Which Route? 1:1 or 5:4?

C

Canon20d

Weaksauce
Joined
Feb 14, 2005
Messages
86
Im just trying to Oc my P4 2.6 these days more than 20% which is proving interesting. And since im sure this has been covered before ill save the backstory
Solution A: Run @ 5:4 with 3.31ghz and 408mhz DDR memory or
Solution B: Run @ 1:1 with 2.99ghz and 460mhz DDR memory
Looks like the Ram tops out around 230 and the GPU around 3.4. Which would give the better game/multiapp performance?
thanks!
 
Most P4 mobos only have 3 settings for cpu:ram ratio, 1:1, 5:4, and 3:2? I think.
 
yeah i only have 3, 1:1, 5:4 and 3:2 but thats not so fun cause i cant get high enough on 3:2 to make it worth while
Another question, what kinda volts should i be at? i tried to do 270 (3.5ghz) w 1.95v, it posts and heads to windows, but it locks as soon as windows starts. i tried some other settings like 1.7v and it wont post
any suggestions on FSB vs V?
 
Go for 5:4. I had to make a decision like this before...

1:1, 3.25ghz 271FSB
5:4, 3.4ghz 284FSB

I chose 5:4
 
ummmmmm a few hundred mhz on the cpu would equal 1ghz of memory :) lol..maybe not..but cpu mhz> ram mhz...... and there's always clockgen :) unless it can't change the multiplier of p4's.. just configure clockgen...follow thier intructions and toss it in the startup folder :)
 
well i was using clockgen before and got 3.4, i just wanted to do it hardware wise cause i figured it would be more stable, i dont know what the difference is, but i can hit 3.4 n/p with it, but cant on the bios..??
 
well, the answer is benchmark both settings and see which is faster.

But from experience I can tell you that the 5:4 ratio with the 300Mhz faster CPU should be much faster, plus you should be able to tighten up your ram timings at a higher FSB which should almost completely offset using the ratio. (lower latency)

Check sisoft for both settings to see. So you get a much faster CPU and ram that benchmarks at probably about the same mb/s.

Personally, I use 3:2 ratio to hit 3.6Ghz with my 2.4C.
http://service.futuremark.com/compare?2k1=8061769
 
damn i cant even hit 3.4 :(
3.38 is about as fast as i can get - must need to fiddle more
 
3.38 is good.

I've been running at 3.4Ghz lately just b/c I can slow down the fans in my case, and have things run cooler & quieter, etc...

actually, I'm only dual prime95 stable up to 3.45Ghz.
But can run / benchmark ANYTHING else up to 3.6Ghz
and can run simple benchies/ screenshots, etc... at about 3.7Ghz
http://www.crowdcontrolusa.com/overclock/shizzy.html <-- pics & details
 
what kind of voltages are you running, mine seems to like 1.75 @ 3.38, but if i go higher, like 3.4xx it wont get to windows until i get to 1.85-1.9 but then it only hangs in windows, but if i run even Higher 3.45x it wont post unless its like 1.800? I dunno i think theres just to many settings for me to try, i need a someone elses baseline ya know
 
If you game, you want 1:1 for synchronous memory access, 5:4 (or any divider) is shit in that aspect.

However, if you want it for foolish little synthetic benchmarks like Sandra or worthless ones like SuperPI, then running a 5:4 divider would be ideal since it isolates the CPU and you can get the most clockspeed since there would be a motherboard limitation before memory limitation (in most cases)
 
AH hA! the truth comes out!
Yea, i want it for games and multi-apps, i never whetstone, i dont know what that is :p
i just want a 3.4ghz but since that doesnt look likey, i might be able to settle with a 3055, stupid memory holding me back from 250..
If i got Faster memory, like 533, could i then set the FSB to 1:1 and GET 250fsb? (with a little OC?)
 
ScHpAnKy said:
If you game, you want 1:1 for synchronous memory access, 5:4 (or any divider) is shit in that aspect.

However, if you want it for foolish little synthetic benchmarks like Sandra or worthless ones like SuperPI, then running a 5:4 divider would be ideal since it isolates the CPU and you can get the most clockspeed since there would be a motherboard limitation before memory limitation (in most cases)
Click to expand...

ScHpAnKy is a fool and talking out of his ass.

Obviously 1:1 is better than 5:4 and thats better than 3:2. However, if you can get a more CPU speed and/or use tighter timings it almost always offsets the SMALL hit you take by dropping your ram ratio to a slow ram speed. (other variables, besides his dumb blanket statements)

In your case your max FSB / CPU speed is 260Mhz.... so for the best performance you need some ram that can run DDR520. It just depends on what type of ram you have, and your goal. For me the best case would be some DDR600 which I could run at 1:1, or next best DDR480 at 5:4 ratio, but my ram is only DDR400, and I've got to use 3:2 ratio! B/c I can't get even close to DDR480 or DDR600 with my ram. If I stuck with the 1:1 ratio, my OC would be limited to 2.6Ghz.... theres no way it would be close to matching my 3.6Ghz real world performance or benchmarks.

Benchmark the games you actually play at both settings and see which is faster.

I've seen many benchmarks at 250FSB with 1:1 at loose timings like 3-4-4-8 vs 250FSB with 5:4 at tight timings like 2-2-2-5, and the 5:4 is FASTER IN ALMOST ALL REAL WORLD GAMES & BENCHMARKS. The 1:1 was faster in Sisoft memory bench. (EXACT OPPOSITE OF what ScHpAnKy said)
 
Schpanky is actually quite right, the only time a divider doesn't really matter is on A64s because the memory controller is on-die, the P4 is a bandwith hog that eats up RAM bandwith like no other (dual-channel RAM anyone?). You can only really stretch its legs in real-world tasks if the RAM is 1:1. Do some research before you start trolling and shit.
 
chrisf6969 said:
ScHpAnKy is a fool and talking out of his ass.
Click to expand...
Thank you for insulting me on a public forum, I take it without retaliation and offer you the advice that continuing to post in a derogatory way will end in nothing more than your removal from these forums, so watch what you say in the future, as the rules clearly state that there is to be no flaming.

chrisf6969 said:
Obviously 1:1 is better than 5:4 and thats better than 3:2. However, if you can get a more CPU speed and/or use tighter timings it almost always offsets the SMALL hit you take by dropping your ram ratio to a slow ram speed. (other variables, besides his dumb blanket statements)
Click to expand...
The reality is that it almost never offsets the actually large hit you take by using a divider, take for example... which would be better: 275FSB 1:1 @ 2.5 Cas Latency, or 275FSB 5:4 @ 2 Cas Latency (resulting in about 220FSB on the memory).

Answer: Obviously the first because with a 275FSB in dual channel the memory is read and written that many more times and since it is synchronous with the CPU it can actually perform the work it was assigned in realtime. The only way to successfully overtake a 1:1 ratio is to gain which you have lost, which results to astronomically high numbers... of which most cpus can not attain.

chrisf6969 said:
In your case your max FSB / CPU speed is 260Mhz.... so for the best performance you need some ram that can run DDR520. It just depends on what type of ram you have, and your goal. For me the best case would be some DDR600 which I could run at 1:1, or next best DDR480 at 5:4 ratio, but my ram is only DDR400, and I've got to use 3:2 ratio! B/c I can't get even close to DDR480 or DDR600 with my ram. If I stuck with the 1:1 ratio, my OC would be limited to 2.6Ghz.... theres no way it would be close to matching my 3.6Ghz real world performance or benchmarks.
Click to expand...

This is the evidence you have, with which you benchmark and record the statistics... but since you are stuck with the DDR400 ram, you are of course going to see minimal results with a 1:1 ratio... it only makes sense.

chrisf6969 said:
Benchmark the games you actually play at both settings and see which is faster.

I've seen many benchmarks at 250FSB with 1:1 at loose timings like 3-4-4-8 vs 250FSB with 5:4 at tight timings like 2-2-2-5, and the 5:4 is FASTER IN ALMOST ALL REAL WORLD GAMES & BENCHMARKS. The 1:1 was faster in Sisoft memory bench. (EXACT OPPOSITE OF what ScHpAnKy said)
Click to expand...

Sandra is a synthetic benchmark and results are nothing but numbers without real-world benchmarking. But since it's really bad ram (CL3 @ 250FSB is quite weak) of course you see bad performance. With some real ram, CL2.5 @ 250FSB would undoubtedly slaughter the same setup @ 250FSB with a 5:4 ratio with a cas latency of 2. Also, you are speaking of sandra's memory setup, that's an obvious deduction. Let's try the cpu arithmatic benchmark, as that was what I was speaking about (please forgive me for not specifying that)
 
when i was playing with my 2.4c a long time ago 5:4 with memory at 220 preformed the same as 1:1 did, the only thing it changed was sandra and stuff
 
ryuji said:
when i was playing with my 2.4c a long time ago 5:4 with memory at 220 preformed the same as 1:1 did, the only thing it changed was sandra and stuff
Click to expand...

That was 220 1:1 FSB though on the old ram with a cas latency of 3 since it was DDR333, with the new ram it was a good deal faster.
 
ScHpAnKy said:
That was 220 1:1 FSB though on the old ram with a cas latency of 3 since it was DDR333, with the new ram it was a good deal faster.
Click to expand...
qft
 
So your saying spanky that if i got some new-fangled ram @ ddr 5500 or 6000 to run @ 250FSB then i could run my CPU @ 1:1 with the stock timing and OC the FSB to 250, to keep the ram at its original frequency, and not deal with a 5:4 and get the best of both
Right? :)
i only have some XMS3200 DDR 400 in there now so im limited to ~231 1:1 (even thats flakey in games, but doesnt crash) where as my CPU tops @ 260 @ 5:4
Take 260FSB and put in DDR 533 ram and run cpu @ 1:1
 
you only need ddr 500 for 1:1 250 if you get good tccd ddr500 an bp pcb, you can hit around 300 :D
 
well, isn't worth the money at all...you won't notice ANYTHING @ all.....but your pocket will just be light, that's the only thing you will notice :)
 
aZn_plyR said:
well, isn't worth the money at all...you won't notice ANYTHING @ all.....but your pocket will just be light, that's the only thing you will notice :)
Click to expand...

That was my point. For absolute best performance 1:1 is better, but the ratio would works fine also. Given your ram that you already have, the ratio is the best way to go. AND its NOT that big of a hit on performance on Intel 865/875 systems.

If you want to go buy some new ram... go buy some DDR500, that will probably get you to FSB 260 1:1 to maximize your config. But like I said at 5:4 ratio, you can use your ram at Cas2 and max your CPU clock which is much more important for most benchmarks than ram speed.
 
Cas is the most important right? cause mine is at 2.5-3-3-8, i got it to 2.0-2-2-6 but there was a lot of artifacts..
got it running @ 3.38ghz, but since 3.1 was crashing during the CPU test on 3d03 i havent rerun it yet :)
So why would it be that my CPU can run @ 3.38 and sit here all day, and when i run a test itll crash even though the temp only rises like 5 degrees, and i just cooling it more than its able to oc? (i thought more cooler, more fa$teR) Or is there something else afoot?
 
Its a known fact that running your ram out of sync with the P4 you starve it of performance. Its something you really should not do. It IS a large hit on performance.
 
http://www.abxzone.com/forums/showthread.php?threadid=53642

In both of these cases, Super PI performed better with the ratio, and that was at the same FSB, same CPU speed, just tighter timings with the ratio.

More actual results:
http://djtech.org/hardware/memory/latency/index4.html < 3d benchmarks
http://djtech.org/hardware/memory/latency/index5.html < 3d games
http://djtech.org/hardware/memory/latency/index6.html < conclusion

from conclusion said:
2-3% Improvement
While bandwidth is still very important to the Intel Pentium 4, it's not as important as it once was in the i845PE days of single channel memory controllers. Thanks to the i865PE/i875P's dual channel memory controller things are much brighter. On average, the system with the memory running at 400 MHz (5:4 memory divider enabled) with aggressive memory timings performed 2-3% faster than the system using high speed memory with loose timings.

If you're in the market for new memory for your Pentium 4 system and you're only thinking about gaming performance, then you're best bet is to get DDR which is rated to run aggressive timings. Some examples include Mushkin's PC3500 Level II which is rated to run 2-2-2-5 at 217 MHz FSB or Corair's TwinX-3200LL which are rated for 2-2-2-5 at 200 MHz.
Click to expand...

Now granted some of the new DDR500+ can run at 2.5 all the way up to near DDR600 speeds so they would be better than the loose 3-4-4-8 timings. But just supports my statements earlier. That using a ratio with tight timings IS NOT A BIG PERFORMANCE HIT, AND IN MANY CASES IF YOU CAN USE TIGHTER TIMINGS ITS ACTUALLY FASTER. SO I still stand by my statement that spanckhy is talking out of his ass. And if you don't like it, go cry to an admin! That I spanked you for diseminating bad information and stupid blanket statements which don't take all variables in to account.
 
DanThePlanMan said:
1:1 will give you the most memory bandwidth.
Click to expand...

sacrificing memory bandwidth for higher clock isn't a good idea unless you are running a64. I could run my 2.8 @ 4.0ghz easy with a 5:4 divider. but my P4 running 1:1 at 3.5ghz (667FSB) ran circles around it.
 
well on my rig it seems to be about even
CS:S is .1Fps off, 39.51 @ 5:4 or 39.66 for 1:1
3dmark 03 is just as convicing as CPU tests came back as 714/713 respectivly, with the troll lair coming in a few FPS ahead w/ the divider (14.1/13.5)but flight had a few loss (74/71) so i think its more of which end will be more stable, OC'ed memory which is seeming to give me a lot of crashed programs and uneasy ejects from games, or a divider which tends to just lock up
ugg they can never be easy :)
maybe ill get a watercooling system..;)
 
you need to test with Prime 95 (preferrably 2 instances of it). If your system won't run it for at hour or so, its not stable enough to game, etc....

write down each FSB, ram timing, etc... and test each step along the way (w/Prime95) until you find your highest CPU OC and then do it also for your highest 1:1 memory OC... THEN use those 2 100% stable tested spots to compare performance.

Your main problem is not thoroughly testing. Also, I think you might be putting too much volts in your system. If you're not stable at about 1.6v increasing the voltage will probably only increase your chance of locking up.

I was very lucky though b/c my chip will hit 3.4Ghz at default voltage.
 
what is the stock voltage for mine?
I can only set it to "auto" but that seems to cause more problems
if i had a stock/known voltage i could do what you said, but trying to do cpu steps + voltage, then memory, that would take a week lol
it seems to be most stable at 1.775 but maybe thats me.
 
yeah, thats an assload of voltage, turn it down right now unless you have it on phase, even so it will most likely croak
 
Canon20d said:
what is the stock voltage for mine?
I can only set it to "auto" but that seems to cause more problems
if i had a stock/known voltage i could do what you said, but trying to do cpu steps + voltage, then memory, that would take a week lol
it seems to be most stable at 1.775 but maybe thats me.
Click to expand...

Like I said you're putting too much volts.
Try 1.6v
maybe 1.65v

But I've found that I rarely could get higher with more than 1.6v, usually anything higher than 1.6-1.65 would cause the processor to lock up after a little use.
 
Hmmmm so the 1.9 i used before was to much to eh?
:-/ oops
im more familair with P-M with centrino hardware control, i tried to look at this one using [auto] on stock speeds, and the mobo said it was 1.824v.. so i was working from that. :eek: :eek: :eek:
ill turn it down ASAP
 
ok, ive been using 1.5875 and 1.6 since thats the 2 CPU-z seems to be toggling between, but when i boots windows it BSOD's, im thinking it might be the ram voltage, what should i be running? i usually use 2.765 but i dont know how to see what its at stock?
Info on XMS3200 512x2?
 
Canon20d said:
ok, ive been using 1.5875 and 1.6 since thats the 2 CPU-z seems to be toggling between, but when i boots windows it BSOD's, im thinking it might be the ram voltage, what should i be running? i usually use 2.765 but i dont know how to see what its at stock?
Info on XMS3200 512x2?
Click to expand...

If windows is showing it as 1.5875 - 1.6v.... maybe put it as 1.625v in BIOS, hopefully that will keep you at about 1.6-1.62v which is about the max for long term stability. Higher voltage may allow you to boot in to windows at a higher overclock, but the chip won't run stable for long.

stock voltage on Northwoods P4's = 1.525v
stock DDR ram voltage = 2.5v

Prescott and DDR2 both have lower voltages. (newer systems)
 
ok, thanks for the info, i think i was trying either to much voltage on the CPU or ram, cause it posts ok, then windows bluescreens as soon as it trys to load :p
and some voltages it likes (ie boots) others wont even post..takes a while to turn it off and get it reset itself - least i deleted the bios image so it doesnt take 20 seconds to check the nvram anymore
 
Ok, so why is that when i set the FSB @ 257, witha 5:4 divider, and set the cpu @ 1.585v, when i look @ Asus monitor or CPU-z it shows 1.6845v??
Im really trying to hit 262 (3.4ghz, i know 60mhz wont make a difference but its nicer than 3.34, and its not That far away, and so it shouldnt be hard to get that little bit further :)
 
You must log in or register to reply here.
Back
Top