Why stress out about timings?

[ap673]

Why is it that everyone will be willing to pay whatever and do watever to get some ram that can do 220-250 with tight timings, like 2-2-2-5. I'm sure you've all seen benchmarks, tight timings don't do much in games and normal stuff. I'm sure it helps in video encoding and professional machines, but why would a gaming rig need 2-2-2 timings? Higher FSB and HTT make more of a difference that extremely low timings. So why not do 2.5-3-3-11 or something at 250mhz, you won't see ur FPS or anything drop. And why are there such things as PC4000 and PC4200 when the base mhz in everything is still 200mhz or PC3200. Are they soley for the purpose of overclocking? By the way all of this is more of a question than an arguement.

 

[Gautam]

This is one reason why gamers may like tight timings:

Identical everything, 24k with 7-2-2-2, 23.3k with 7-3-3-2.

You won't notice a lot of things. But, like that last 100MHz, having tight timings just feels good to have, and definitely gives an edge in just about every benchie you can throw at it. It's effects in SuperPI and PIFast are quite pronounced. But for everyday usage, I agree, the timings aren't really a big deal in the grand scheme of things.

Higher FSB though, is becoming rather meaningless. Games definitely do not require the sort of bandwidth that PC4000+ offers. But they always benefit from tight latencies.

And yes, I think PC3200 is the highest that's JEDEC approved, so everything else is techincally overclocked, and for overclocking only.

 

[kiyoshilionz]

I heard someone say something like going from 3-4-4-8 timings to 2-2-2-5 gave him the same increase in performance as going from like 200fsb to 220. But I really doubt that's possible, it's just a small adjustment that may give a tiny little speed bump in synthetic stuff. And as overclockers, we want all the speed possible.

 

[pelikan]

On my system the UT2003 benchmark shows going from 2.5-3-3-x to 2-2-2-x equals 200 cpu MHz.

 

[Gautam]

I heard someone say something like going from 3-4-4-8 timings to 2-2-2-5 gave him the same increase in performance as going from like 200fsb to 220. But I really doubt that's possible, it's just a small adjustment that may give a tiny little speed bump in synthetic stuff. And as overclockers, we want all the speed possible.

The difference is more like from 200MHz to 240ish, if you take into account most benchmarks. As overclockers, we want all the performance possible, at least I do. Don't get speed confused with performance.

I was bored so I did some testing allbeit limited.
A64 3200+ run at 2407 in all test. Ram is single stick of KHX3000 with 2-2-2-5 timmings @ 200 (CPU/12 divider as listed by CPU-Z)and 2.5-4-4-8 timmings @240 ( I know I could do better timmings here but these are average crap timmings for PC4000 ram)

Hexus Pifast
1:1 52.05 seconds
CPU/12 52.14 seconds

SuperPI 8meg
1:1 7 min 50 sec.
CPU/12 7 min 38 sec.

SuperPI 1meg
1:1 37 sec
CPU/12 37 sec

200 with 2-2-2 either ties or beats down 240 4-4-2.5 in all situations.

 

[ap673]

Wow that is pretty amazing. But still, for games, you won't be missing out on much.

 

[Hardin]

Then why did everyone complain about RDRAM? Was it because it was expensive or because of the latencies, or both?

 

[glasszon]

mainly because of the price, it is MUCH more expensive compared to ddr.

 

[Busty St. Clair]

isn't RD-Ram trying to make a comeback on SIS chipsets for intel. i read an article on that about a month ago i believe.

 

[ap673]

Ya, quadchannel rdram, sounds gay.

 

[BeerHunter]

I hear LL is even more important to A64? Any truth to that and why?

Personally I got smoking deal on BH-5 so I thought, what the hey, it's only $25 more per stick than cheap value ram and all these guys talk about it so it must be good.

 

[Phrenetical]

Just on tight timings for video encoding, defenently helped me out to move form 2.5-4-4-8 to 2.5-3-3-7

3.55 @ 2.5-4-4-8 encoding mpeg movie to DivX i was getting 110 - 130 fps (depends which pass its doing) thats pretty good

but at

3.55 @ 2.5-3-3-7 encoding same mpeg movie to DivX i was getting 130 - 160 fps (depends which pass its doing) thats dam good

i can now encode mpegs and other media to real DivX's at the rate of 24 minutes real time = 1hr of movie encoding into divX with two passes for quality

i didnt even believe what i was seeing, but its true

ps.. this was done using DR.DivX 1.0.4 full version