Do I benefit from expensive RAM if I only run at Stock speeds

imported_MaxPayne

Junior Member
Jan 19, 2005
4
0
0
I never ever overclock my computer. I always run it at stock speed.

Since I only run it at stock, is there any advantage to me getting better RAM?

For example, is there a benefit to buying PC3200 over PC2700 if I only run at stock speed?
Or do I really need a low latency RAM if I don't overclock??

 

miketheidiot

Lifer
Sep 3, 2004
11,060
1
0
more bandwidth. almost all modren desktop processors run at 400mhz (3200) by default. Second cheap pc2700 isn't much cheaper than cheap pc3200, if it is at all.
 

Tomahawk513

Junior Member
Jan 9, 2005
20
0
0
well it depends really... if you only run things at stock settings, probably not, unless you game... if you play HL2 or UT 2K4 or other recent games like that, it would make a big difference, but if all you're using your computer for is Wolfenstein 3D and Word Processing (Or other business related activities) it wouldn't really make a difference
 

BonzaiDuck

Lifer
Jun 30, 2004
16,041
1,679
126
If the motherboard supports 800 Mhz front-side bus, it does not make sense to run PC2700 or DDR333 memory with a processor that supports 800-FSB. That is, it works fine, but forces the memory to run at DDR320.

My philosophy is this -- whether gaming, vid-processing, running spread-sheets and database management -- you should exploit the reasonable limits of stock specifications for all components. Don't get a 2.53 "B" processor for a motherboard compatible with 800 Mhz FSB, when you can get a 2.4 "C" processor. Don't get DDR333 memory for the same mobo when you can get DDR400 -- provided you didn't already limit yourself to the 2.53 "B" CPU. Barring that, try to reduce the number of compromises so that future enhancements only cost you the replacement of one item.

DDR400 or PC3200 (same) is industry-standard right now. Further, you can increase performance without over-clocking the FSB by looking for lower-latency RAM, even if it costs a bit more. At that point, it all depends on whether you want the best performance without over-clocking, or just reasonable performance without over-clocking.

I have a friend in the Bay Area who bought a custom-configured rig from a local independent computer shop. She does art-work, but has yet to explore the limits of Corel and other graphics programs for her pastime. They set her up with an Intel D865PERL mobo and a 2.6C processor (so far, so good), but only socketed a single 256 MB module of DDR400 memory -- therefore operating in single-channel mode. If she thought she could live with 256MB, my choice would have been a pair of 128's. But if she didn't "know", they didn't "care". In fact, if she thought she could "live" with 256MB, I would have cajoled her into gettng 512 MB as 256x2.

If you buy average-latency DDR400's now, and six-months from now you decide you want to over-clock the system, how will you feel about it then?
 

hundesau

Member
Dec 25, 2004
157
0
0
dual channel is a hoax, speed improvements are 3% maiximum.

timings are a hoax, differences between high and low timings are maximum about 5%.

DDR400 is standard and not pricier than 333Mhz Ram, so why buy the old stuff. Bandwith improvement depends on what kind of Frontside Bus ur CPU has and if it works good or bad when running asynchonous to the Ram Speed of 400MHz or 333Mhz.

Athlon XP always had the weakness that it wouldnt win much out of overclokcing as long as u run it asynchron, eg having it run with 200MHZ frontside bus and 333MHZ Ram. U needed 400MHz ram so that ur overclokcing would really have performance gains, On the other hand Intel never had that much of a problem iwth asynchron clock speeds. So better Ram (400Mhz) always gives u more space regarding overlocking an synchronous clock speeds. Tahts basically all. The higher u want to overclock the higher speeds u need for ur Ram. The timings as i said dont matter much, u pay for maximum 5% when u buy low latency ram, not worth the expensive Rams imo.

If u run ur CPU on clock speed than all u need is Ram that fits to the FSb of ur CPU. If its running at 166Mhz like Barton u need 33Mhz Ram, if u run a P4 533Mhz FSB(quad pumped, eg 4x133) u can go with 266Mhz Ram. If u run P4 800Mhz FSB guess? Yep, 400MHz works best. Any additional Speed will give more Memory badnwith and CPU mite slightly can use it, but not much.

Anyways, as 400Mhz Ram isnt much more expensive than the slower Ram, u can buy the fast Ram even if u dont really need it, as its better for the case of overclocking and "futureproof" for later systems or when u want to sell it at ebay.
 

Stormgiant

Senior member
Oct 25, 1999
829
0
0
Originally posted by: hundesau
dual channel is a hoax, speed improvements are 3% maiximum.

timings are a hoax, differences between high and low timings are maximum about 5%.

Do you have numbers to backup what you are saying ? Can you elaborate more on this subject.
I would be very much interested in hearing more from this two statments...
 

hundesau

Member
Dec 25, 2004
157
0
0
Okay, u asked for it :)

1. link concerning Ram timings = http://www.tbreak.com/reviews/...pu&id=333&pagenumber=1


2. Now for that dual channel thing:
the problem with dual channel on Athlon XP was that the CPU bandwith was 2x200MHz FSB (if it was clocked to 200Mhz, like a 2500+@3200+niveau), and the Memory bandwith using 400MHz DDR Ram was 2x200MHz Ram clock in single channel aswell. So it was almost perfect already and the new feature of 4x200MHz memory bandwith was more or less pointless, except for some minor effects, eg that the latencies (which have no big effect on performce as we already know) were a lil lower. And that in the case the full memory bandwith was already used by the CPU there was no extra room for the needs of agp and southbridge bandwith. with dual channel this would work better, so that southbridge and agp would have enuogh bandwith even when CPU was reading/writing memory with full bandwith. These were minor effects but at least something.

With Intel its in theory a lil bit different. Lets do some math in detail:
This example is for a Pentium4 Northwood with 533MHz Fsb, which is able to read/write eg has a CPU bandwith to the Core of:
133 MHz FSB * 64 bit (single channel) * 4 (quadpumped) Datapackets per Clock = 34048 MBit/s bzw. = 4200 MB/s

With dual channel DDR 266MHz Ram the memory bandwith to the Core would be:
133 MHz MemoryClock * 128 bit (Dual Channel) * 2 (DDR) Datapackets per Clock = 34048 Mbit/s = 4200 MB/s
So as u can see the slow PC2100 seemed to be a good choice for that CPU, but it turned out that due to the architecture of DDR Ram the maximum bandwith of the Ram could not be reached and that faster Ram with a higher bandwith would lead to optimal performance.

Finally Intel increased the FSB to 800MHz that offers this CPU bandwith:
200 MHz FSB * 64 bit * 4 (quadpumped) Datapackets per Clock = 51200 MBit/s = 6400 MB/s

Now, there is no other way than to use 400MHz Ram:
200 MHz Memory Clock * 128 bit (Dual Channel) * 2 (DDR) Datapackets per Clock = 51200 MBit/s = 6400 MB/s
Still theres the problem, that faster memory would give slightly better memory benchmark results cuz the Rams maximum bandwith cant be reached, eg its only an perfect Team in theory.

By now u will be sayin: Told ya, this guy is nuts. First he says Dual Channel (DC) is crap and now he proves u need it and u need even higher clocked Ram for best performance.
As these math above in theory is right, it will differ in practice and wont give u much performane due to one thing and i hope u understand the subtext ;):

"3DMARK is cool but i cant play it with my friends!"

When u take a look at the benchmarks in this german review here (sry, didnt find a english one): http://www.hardtecs4u.com/revi...ntel_i875p/index16.php
u will see that there is only one benchmark that benefits much and this is Sandra Memory benchmark. And this benchmark is a synthetic, a theoretical benchmark meassuring what is the possible bandwith and not what is the real bandwith used by an application.

The term dual channel makes people think their memory is workink twice as fast as in single channel mode. but the results in the benchmarks show that its far away from doubling the performance in real life siutations. Dual Channel wont help u increasing ur fps in games, wont make ur video encoding, zipping files or compiling work significantly faster, i mean not that much that u will even recognize it unless using a benchmark.

I dont wanna advise not to buy Dual Channel, it surely gives a lil extra performance and has become a standard in AMD and Intel CHipsets, but the performance gain is rather small. But its better to buy dual channel and think: hey, i doubled performance!, then to waste cash for low latency Ram and think: i got the lowest latency, my comp must be the fastest in town.


Hope this helps ;-) Maybe not, my english sucks, this text must sound like written by a 5 year old, lol, anyways, greetz.


e: here a benchmark showing the effects of dual channel regarding Athlon XP Systems. http://www.tweakpc.de/hardware...rce2_benchmarks2_2.htm
 

maluckey

Platinum Member
Jan 31, 2003
2,933
0
71
That's all fine and dandy,

8 percent (tatal) difference you are talking about can mean the world in a fast-paced FPS game. It can translate to about 3 FPS in a RAM-speed dependant game that averages 60 FPS. Below 60 FPS, games aren't totally fluid. Slower RAM also is a hit for certain games that use the AGP aperture heavily. certain physics events, arrays and designated non-local textures are sent there, and the faster the RAM at any given speed, the less that the game has to wait to render.

If you rely on in-game FPS for readings, you will see that some are faster in dual channel, and faster still with tight RAM timings. This is especially true as your near the threshold of Saturation, which is almost NEVER dealt with in these tests. You will not find an experienced gamer that would willingly run in single channel with loose timings for a few dollars difference (if money is not the object).

If RAM "A" is at 89 percent bandwidth efficiency and RAM "B" is at 96 percent efficiency, there is a solid 7 percent change in speed right there. WIth slower latency comes decreased efficiency at the same FSB. Dual Channel gains are truly minimal, but still is worth it if your system can handle it. Increased bandwidth alone is not a huge change, and losing timing OR bandwidth is not a huge setback it's true, but both sacrificed for a few bucks is less acceptable.



 

hundesau

Member
Dec 25, 2004
157
0
0
how about that: instead of buying features that are expensive and bring only lil performance gains- like low latency ram - one should buy a new CPU or a new graphics card with a good relation between price and performance. Dual Channel is already included in the all new Chipsets, so its a common feature. The myths about LL Ram are still alive and pricy, here in germany u can buy 512mb value Ram for about 80$, low latency ram with tight timings starts at about 140$, and can rise up to 250$ for a 512mb module. so instead of the 5% more performance due to better timings i would rather upgrade my CPU or pay more for my graphics card. that should give higher fps.