probs there. I could always install a dual boot if I get a bit annoid with anything in particular, but from the looks of it, ill be allright.


Side Note:
Problem im finding is that some people do benchmarks without considering the need for higher FSB speeds, so they go on DDR2 comparisons based on 1066MHZ FSB, and 533 will perform admirably, but from seen at X-bit labs, its a 30% increase with DDR2-800 and a 400MHz FSB. Does anyone have any real world benchmarks that translate that? Preferably media apps. Thanks everybody!!!!

 

[FLA94FD]

Core 2 doesn't have bandwidth issues so a 30% increase wont yield the that much of performance increase. There are loads of published application benchmark out there by all means see for yourself.

 

[alex30303]

I Usually just lurk around these forums. But dammit its time to start posting.
With the right motherboard you can easily get your e6600 to 365fsb with little increase in voltage and virtually no increase in heat; and yes rock solid stable; likely 400fsb if you have a good cooling solution(extremely good air, watercooling). Fsb and ram are the only issues they you need to consider. Many boards can easily do past 400fsb without sweating(therefore stabley). I don't know about that MSI board but i imagine it can easily do 350 for 3150 clock speed. I would consider a different motherboard like the P5b Deluxe or DS3 or DQ6. I would get 2-4gb of corsair and in 6 months to a year get another 4-6 gbs of ram. Maybe you can even drop a digit of the CAS latency of the ddr800 if you run at say 360fsb. I would not cripple my cpu with less than stellar ram. Shoot, if you can find it you could even get some ddr 2 667 and run your fsb at 340 without changing a single setting, except fsb on the right motherboard.

 

[Logicsequence]

i built my rig (specs in sig) for the same reasons as yours. I'm a PC Game designer specializing in 3D modeling. Anyways, i chose to go with the 4GB of DDR2-800. I've been very pleased with my decision. I work with a lot of industry standard high polygon models, and i'm also a graphic designer that uses after effects, photoshop, and all that jaz on a regular basis. I've been very pleased with the computer i built, so i would recommend the 4GB of faster ram.

 

[Logicsequence]

Side-note:
I plan on moving to a Quadro when I get into more intensive video and 3D work, as you all know from the benchmarks on this site, I will see 100% boost for 3D rendering, 80% boost for premiere rendering, and 25% boost for photoshop. Thats a good investment in my opinion. I also plan on moving from a X1900 XT 256MB to a Quadro 1500 when the time is right. Just want to throw this info out there just in case it affect the review of this topics question.

This is completely UNTRUE, you WILL NOT see those kind of gains, no matter what you've read here. I'm speaking from actual experience.

Ok, as you know from my previous post i work in the gaming industry, focusing on content creation, 3D modeling, and graphics. I've been doing this for a few years now. You are probably don't like what i have said, and might not beleive it, but it's coming from a a few years experience in this feild and alot of first hand knowledge and information.

Here's a quote from another post a made about the subject a few weeks back:

Anyways, about the Quadro/FireGL/Matrox, i've answered this one alot, so i'll be breif. Most consumer graphics applications derive absolutely no benefit whatsoever from OGL workstation cards. All image editing program, like photoshop or after effects, don't have anything to do with your graphics card whatsoever, it's all CPU.

Now 3D modeling applications (which i use all the time) like Lightwave 3D, 3DS MAX, Maya, etc... obviously require a decent 3D accelerator, but still don't benefit from an OGL workstation card. Rendering is all CPU power, has nothing to do with the graphics card. The only aspect that does have to do with graphics card is the actual real-time rendering of models when your editing them and laying them out, and that is as simple as any modern game does, in fact simpler in most cases.

Now people that really benefit from Quadro/FireGL/Matrox cards are people who do CAD work, with programs that use very complex vector algorithms, computers with extremely large displays (5120+ pixels) and people who need things like dual link DVI. I used to own a FireGL in my previous home workstation, there was no more benefit from it for me than any high end gaming card could provide. So when it came time to build this workstation, i choose a much cheaper high end gaming card, that would perform orders of magnitude better in games and still perform the same in other applications.

For your kind of work, and mine, there is simply NO reason to get an OGL workstation card, at least not one that outweighs the cost. the only, and i mean the ONLY two possible benefits you will see is if you are want a REAL TIME FINAL RENDER QUALITY render of an object being layed out in a 3D environment. And in all my time working in this field, i've never done that, and never wanted to. That would be a nightmare to work with. The other reason is if you use MAX or Maya a lot you can get specialized plugins that will take some benefit from an OGL workstation card. But contrary to popular belief, and propaganda, that benefit is less than 10%.

Don't waste your money on an OGL workstation card now, you can spend it better elsewhere, with things that will affect your performance 10fold more than an OGL workstation card will. Wait until you are building a Xeon or Opteron dual-processor workstation that is DEDICATED to work only before you get an OGL workstation card.

 

[alex30303]

Nice Rig Logic. BTW I want your job, lol!! I would also add that you can purchase 4 by 1gb sticks of faster and cheaper ram for roughly 500 dollars or less. If you find you need more when Adobe releases new apps or Vista comes out you can always EBAY what you have and upgrade then.

BTW, this may be a noob question but where are these 2gb sticks gentrinity is talking about? At a cursory glance the only 2gb sets on Corsairs website are server ram, and i they max at ddr 667.

 

[cd14]

You can run your ram at the rated speed by using the multiplier while keeping your CPU stock. Of course it wouldn't be very efficient due to the FSB bottleneck. I did some benches using Sandra:

1st Run: 266x8=2128, Ram 2x (533mhz) 112% Bandwidth Efficient, 4777 MB/s int, 4785/MB/s float

2nd Run: 266x8=2128, Ram 2x (533mhz) 112% Bandwidth Efficient, 4785 MB/s int, 4774/MB/s float

3rd Run: 266x8=2128, Ram 2x (533mhz) 112% Bandwidth Efficient, 4767 MB/s int, 4787/MB/s float


1st Run: 266x8=2128, Ram 3x (800mhz) 85% Bandwidth Efficient, 5451 MB/s int, 5477/MB/s float

2nd Run: 266x8=2128, Ram 3x (800mhz) 85% Bandwidth Efficient, 5453 MB/s int, 5475/MB/s float

3rd Run: 266x8=2128, Ram 3x (800mhz) 85% Bandwidth Efficient, 5459 MB/s int, 5469/MB/s float


1st Run: 266x8=2128, Ram 3.3x (877mhz) 78% Bandwidth Efficient, 5571 MB/s int, 5592/MB/s float

 

[gentrinity]

There are loads of published application benchmark out there by all means see for yourself.

Ive tried looking but sadly all I can find are benchmarks for more memory, not speed. In that case, going from 2GBs to 4GBs showed definate improvement in Photoshop, but after 4GBs, there wasnt much improvement, which is online with what Mondoman said. I would basically max out most Adobe products so I can multitask beautifully.

I have seen a few apps benefit from faster RAM, like WinRar, but thats about it. If you can help me out man, id really appreciate it.

I would get 2-4gb of corsair and in 6 months to a year get another 4-6 gbs of ram. Maybe you can even drop a digit of the CAS latency of the ddr800 if you run at say 360fsb. I would not cripple my cpu with less than stellar ram. Shoot, if you can find it you could even get some ddr 2 667 and run your fsb at 340 without changing a single setting, except fsb on the right motherboard.

I really dont want to wait that long I know what you mean, but not all apps and processes are sensetive to memory speed. Rendering for instance, benefits very little, thats why im looking for real world benchmarks that focus on content creation to understand which processes will ultimately benefit from faster memory and by how much, that way I know for sure if I really need the speed or not,

This is completely UNTRUE, you WILL NOT see those kind of gains... Don't waste your money on an OGL workstation card now

Oh crap, I just realised my mistake. Sorry dude, I wasnt talking initially about a Nvidia Quadro, I was talking about a Kentsfield Quadro I got those % from THG in their Core 2 Quadro benchmarks, where the performance increases are what they say in that article. I feel like a numb nuts. Should have clarified the first and last Quadro reference.

I KNOW a Quadro video card will never give me that type of a performance boost!!!!

I actually did a lot of research on those cards, and what youre saying is more or less what I had concluded, thats why I got a X1900XT 256MB instead of a Quadro 560. I realised that those cards are for the exact same purpose you mentioned and I realised that my 3D work isnt going to get that intensive just yet. Thanks alot, you helped reassure myself of my decision, and I really appreciate that. Do you think the card Im getting will be sufficient, or should I got with the 512MB version?

... I would also add that you can purchase 4 by 1gb sticks of faster and cheaper ram for roughly 500 dollars or less. ...you can always EBAY...

... where are these 2gb sticks gentrinity is talking about?

Id rather keep my upgrade path inline, it would still be a more expensive proposition to get 4GBs and then sell them to get 8GBs, remember I only have 4 slots to work with so I need to do this right.

http://www.mushkin.com/doc/products/memory_detail.asp?id=514

Here you go noob

Aeneon and A-data also carry 2GB sticks, but one is sold out(havent checked back, I prefer Mushkin since they offer awesome service, I mean, I have their Tech guy on MSN Messenger, thats sweet) and the other was impossible to find in North America.

You can run your ram at the rated speed...

Sorry dude, those benchmarks dont help, theyre pretty much what everyone else has on their benchmark articles, I was looking more for real world benchmarks, with Apps that involve Content Creation like Adobe products, and I appreciate you posting them, but to be perfectly honest, I dont know what to make of em

 

[Mondoman]

...it would still be a more expensive proposition to get 4GBs and then sell them to get 8GBs, remember I only have 4 slots to work with so I need to do this right....

While I'm not psychic, I think this is likely to be wrong. 1GB sticks are now mainstream, at a cost of roughly $100/GB. 2GB sticks seem to cost $200/GB or more. 2GB sticks should be *much* more widely available in 6 months' time, at a substantially reduced price premium over 1GB sticks. Conversely, used name-brand modules in popular sizes don't seem to loose too much value when sold on eBay.
If we assume that the 2GB stick premium will drop from over 100% to 50%, one 2GB stick at that time will cost 150% as much as two 1GB sticks. Thus, 8GB as 2GB sticks will cost the same price as 12 new 1GB sticks (8GB x 150% price ratio).

So, let's look at the two options:

1) Buy 4 GB now, sell later and buy 8GB.
a) buy 4x1GB sticks today (cost: 4 current 1GB sticks)
b) sell later; if used sticks are 25% cheaper than new at that time (similar to or worse than today), will recoup 4x 0.75 = cost of 3 future 1GB sticks.
c) buy 4x2GB sticks later (cost: 12 future 1GB sticks -- see above)
Net cost: cost of 4 current 1GB sticks + 9 future 1GB sticks.

2) Buy 8GB now.
a) buy 4x 2GB sticks today. Given cost premium of 100%, price is 8GB x 2x cost/GB = 16 current 1Gb sticks.
Net cost: cost of 16 current 1GB sticks.

Thus, buying the 8GB now is only cheaper if you assume BOTH that the price premium for 2GB sticks over 1GB won't drop significantly AND that the overall price of RAM will not drop significantly. RAM prices have recently peaked and have already started to decline a bit. A drop in the price premium of 2GB vs 1GB is just a matter of the normal development of memory production.

 

[gentrinity]

Well, while that comparison is fair, we would have to check back when I finally buy them, which would be in about a month or two, so in other words, I might luck out, and RAM prices will be fairly good. But yeah, the Mushkin RAM i wish to buy is currently selling for $130 per 1GB. Its ddr2-533.

I dont know, lets see what happens. But what about performance?

 

[Mondoman]

...But yeah, the Mushkin RAM i wish to buy is currently selling for $130 per 1GB. ...

That's a good price! Looks like the price premium is dropping already.

... But what about performance?

Asked and answered. Ball's in your court now.

 

[gentrinity]

Sadly it was even cheaper, about $5 per stick cheaper a week or so ago.\

About the performance, what I meant was that I still havent found the benchmarks im looking for, which are basically content creation apps, and more importantly, adobe apps when increasing frequency, not size.

 

[Mondoman]

Exactly -- folks here don't seem to know of any such benchmarks.

 

[gentrinity]

Exactly -- folks here don't seem to know of any such benchmarks.

 

[bliq]

It seems to me that at those price levels, it doesn' t make a lot of sense to save 5-6% to lose half the capacity. If you're strapped for cash, obviously, the 4GB. Since this is for a business, and if $80 won't break you or if it means saving another week or two or throwing it on the credit card, if the extra RAM is justified because it increases your efficiency, then it's $80 well spent.

In fact, even if you're strapped for cash, if you can throw it on a credit card and you can quantifiably justify the need, I'd say do it, as long as it's for a business (meaning, there is a return on investment) as opposed to just for bragging rights or gaming(meaning, no return on investment).

 

[gentrinity]

Yeah, thats more or less what ive been thinking. I am more than sure that those 8GBs will come in handy when I really start to kick things in gear. First I plan on going through a learning curve.

 

[elbert]

Thats true but isnt there a way to have the best of both worlds? An OC memory that can both clock to 1067 and underclock when he wants to go beyond 4GB's.

 

[gentrinity]

Thats true but isnt there a way to have the best of both worlds? An OC memory that can both clock to 1067 and underclock when he wants to go beyond 4GB's.

uhmmm.... no.

Socket 775 doesnt have more than 4 slots, so I need to get 2GB sticks if I want to use Core 2 Duos. The only available 2GB sticks availableare Mushkin ddr2-533