BH-5 vs TCCD @ 300+

[jikdoc]

my own personal experience with this is that it takes about 30mHz extra for 2.5-3-3 timings to equal 2-2-2. that is 2-2-2 @ 250 is equivalent to 2.5-3-3 @ 280

[HARDCORECLOCKER]

Very nice results hardcoreclocker - Are these results on the prommy?

THX bias - yes done with MACH II GT:

IDLE -63°C / Normal load -56°C / full load -52°C

@MarilynMX:

Ya' right, stock timings of G.SKILL 4800LA is 2,5-4-4-8 at 300.

AGAIN for all who didn't read: If pics do not work try later - sometimes to much access on my HP.

[Zeus]

WOW
Jeus are you water cooled? my 3400NC on air does 2.66 its a 0440 - whats urs?

*really starts to consider WCing :p

Yes, just with plain water, CPU is a 0434MPMW, a serious kickass one.
Here's some more:


@HARDCORECLOCKER,
The BH-5 has a little disadvantage at CPU speed in your benches but good info anyway.
Could you repeat those benches with the same CPU speed and Process Timer running?

People are talking about rasing HTT and running BH-5 with divider. Well, I did that with my BH-5 and DFI DLI-DR.

10x270 2-2-2 1:1 (LDT @ 1080) v 7x385 2-2-2 7:10 (LDT @ 1155)

And 10x270 was faster in both sandra and everest. Didn't take any ss but then again I though that this was common knowledge.

I didn't get that divider thing myself.
Good you brought it up, i was thinking the same but forgot all about it.

[perry_78]

Ah well. I still like TCCD, i think they're much "safer" to run in a 24/7 rig. Giving those volts to BH5 and UTT sounds rather risky

[Ulti]

Let me explain the reason why i got UTT.
I've a 3200+ C0 claw, it clocks quite well seeing I can run 2600mhz at watercooling. It has a rotten memcontroller so anything above 266mhz fails (yes I've tried TCCD that can do 280+). I had a gig of OCZ rev2 so running 10x263 2.5-3-3-6 1T was okish. I had the opportunity to trade my OCZ for a nice Asus 6800GT 128mb so I could get some UTT. My 44D UTT will probably run 260 2-2-2-5 when I get my Booster here, if not 250mhz shouldn't be a problem. I figured 250 2-2-2-5 is better then 263 2.5-3-3-5. And I also got a nice 6800GT without having to pay anything for it If your CPU can't do 280+ UTT just looks like the best solution to me.

[Zeus]

Let me explain the reason why i got UTT.
I've a 3200+ C0 claw, it clocks quite well seeing I can run 2600mhz at watercooling. It has a rotten memcontroller so anything above 266mhz fails (yes I've tried TCCD that can do 280+). I had a gig of OCZ rev2 so running 10x263 2.5-3-3-6 1T was okish. I had the opportunity to trade my OCZ for a nice Asus 6800GT 128mb so I could get some UTT. My 44D UTT will probably run 260 2-2-2-5 when I get my Booster here, if not 250mhz shouldn't be a problem. I figured 250 2-2-2-5 is better then 263 2.5-3-3-5. And I also got a nice 6800GT without having to pay anything for it If your CPU can't do 280+ UTT just looks like the best solution to me.

Nice trade man!
I agree 10x250 2-2-2-5 will perform better than 10x263 2,5-3-3-5.
In my case the TCCD was the better choice but if i had some BH-5 that could do 286 there's no doubt which would dominate.

[HARDCORECLOCKER]

It's always a good idea to keep some BH-5 for future - someday You might need 'em.

[Rabbi_NZ]

I agree 10x250 2-2-2-5 will perform better than 10x263 2,5-3-3-5.

I highly doubt it...
maybe 10x250 2-2-2 (2500 MHz) would out perform 9x277 2.5-3-3 (2500 MHz), but 10x263 2.5-3-3 has 130 MHz advantage on the CPU... Im pretty sure it would win out every time.

[Zeus]

I highly doubt it...
maybe 10x250 2-2-2 (2500 MHz) would out perform 9x277 2.5-3-3 (2500 MHz), but 10x263 2.5-3-3 has 130 MHz advantage on the CPU... Im pretty sure it would win out every time.

Please, don't make me test it.

[Rabbi_NZ]

Please, don't make me test it.

hahahahahaha, looks like there's only one way to be sure!

[jumanji969]

vx and bh-5 need massive volts to do that though. tccd doesn't need the voltage to do stock timings.

Who runs 3.6 volts through their memory in a system 24/7?

Running 3.0 volts through the mem heats it up like an inferno. I would hate to see how hot 3.6 volts makes it.

3.0v hardly gets bh-5 hot. 3.3v with limited air flow over bh-5 is usually fine.

[bachus_anonym]

It would be better to see this kind of comparison:

BH5 @ 286x10 5:6 (divider166) vs TCCD @ 286x10 1:1 (divider200)

48Mhz instead of 57Mhz (238Mhz vs 286Mhz) difference in memory clock would most likely give equal times or very close to it.

BTW: nice TCCD you got there, Zeus :thumbsup:

[felinusz]

What kills me is so many people calling your 260 MHz, 2-2-2, with BH-5 "bad" . Pretty decent for only 3.4V if you ask me - those sticks have at least another 10 MHz in them Zeus, at 3.6V, 270 should be possible .

Interesting comparison here, thanks for posting it Zeus. Would you be able to post some simple game benchmarks with the same settings?


I would like to see an in-depth comparison between 2-2-2 timings, and a "maxed out" set of BH-5 running ~270, with some 2.5-3-3 TCCD doing ~320, at similar clock speeds. The goal being, to find the drop off point, what exactMHz increase, at 2.5-3-3 is needed to overcome 2-2-2.



The new G-Skill DDR600 can do such high speeds fairly consistantly with 2.5-3-3 timings, and with low voltage to boot.


I am a little perplexed with so many people saying that "not many people are running 2.5-3-3 with TCCD, at DDR600+ 24/7", do these people not realize that very very soon, a whole heck of a lot of people will be running such memory speeds 24/7? DDR600 is upon us, and it flies like a bat out of hell.

Originally Posted by kickassclone

Who runs 3.6 volts through their memory in a system 24/7?

I did for a while, as do many others. BH-5 can take this kind of voltage, if you actively cool it.

[Zeus]

Thanks for the kind words guys.

I didn't want to do any divider stuff as this always kills performance, and 1:1 is what most of us run--i know i do.
Neither did i bother with the TCCD running it below 300MHz, imho running TCCD under 300MHz doesn't justify the slack timings, in other words, one would be better off with VX/BH-x.

But, if i can find some time i will do some more tests to find out at what exact speeds both mem will match.

Like suggested by some, i think the TCCD needs some 30MHz extra to match the BiatcH-5 with it's tight timings, give or take a few MHz depending on the benchmark.

Actually, i mainly did this test to show that BH-5/VX isn't always the answer to our quest.
In this specific case the TCCD was the better choice and can only be equaled or topped by some magic BH-5 that runs 286MHz.

I'm not a fanboy of both, i just like to have the best performance.

[zakelwe]

Do you get much gain with the bh-5 on pifast if you set TrTw to 1 rather than 2 ?

Regards

Andy

[blinky]

picking two numbers and comparing performance doesnt allow us to draw any conclusions at all except that in that exact setup, or in the same extremes (less bh5 clocks, more tccd clocks) the same results are true.... but that doesnt tell us anything

could someone run extensive tests and find teh equilibrium point (approximately) where bh5 @ 2-2-2, is equal to tccd @ 2.5-3-3, and what i would use as a benchmark is 300mhz on the tccd, seeing as most people can hit around that, but most cant hit much more on 2.5-3-3,

if someone wants to do lots of work they could see where 2.5-4-3 matches bh5, prolly a couple mhz higher than 2.5-3-3

[NapalmV5]

Anyone run 2 gigs 4x512 of TCCD @ 300/300+ ?

[saaya]

Sounds about right, 10x286 with the BH-5 would have probably killed the TCCD but my BH-5 couldn't do that, who's can?

i know a guy who has a 256mb bh6 stick that does 280 with 3.6v
but yeah, thinking about the average oc tccd beats low latency memory... :/

[Formann]

i know a guy who has a 256mb bh6 stick that does 280 with 3.6v
but yeah, thinking about the average oc tccd beats low latency memory... :/

I know one to

A Twinmos 256mb 280 @ 3.6v and a Twinmos 256 280 @ 3.75v .. need 3.78v for 280 DC tight.

280 2-2-2-5-7-12-2-2-1-1 .. need some SERIOUS kickass TCCD to beat that.

[[XC] serlv]

But then again, who has TCCD that does 317mhz? My OCZ rev2 stops at 288 My UTT can do 220 2-2-2-5 at 3.1V, waiting for that booster to arrive

A little OT, but have been looking for an answer to this, as I've seen it a lot lately - What, exactly, is UTT?

BTW, nice OC's, Zeus.
Very informative comparison., too

[Marquzz]

I didn't get that divider thing myself.
Good you brought it up, i was thinking the same but forgot all about it.

Well, first I ran 10x270 with ram 1:1 with 2-2-2 timings, 4xLDT then I lowered the multi and raised the FSB and ran 7:10 ram divider so 7x385 and LDTx3, because of the ram divider of 7:10 the ram was clocked at almost the same speed as scenario 1 (270), 385/10 * 7 = ~270. And the cpu was also at the same speed, 385*7 = ~2700. So in both scenarios cpu and ram had the exact same settings, only the fsb and htt was raised. So even with fsb and htt much higher I still got lower scores in both sandra and everest.

Hoped that cleared things out

Summery:
Scenario 1:
10x270 (2700), 2-2-2 (270), 1:1 4xLDT

Scenario 2:
7x385 (2695), 2-2-2 (269.5), 7:10 3xLDT

[Ulti]

A little OT, but have been looking for an answer to this, as I've seen it a lot lately - What, exactly, is UTT?

BTW, nice OC's, Zeus.
Very informative comparison., too

The newest WinBond chip, it stand for UnTesTed. It's revised CH5 if I'm correct.

[eva2000]

can you add 3d2k3 benchies to the original comparison ?

[Zeus]

Do you get much gain with the bh-5 on pifast if you set TrTw to 1 rather than 2 ?

Regards

Andy

To be honest, i don't know, i never used anything but 2.

I know one to

A Twinmos 256mb 280 @ 3.6v and a Twinmos 256 280 @ 3.75v .. need 3.78v for 280 DC tight.

280 2-2-2-5-7-12-2-2-1-1 .. need some SERIOUS kickass TCCD to beat that.

Lucky b@stards i can't shove more than 3.4V up mine.

can you add 3d2k3 benchies to the original comparison ?

I could but i don't think you would want to see them, done with a GF4ti4200

[saaya]

%age is %age... whether the number is small or huge :P