is not running a divider better?

[jonesmaster]

Im going crazy trying to decide on what ram to get for my new x2 4200.

Ive pretty much settled on 2 sets (2x1gb). OCZ platinum
http://www.newegg.com/Product/Product.asp?Item=N82E16820227210

or Crucial value.
http://www.newegg.com/Product/Product.asp?Item=N82E16820146545

OCZ is not very overclockable, maybe only to 220. But the timings are tight 2-3-2-5. THe crucial goes to 250 @ 3-3-3-8.
Ive read that X2's like bandwidth, ive also read they like tight timings. Im not trying to max out the overclock, only going to 2.5 or so. Is it better to get the OCZ ram on a divider or the crucial and possibly go 1:1?

On A8N SLI-deluxe MB.

 

[imported_Tick]

Crucial if you want to OC.

 

[woodchuck69]

A divider won't matter. I am running the OCZ Platinum @ 166 divider and it's doing 204mhz while my 3000+ is 2250mhz ... so 450mhz over stock and I still have tight ram timings which is mostly what the A64 architecture cares about. The system is rock solid and cool at this speed.

 

[w00t]

links are broken.

running a divider is not bad chances are your wont even tell a difference.

 

[w00t]

Originally posted by: woodchuck69
A divider won't matter. I am running the OCZ Platinum @ 166 divider and it's doing 204mhz while my 3000+ is 2250mhz ... so 450mhz over stock and I still have tight ram timings which is mostly what the A64 architecture cares about. The system is rock solid and cool at this speed.

"I still have tight ram timings which is mostly what the A64 architecture cares about."

wrong.

 

[Anubis]

Originally posted by: woodchuck69
A divider won't matter. I am running the OCZ Platinum @ 166 divider and it's doing 204mhz while my 3000+ is 2250mhz ... so 450mhz over stock and I still have tight ram timings which is mostly what the A64 architecture cares about. The system is rock solid and cool at this speed.

ummm the timeings dont really mean sh!t on A64s

 

[monster64]

Consider this: In everest running at 3-3-3-6 at 245 MHz I get lower latency/higher bandwidth than running at 2-2-2-5 @ 200 MHz.

 

[n7]

Originally posted by: Anubis

Originally posted by: woodchuck69
A divider won't matter. I am running the OCZ Platinum @ 166 divider and it's doing 204mhz while my 3000+ is 2250mhz ... so 450mhz over stock and I still have tight ram timings which is mostly what the A64 architecture cares about. The system is rock solid and cool at this speed.

ummm the timeings dont really mean sh!t on A64s

Meh, not entirely true.

While both bandwidth & timings aren't everything for A64s, it does matter.

Mushkin 2GB XP3700 3-3-3-8 Dual Pack (2x1GB)

There's a damn nice kit that would allow you to run "1:1" even with some OCing.

Pricey though...

 

[woodchuck69]

Originally posted by: w00t

Originally posted by: woodchuck69
A divider won't matter. I am running the OCZ Platinum @ 166 divider and it's doing 204mhz while my 3000+ is 2250mhz ... so 450mhz over stock and I still have tight ram timings which is mostly what the A64 architecture cares about. The system is rock solid and cool at this speed.

"I still have tight ram timings which is mostly what the A64 architecture cares about."

wrong.

Compared to the overbloated bandwidth the P4 thrives off of.

right

 

[Hacp]

http://www.xbitlabs.com/articles/memory/display/ocz-ddr600_4.html

The divider gives less than .5% less performance. Almost 0!. Also timings mean alot for the A64 platform. Anubis, would you say that an overclock of up to 200MHZ means poop for A64s? Because thats around the difference 2-2-2-5 ram makes over 2.5-3-3-X ram, especially in games.

Ya btw, go with the tight timings.

 

[jonesmaster]

thanks for that link hacp, looks like ill be going with the platinum since I cant afford pc4000+ or crucial ballistics.

SOrry about the broken links, theyre fixed now.

 

[Unkno]

since there isn't much performance difference in low timings or running "1:1", why not just save MORE money by buying a cheaper ram with loose timings that can't overclock well...

 

[rise]

Originally posted by: Unkno
since there isn't much performance difference in low timings or running "1:1", why not just save MORE money buy going to a cheaper ram with loose timings and can't overclock well...

if you want the extra performance and you're not spending money that would be better used on other components, get good ram.

i'm not sure where the ".5% almost 0 difference" is coming from but whatever. certainly not in gaming. run some benches yourself on the games or apps you use most and see whats worthwhile to you.

 

[Suspicious-Teach8788]

Ok, remember, when we say dividers, we're not saying DDR500 vs DDR433 or something like that cuz someone had to use dividers, it's more like same CPU speed but with mem dividers and same MEM speed.

i.e. 3700+ with 10x or 11x multi, and the 11x uses 1:1 while 10x uses a divider and still the same mem and cpu speed.

the net result is THE SAME.

On AMD there is no difference because an A64 uses hypertransport and mem speed is a divider anyways. It's CPU speed / mem divider = mem speed. You are on a mem speed even if it's "1:1" or "9:10" because 1:1 is just so us old skoolers relate mem speed to HTT when in reality mem speed is determined by CPU speed. Thus there is no waiting around like in FSB.

Now, for Intel, there is a difference because if your memory runs slower than your actual FSB, your north bridge waits for data from your memory while the CPU is ready to process stuff already (200 FSB and 166 MHz ram for example). Your CPU is ready to receive/send data, but your memory data is still fetching/writing. FSB/NB traditional architecture has this problem.

In brief, A64 = no difference, Intel = yes.

There was a long discussion on this on XtremeSystems because many people with A64s are still m ore familiar with Intel architecture, and so we tend to think dividers are bad, but on A64s.. no difference.

 

[Suspicious-Teach8788]

As for the timings vs clock speed debate, there was a benchmark between DDR600 and DDR500 2-2-2 on OCForums (TCCD vs UTT memory), and I believe DDR600 wins out SLIGHTLY. Oh and this was done on an A64 system.

I for one believe that I will never hit 300 HTT anyways and even if I lowered my multiplier to 9x, I dont like 9 x 300 (plus that's not too great of a 24/7 speed for me), so I went with UTT ram because I won't be hitting DDR600 with TCCD anyways, and 250 2-2-2 works great for my CPU as well.

 

[rise]

Originally posted by: DLeRium
As for the timings vs clock speed debate, there was a benchmark between DDR600 and DDR500 2-2-2 on OCForums (TCCD vs UTT memory), and I believe DDR600 wins out SLIGHTLY. Oh and this was done on an A64 system.

I for one believe that I will never hit 300 HTT anyways and even if I lowered my multiplier to 9x, I dont like 9 x 300 (plus that's not too great of a 24/7 speed for me), so I went with UTT ram because I won't be hitting DDR600 with TCCD anyways, and 250 2-2-2 works great for my CPU as well.

you should just delete your first post as i have no idea what you're trying to say

obviously 10x220 with 200mhz memory
and 11x200 with 200mhz memory will be almost exactly the same.

the "divider" most people talk about is going from
10x220 with 220 memory vs
10x220 with 200 mhz memory

as for your ocforum thread, needless to say the tccd wasn't running 2-2-2 @ ddr600. so more likely it was a comparison between ddr500 2-2-2 and ddr600 2.5-3-3.

but again, most people run a divider to keep there ram ~200mhz. do you think you'd see a difference between your ddr500 cl2 and ddr400 cl3

 

[Anubis]

Originally posted by: rise4310

Originally posted by: DLeRium


but again, most people run a divider to keep there ram ~200mhz. do you think you'd see a difference between your ddr500 cl2 and ddr400 cl3

not in anything other then a benchmark

 

[MobiusPizza]

There are NO performance difference using dividers except for the marginally more bandwidth proveded by higher frequency memory
However as AMD CPUs prefer faster timing to bandwidth, you better use dividers to allow RAM to run 2-2-2 timings instead

 

[Hacp]

Ok what Delerium is trying to say is that if you have a Venice that is 3200+, and use the divider to get it to 2.4GHZ using the 5:6 ram divider, It will run ALMOST exactly like a 3800+ not using the divider. Thats what were trying to say by saying there is Almost (like .05% if you calculate the difference using the article I linked to above) no performance difference from using the Ram divider.

Ow ya, the article also shows that DDR600 with somewhat loose timings losese to DDR400 with tight timings, although its VERY close. Seeing that its cheaper to get DDR400 2-2-2-5 and use the ram divider, than getting Samsung TCCD and having the 2.5-3-3-7 timings at DDR600 to beat 2-2-2-5 DDR400, its hard to justify going with anything but cheap tight timing ram.

Finally, do the math. I'ma link to an Xbit labs article, which shows the FX-57 vs the FX55.
http://www.xbitlabs.com/articles/cpu/display/athlon64-fx57_9.html
Calculate the % difference the FX-57 has over the FX55 in games, and things like Mpeg-4 encoding and Data compression.

Now I'm going to link to an article that tests DDR600 ram.
http://www.xbitlabs.com/articles/cpu/display/athlon64-fx57_10.html
It also shows the ram performing with various timings at DDR400. Look at the 3rd slot on the graph, where it says FSB=200 DDR=400. Calculate the differences in games and Mpeg-4 and Data compression between 2-2-2-5 ram and 2.5-3-3-6 ram. In the games and Data compression, the % difference between Tight timings and loose timings is GREATer than the difference between the FX57 and FX55!!!! Of course, in Mpeg-4 encoding, the FX57 % difference over the FX55 wins by a large margin.

So, in conclusion, if you want performance and already have things like the latest video card, spending the extra 20-25 dollars on performance ram over Corsair Value Select is WELL worth it.


FInally, to the OP, get the OCZ plats. The crucial are NOT using the same chips as the ballistix anymore. Hardforums has confirmed this, as have other forums.

 

[rise]

you link to the same article twice. i dunno why you're comparing the difference between an fx57 and fx55 but regardless your final conclusion is what i said before i just don't agree with the .5%-0 difference you stated.

if you want the extra performance and you're not spending money that would be better used on other components, get good ram.

you just seem to take a round about way of saying it

Originally posted by: Anubis

Originally posted by: rise4310

Originally posted by: DLeRium


but again, most people run a divider to keep there ram ~200mhz. do you think you'd see a difference between your ddr500 cl2 and ddr400 cl3

not in anything other then a benchmark

or FPS but whichever way you want to look at it.

 

[Hacp]

Originally posted by: rise4310
you link to the same article twice. i dunno why you're comparing the difference between an fx57 and fx55 but regardless your final conclusion is what i said before i just don't agree with the .5%-0 difference you stated.

if you want the extra performance and you're not spending money that would be better used on other components, get good ram.

you just seem to take a round about way of saying it

Originally posted by: Anubis

Originally posted by: rise4310

Originally posted by: DLeRium


but again, most people run a divider to keep there ram ~200mhz. do you think you'd see a difference between your ddr500 cl2 and ddr400 cl3

not in anything other then a benchmark

or FPS but whichever way you want to look at it.

Look at the first article I linked to. DDR400 with HTT at 300 compared to DDR400 with HTT at 200. Thats running the ram divider. CALCULATE the percentages by hand if you want.

 

[rise]

Originally posted by: Hacp
Look at the first article I linked to. DDR400 with HTT at 300 compared to DDR400 with HTT at 200. Thats running the ram divider. CALCULATE the percentages by hand if you want.

you mean this article?

ok, lets see... we'll go with -
300htt and ddr600 2.5-4-4-10
300htt and ddr400 2.5-4-4-8
as those are the nearest timings shown between running 1:1 and a divider, ok? same cpu speed just running the ram slower...

doom3
ddr600 110 fps
ddr400 104 fps

farcry
ddr600 191 fps
ddr400 169 fps

hl2
ddr600 86 fps
ddr400 77 fps

q3
ddr600 500fps
ddr400 451 fps

ut04
ddr600 188
ddr400 170

 

[Hacp]

Rise your looking at the WRONG numbers. Can you read? HTT at 300 DDR400 VS HTT at 200 DDR 400. NOT HTT at 300 DDR 600! Now type all that out again using what I said again and you will see that Using the divider gives almost NO performance hit!

 

[rise]

dude are you out of your mind?

we're talking ram speed, or at least i am. wth would i compare ddr400 vs ddr400?

the rest of the free world refers to a divider as the relation between ram speed and htt/fsb- raise the htt and you raise the ram speed, unless you use a divider. understand?

as i said before,
10x220 and 11x200 won't mean a thing if ram speed is kept at ddr400.

 

[fluxquantum]

hello everyone,

since you guys are on the topic of dividers, i was wondering which is better in my situation.
i've played a little bit with my settings and this is what i was able to achieve so far.

180 MHz (DDR360) using a 266 divider with CAS at 2-3-2-5 1T at default voltage (2.6 vdimm)

225 MHz (DDR450) using a 333 divider with CAS at 3-4-4-10 1T at voltage 2.7 vdimm
the rest of the information is in my sig.