• We should now be fully online following an overnight outage. Apologies for any inconvenience, we do not expect there to be any further issues.

My Gaming Rig

Page 2 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

RussianSensation

Elite Member
Sep 5, 2003
19,458
765
126
Originally posted by: Acanthus
Again i must reiterate, although memory speed does make a difference, timings do not.

At no point without benchmarks can you take 2 identical PCs side by side, one at 2-2-2-10, the other at 3-4-4-15, and say HOLY CRAP LOOK AT THE DIFFERENCE! As a matter of fact, i bet you couldnt even tell which was which, i know i couldnt. (from experience in this exact situation)

Good point since the performance difference in timings is hardly more than 5%. Real world difference must exceed 10% for the user to "feel" it. At the same time I can't disregard that often running 250FSB : 200 memory (5:4 ratio) @ 2-2-2-5 is actually faster than running 250FSB:250 memory (1:1 ratio) @ 3-4-4-8. Of course since OCZ rev 2 cant run 250@ 2-2-2-5, spending money on ocz and running it 1:1 at looser timings is going to alleviate any advantages gained by higher memory speed. So if a user reverts to running 200@ 2-2-2-5, why buy super expensive ram when you can get cheaper PC3200 2-2-2-5.

Here are some examples of how tight timings can benefit (often vs. higher speed ram):

Review 1 - XP systems
Review 2 - P4 systems 1
Review 3 - P4 systems 2
Review 4 - A64 systems

Now for A64 memory timings matter less since it has on-die memory controller but they still matter. Since no ram except for BH-5 @ 3.3+V can run 250 @ 2-2-2-5 or even 2-3-3-6, I dont ever see the point of people buying the fastest ram, since all you need is PC3200 or maybe PC3500 @ 2-2-2-5 and you'll be just as good as the person running their memory at 270mhz x 2 @ lose timings. But you save loads in the process. Yes, you are right, timings don't matter so much, but then if you want to be fair, higher memory speed doenst matter either since its small advantage performance is reduced due to worse timings.
 

Zinn2b

Banned
Jan 9, 2004
361
0
0
IThere was a time when it really didn't matter much with AMD what ram u used but if you buy a 939 socket the memory timings do matter. I would urge you to look real hard at Crucial performance ram It runs great with AMD AND you can O/C the hell out of it. This memory runs @ 2-5-2-2 up to 225 thats PC 500 memory @ 2-5-2-2 and its the cheapest high performance ram its rated @ PC3200 great ram 4 AMD"s
 

stelleg151

Senior member
Sep 2, 2004
822
0
0
I would say deeeffffinately get the Corsair XMS if you can get it for 190 american, that is a sweet deal. The difference between that stuff and the OCZ Platinum revision 2 is DEFINATELY NOT WORTH 80-90 DOLLARS. Goodness, just because Anandtech says its the best doesnt mean all other kinds arent any good. All of the TCCD Ram is sweet, and with the 3500+ you wont need super high FSB, because of the 11x multiplier.
 

Ressurection

Junior Member
Nov 14, 2004
12
0
0
Originally posted by: stelleg151
I would say deeeffffinately get the Corsair XMS if you can get it for 190 american, that is a sweet deal. The difference between that stuff and the OCZ Platinum revision 2 is DEFINATELY NOT WORTH 80-90 DOLLARS. Goodness, just because Anandtech says its the best doesnt mean all other kinds arent any good. All of the TCCD Ram is sweet, and with the 3500+ you wont need super high FSB, because of the 11x multiplier.

Thx for the advice. ;)
 

Acanthus

Lifer
Aug 28, 2001
19,915
2
76
ostif.org
The only results linked in your 4 examples that make a large difference are in artificial benchmarks.

I dont consider 3% to be a wide margain.

In the case of the 2.4C example, the CPU doesnt use the additional memory bandwidth given to it, timings make a larger difference there because the clockspeed of the memory basically does nothing. The DDR500 results @ 3-4-4-8 might as well have been DDR1000 3-4-4-8, the cpu simply wont use the bandwidth @ 3.0ghz. At 3.7ghz+ you barely start to see increases with mem bandwidth and fsb, the cpu simply doesnt use it.

What im saying essentially is, if you skimp on memory timings and get a better cpu/vid card/cooling system you will get better results.

If youre spending a rediculous amount of cash, then get it, and be happy with your 2%.

I am building a 3.2E @ 3.7-3.9ghz system soon, and will more than likely post some benchies to support what i am saying (about ddr clockspeed not effecting the pentium 4 @ 3.0ghz, dual channel ddr400 is more than enough).
 

caz67

Golden Member
Jan 4, 2004
1,369
0
0
I would forget about the Corsair. OCZ Plat Rev2 or Geil Ultra X, these are the most overclockable chips at the moment. These chips are almost universal, as the achieve DDR500 speeds and beyond with 2.2.2.5 timings.
 

Gothgar

Lifer
Sep 1, 2004
13,429
1
0
look at all these OCZ RAM fanboi's... far as the eye can see..

guess what people, the Corsair that he wants is the Samsung TCCD memory modules...

It is basically the exact same memory in both RAMs, the Corsair will do whatever the OCZ will do for a lot less...
 

Zinn2b

Banned
Jan 9, 2004
361
0
0
Got to love Acanthus here is gentleman that does understand things pretty well! If it were not for one thing I would have agreed 100%.IN my personal system I presently run Crucial ballistic @ 2-5-2-2 and you are absolutly correct to a degree. But I also have About 10 matched pairs of the CorsairXMSLL BH-5 sticks and I could run them with Pat enable AT F1 @ 210 and its actually about 8% now I use the crucial most the time because thats what I the put in customer machines. The most I ever got the Corsair to O/C was 216 at pat in auto now the crucial well run at 219 Pat is auto and everthing is stable . I couldn,t run V above 2.8 on my abit card because V. was all over the place. After installing PC Power & cooling 500 watt PSU ram voltage was stable @ 3.04 I was amazed. So I decided to check things out I installed the Corsair BH-5 put V. @2.94 and was able to get 224 pat at auto. but heat was a problem so I backed Off to 222and ran it there For a weak not a burp AND it was 8% differance. And my PC mark scores show it.I want to install one off my watercooling systems but I can't keep them on the Shelf soon I hope though. Dam machine shop is slow. I know or at least I think I do that withV.3.04 I believe I can hit 230 @ 2-5-2-2 pat at auto with watercooling thats around 3.70 GHz and that should be one asskicking Intel. Thought about 3.4c that would put me at 4+GHz @ 2-5-2-2 pat @ auto. The problem is this little 3.2c is an asskicker in it own right. and smooth its like silk. Nice Job Acanthus but your theory does not apply to BH-5 ram it is an 8% gain @ 2-5-2-2. It is also gone forever. WHEN I say 8% differance thats compared to the Crucial at same settings it is truely special ram. ITS one of the things me an Anad agree on!!!!!!
 

Ressurection

Junior Member
Nov 14, 2004
12
0
0
I have been thinking about the HD's for a while now... and I came up with this solution. How about one 36GB Raptor for Windows while the Seagate 120GB (7200.7 RPM) is for normal storage? Would that be good?

Also, what's the advantage of having Windows installed on a Raptor? Would booting up take lesser time?


Thanks,
Ress
 

Interitus

Platinum Member
Jan 28, 2004
2,143
9
81
I think your setup looks good. Only thing I would change is the motherboard.

If you're going to wait until January anyways, I'd go with an Nforce 4 motherboard from Asus or possibly DFI if they release a non-SLI version.

As for the RAM, stick with your Corsair. You wanna spend 2x the cash for 200 points in a benchmark, go ahead, but that Corsair will do you just fine.

The Raptor would be nice, but they're noisy as hell and personally I'd rather have more space than 2 less seconds loading time.
 

RussianSensation

Elite Member
Sep 5, 2003
19,458
765
126
Originally posted by: Acanthus
PC mark scores are artifical benchmarks.

They do not reflect actual performance in real applications.

You and I are essentially saying the same thing that timings do not really matter. What i am saying however, is that not only is this true, but what's worse is that higher FSB with looser timings is even worse than lower RAM speed with tighter timings. So to summarize, you are saying that ram timings do very little (which is true for the most part) and I am saying that ram speed at high timings doesnt matter either. That's why I'd just get 1 gig for $170 and put the rest towards a faster cpu or gpu. But if he wants a balance, then buying PC3200 that gives 2-2-2-5 is better than running memory at 500mhz with 3-4-4-8 and costs less.