CH-5 benchies on Infinity

[Fewture]

Ok guys, I knew my KHX3500 CH-5 was pretty good but check this out. Today I got an Infinity NF2 to replace my good old NF7-S. The NF7 was great, but I needed a new toy :D

http://www.fewture.cistron.nl/10x252sisoft2.JPG

http://www.fewture.cistron.nl/10x254sisoft.JPG

Proc maxed on this Vcore. The chip can do 2655MHz in pifast @ 2.27V (plain water).
http://www.fewture.cistron.nl/10x249pifast.JPG

FSB maxed:
http://www.fewture.cistron.nl/255pifast2.JPG

Settings:

2.0V Vcore
3.3V Vmem
1.9Vdd (it can also do it on 1.8 I think, because it's the mem that is maxed)
11-2-2-2
command/clock: ena
cant remember the rest :D

Now I have a prommy coming (to play with only). With that and some mods I think something good will come :D I am pretty amazed by this RAM... You see this with BH-5. And my sticks wouldn't even boot 200MHz ras2cas 2 the first week or so.

[Fewture]

Some more.. now with 1-31 beta bios:

http://www.fewture.cistron.nl/255sandra131bios.JPG

MP = 9.5

Hmm..the 3.3V setting only gives 3.22V. I'll mod it tomorrow. All voltages suck on this board btw.

[blinky]

wow, that is excellently superb for CH5

now are you its CH5
just for kicks you should take the heatspreaders off and see what they really are. if they are CH5 this will prove that in fact CH5 can take a Ras to Cas setting of 2.

another interesting thing is that Grace says that on CH5 they dont even have the pipelines available to run ras to cas of 2, if this is true then somehow they either implemented a fix to that problem and now ras to cas of 2 should be possible, OR they figured out a way to have bios set it to 2 when really inside the memory chip it acts the same as if it were 2.
you can test the second one by testing sandra scores with the ras to cas set to 2 and then run identical tests with the ras to cas set to 3 keeping everything else the same

[Grace]

Originally posted by blinky

another interesting thing is that Grace says that on CH5 they dont even have the pipelines available to run ras to cas of 2, if this is true then somehow they either implemented a fix to that problem and now ras to cas of 2 should be possible, OR they figured out a way to have bios set it to 2 when really inside the memory chip it acts the same as if it were 2.


I also say a thing about Nforce2 motherboards each time I say that.

Nforce2 motherboards can take the timing, it's considered a bug which can be true since the performance altered is nothing near what the real change from 3 to 2 would give anyway, that's the most effective timing. Trying them on any other motherboard would not let them run at 2/6/2/2. All CH-5 I tried could run at 2/6/2/2 on Nforce2 and apparently a few Nforce3 motherboards, none of them could take those timings on any 865/875 motherboard.

[STEvil]

I think we need a 230mhz @ 2-6-2-2 2300mhz direct comparison between BH-5 and CH-5. (I just picked 230mhz/2300mhz out of the air).

Same mobo, same cpu, same everything.. just swap stick of ram for stick of ram and rebench.

[Soulburner]

Well STEvil it would then have to be made sure that both sticks were either DS or SS, as the bandwidth each produces is different at the same clocks.

Most 256mb modules are SS though.

[blinky]

yah we could use KHX3200/256 and KHX3200A/256

those are almost the same PCBs and the only other difference is the ICs used on them

[Fewture]

Originally posted by blinky
wow, that is excellently superb for CH5

now are you its CH5
just for kicks you should take the heatspreaders off and see what they really are. if they are CH5 this will prove that in fact CH5 can take a Ras to Cas setting of 2.

another interesting thing is that Grace says that on CH5 they dont even have the pipelines available to run ras to cas of 2, if this is true then somehow they either implemented a fix to that problem and now ras to cas of 2 should be possible, OR they figured out a way to have bios set it to 2 when really inside the memory chip it acts the same as if it were 2.
you can test the second one by testing sandra scores with the ras to cas set to 2 and then run identical tests with the ras to cas set to 3 keeping everything else the same

My heatspreaders are off for months now ;) 100% CH-5 m8. I got the mem months ago and it would not boot ras2cas 2 at first. It took extensive burning in. Before these sticks I had Corsair CH-5 (3200LL Twinx). Also did ras2cas 2 but needed insane voltage to do it.

These KHX sticks also get a hit when running ras2cas 2, but it works.

Originally posted by Grace
I also say a thing about Nforce2 motherboards each time I say that.

Nforce2 motherboards can take the timing, it's considered a bug which can be true since the performance altered is nothing near what the real change from 3 to 2 would give anyway, that's the most effective timing. Trying them on any other motherboard would not let them run at 2/6/2/2. All CH-5 I tried could run at 2/6/2/2 on Nforce2 and apparently a few Nforce3 motherboards, none of them could take those timings on any 865/875 motherboard.

Well, my pifast times go down when setting ras2cas 2. I also had Twinmos CH-5 in this board and it would not do ras2cas 2. So I think you're wrong here :)

[Fewture]

Ok..I tested ras2cas 2 and 3 on 9,5 x 230MHz

http://www.fewture.cistron.nl/230pifast11-2-2-2.JPG

http://www.fewture.cistron.nl/230pifast11-2-3-2.JPG

Thats more than 1 second.. Says enough imo.

[Grace]

Originally posted by Fewture
So I think you're wrong here :)

It took you extensive burn in and now you can set RTC to 2 when you couldn't before at any speed and voltage?


Well, yes, I'm wrong. If burning in creates pipelines, what can I say?


I didn't say that performance won't be altered at all but certainly not NEAR as much as the real RTC set to 2. That gives a tough boost on performance on its own, not a little faster SuperPi which can be due to luck anyway. You have an Nforce2, you always could run at 2.

[stevenmh]

I just upgraded from 2 x 256 OCZ EL PC3500 DC Platinum CH-5 to 2 x 512 OCZ EL PC3200 DC Platinum CH-5. The P95/Memtest stable limit for both sets is 227 @ 11-2-2-2, 1.95v chipset, 3.1v memory. I'm going to re-mod for ~2.2v / 3.5v and see how much more I can get out of them.

I never did any burn-in. However, on the PC3500 it was doing 217 for awhile limited by my XP2500. Then I upgraded to XP3200 and was able to do 227. When I upgraded to 1GB PC3200 I just left it at 227 11-2-2-2 and plopped the new ones in and fired it up.

I also tried 2 x 512 Mushkin Level II PC3500 BH-5 ...wasn't stable much beyond 200MHz at any voltage or timings with my XP3200. With an XP2500, I couldn't get it stable even at 166. The computer wouldn't even post if I tried CAS3...just sirened.

[skate2snow]

Originally posted by Grace
I didn't say that performance won't be altered at all but certainly not NEAR as much as the real RTC set to 2. That gives a tough boost on performance on its own, not a little faster SuperPi which can be due to luck anyway. You have an Nforce2, you always could run at 2. You cant say that CH5 CANT run 2-2-2-x, more then 20 people in one thread said it was able to do on theyre CH5, NF2, 865PE, 875P have said it can do 2-2-2-x, but ive notice that only people that buyed theyre RAMs recently can do it.

I think we can say its a fact: new CH5 can do 2-2-2-x

[Grace]

Originally posted by skate2snow
You cant say that CH5 CANT run 2-2-2-x, more then 20 people in one thread said it was able to do on theyre CH5, NF2, 865PE, 875P have said it can do 2-2-2-x, but ive notice that only people that buyed theyre RAMs recently can do it.

I think we can say its a fact: new CH5 can do 2-2-2-x

I can say all I like actually, don't you love democracy?


I tried hundreds of such modules, none took such timings on any other motherboard than Nforce2 motherboards. Winbond themselves told me their CH-5 modules can't work at RTC 2. Also, I used modules last week that will be on shop shelves next week or so, what do you consider new, those the Winbond will get out next summer?


If that's a -fact-, go and buy CH-5, why paying double money for BH-5 anyway?

[skate2snow]

The only thing i dont catch its more then 20 people are doing 2-2-2-x and youre still saying CH5 cant do 2-2-2-x

[Grace]

Originally posted by STEvil
I think we need a 230mhz @ 2-6-2-2 2300mhz direct comparison between BH-5 and CH-5. (I just picked 230mhz/2300mhz out of the air).

Same mobo, same cpu, same everything.. just swap stick of ram for stick of ram and rebench.


I'll be setting up a 3200+ system for some testing here at work in a little while, I can't PROMISE it but I'll try to take a benchmark with my Pc2700 BH-6 and HyperX Pc3200 CH-5 at 2/6/2/2 under equal speeds. If I can, I will PM you the links to the screenshots to do whatever you like with them.

....it's just not like me to start threads and the like.


The only thing i dont catch its more then 20 people are doing 2-2-2-x and youre still saying CH5 cant do 2-2-2-x

I could make a list of things that I don't catch and you didn't explain to me yet. Plus I don't know how to count, so I must have missed those 20 people that you said.

Sorry if I'm being a b!tch.

[skate2snow]

Grace: I dont want a war, this is only a link of the more then 20 people, just can you read that please?

http://www.xtremesystems.org/forums/showthread.php?s=&threadid=26847

[Grace]

Originally posted by skate2snow
Grace: I dont want a war, this is only a link of the more then 20 people, just can you read that please?

http://www.xtremesystems.org/forums/showthread.php?s=&threadid=26847


I read it before.

Count how many run on nforce2 motherboards, then come and tell me how many are the rest. I'm too busy at the moment to use my mind on such things.

[Vlad Draculea]

no interesting questions about ram internals.

[Fewture]

While all you guys are talking about the CH-5 thing I took some BH-5 and kicked it's arse :D

http://www.fewture.cistron.nl/264x9.5.2.JPG

Not the highest, but still nice I think :)

[Grace]

Originally posted by Vlad Draculea
grace wont you have any link talking about trcd pipelining, some time ago i read one but now i cant find it, and if my memory doesnt fail, the only thing that was pipelined was cas.

No, I'm sorry but I don't.

If my own memory serves the only thing that isn't affected by the pipelines is the TRCD, it just names the time the memory will be 'open' for a burst of cycles (or whatever, my translation skills are horrible at times).

I'm not taking an oath.




I took the screenshots I said, I'll just PM STEvil with them as I also said.

[blinky]

i think the only way to resolve this, is for someone to email or call winbond and ask them.
any takers??

[skate2snow]

winbond rate theyre chips at 2-3-3-7 for CH5, so if you ask them they will say 2-3-3-7

[Grace]

Originally posted by blinky
i think the only way to resolve this, is for someone to email or call winbond and ask them.
any takers??


I did that....and I said that I did it.....


winbond rate theyre chips at 2-3-3-7 for CH5, so if you ask them they will say 2-3-3-7

He meant to ask if the chips can take the RTC of 2.

[blinky]

Originally posted by Grace
I did that....and I said that I did it..... well did you directly ask them about how some newer CH5s on the nForce2 platform can do ras to cas of 2 and how that is contradictory to what they are telling you.

sometimes you gotta :stick: them a little to get the info you want

[Grace]

Originally posted by blinky
well did you directly ask them about how some newer CH5s on the nForce2 platform can do ras to cas of 2 and how that is contradictory to what they are telling you.

sometimes you gotta :stick: them a little to get the info you want


All CH-5 always did 2/6/2/2 on Nforce2 motherboards.
Or actually they didn't, as STEvil will/may/maybe not show you when he comes online, I PMed him the compare screenshots which will tell why many people (me included) consider the Nforce motherboards running RTC at 2 is a bug.

[Fewture]

Well this thread really sucks now.. tnx guys.

I don't f*ckin care whats the deal with the CH-5. All I know is everybody is complaining it won't go fast and it sucks. Well...it does go fast as you can see. You can also see BH-5 is still better (as we all know)

This stupid argument about CH-5 RTC2 helps nobody. Fact is some don't get it to run RTC2.. And yes, also on NF2 Grace. What the reason is..I don't know and I don't care :) this is a yes/no game..grow up.

[skate2snow]

OK, i think you make this clear;)...LOL

[Bennah]

I just got some CH-5 2 days ago and I havnt really done any buring in yet but this is what I got to...not stable though....temps are abit high...still on AMD stock HSF.

[enzoR]

how much voltage?

[Fewture]

Originally posted by skate2snow
OK, i think you make this clear;)...LOL

LOL :D I just don't see the sense of it..thats all ;)

[Bennah]

Originally posted by enzoR
how much voltage?

2.9v...If I cant remember correctly....

[Penti]

Originally posted by skate2snow
winbond rate theyre chips at 2-3-3-7 for CH5, so if you ask them they will say 2-3-3-7

No, they are rated at Cas 2.5. Where have you got 2-3-3-7 from?
What RAM-manufactures say they will handle or even program them to and what the manufacture of the actual memory module have it rated for is a different thing. BH-5 was rated at Cas 2.5 @ 200MHz too..

BTW I too have seen older CH-5 at 2-2-2-5 at nF2. But on I875/865 is a bit unusual.

[skate2snow]

I meen like compagnies rate theyre PC3500 and most of PC3200 at 2-3-3-7. I know 2-2-2-5 and PC3500+ are not "aproved".

[Penti]

Originally posted by skate2snow
I meen like compagnies rate theyre PC3500 and most of PC3200 at 2-3-3-7. I know 2-2-2-5 and PC3500+ are not "aproved".
But you said "winbond rate theyre chips at 2-3-3-7 for CH5" wich they don't and theres no difference between 2-2-2-5 and 2-3-3-7 in this matter, none of them are "approved" for PC3200/5ns speeds.

[skate2snow]

I know, b ut ive said it the wrong way. Sorry, but meen compagnies(ex: hyperX PC3500: 2-3-3-7).

[STEvil]

Thanks for the PM, Grace.

10.5x235 3200+
Epox 8RDA3+
One pair of Kingston Pc2700 BH-6 of mine and the HyperX Pc3200 CH-5 in compare.

BH-5 - http://users.tellas.gr/~dawn/BH-5.jpg
CH-5 - http://users.tellas.gr/~dawn/CH-5.jpg

There you have it, 2-2-2-6/2-6-2-2/whatever isnt as fast on CH-5 chips as it is on BH-5/6 chips.

[skate2snow]

did you test CH5 at 2-2-3-x and 2-2-2-x?

[STEvil]

Grace, want to add a 2-3-2-6 time to the above with both the BH-6 and CH-5 sticks? May as well run the gamut...

[Grace]

Originally posted by STEvil
Grace, want to add a 2-3-2-6 time to the above with both the BH-6 and CH-5 sticks? May as well run the gamut...

I did that yesterday, I did take screenshots, but I'm not at work at the moment. I'll PM it to you in a few hours.

Running at 2/6/3/2 DOES impact performance but not as much as it should. You'll see what I mean.

[STEvil]

Can you get a sisoft of the BH-6 at 2-2-2-6-1T, too?

I did the L12 mod to my 8rda+ and can do 10x235.9 stable.. so 10.5x235.9 might just go for a sisoft run.

I can put up my scores with mh DC DS 2 x 256 BH-6's for confirmation... :D

[Grace]

Originally posted by STEvil
Can you get a sisoft of the BH-6 at 2-2-2-6-1T, too?


No, I'm afraid that I can't. I didn't take such a screenshot and it would be useless for comparisons anyway, synthetic benchmarks like Sandra are barely affected by timings and they fluctuate due to pure luck, it would be a bad idea for comparisons.

[STEvil]

ok, gimme a link to the program you are using (one of those pi crunchers I assume?), i'll use it instead :D

EDIT

More Benches by grace.

CH-5 2-3-2-6: http://users.tellas.gr/~dawn/CH-5rtc3.jpg
BH-5 2-3-2-6: http://users.tellas.gr/~dawn/BH-5rtc3.jpg


Notice the CH-5 beats the BH-5 in "Final Huge Multiplication Time" but loses overall.

[Grace]

Originally posted by STEvil
ok, gimme a link to the program you are using (one of those pi crunchers I assume?), i'll use it instead :D


It's the pifast. Mine is pre-set since I have it for a while for fast testing equipment here and I don't remember a link, just do a google search, you can't miss it. Or actually shouldn't miss it!

Edit : If you need anything else, send me a PM. Fewture was deeply bothered by my presence it seems, I don't want to ruin his thread.

[Fewture]

Originally posted by Grace
It's the pifast. Mine is pre-set since I have it for a while for fast testing equipment here and I don't remember a link, just do a google search, you can't miss it. Or actually shouldn't miss it!

Edit : If you need anything else, send me a PM. Fewture was deeply bothered by my presence it seems, I don't want to ruin his thread.

No offcourse not. It had nothing to do with you personally m8 ;) Just the way how everybody is playing the yes/no game. I know CH-5 is not as good as BH-5 and I showed that to all of you by posting some BH-5 benchies on the same board.

I was just happy that I had CH-5 running these insane speeds (for CH-5 standards) and then the threads turns into a yes/no game :rolleyes: :p:

I also didn't agree with the RTC2 "bug". It's not a bug. The fact that BH-5 gains more from 3 to 2 is (to me) just a sign that the chips behave differently. I don't care _why_ that is. I just take it as a fact.

[Grace]

Originally posted by Fewture

I also didn't agree with the RTC2 "bug". It's not a bug. The fact that BH-5 gains more from 3 to 2 is (to me) just a sign that the chips behave differently. I don't care _why_ that is. I just take it as a fact.

To that, when comparing modules of the same specifications with different chips (let them be anything that can be compared to each other), the timing changes affect the speed about the same. Only on the Nforce motherboards and the CH-5 the particular timing affects nearly nothing while BH-5/6 (no other can do that at this speed to compare with) do give a good gain. As far as DDR333+ modules are concerned, I never found that behavior on any other comparison between any of the timings.

[Fewture]

Originally posted by Grace
To that, when comparing modules of the same specifications with different chips (let them be anything that can be compared to each other), the timing changes affect the speed about the same. Only on the Nforce motherboards and the CH-5 the particular timing affects nearly nothing while BH-5/6 (no other can do that at this speed to compare with) do give a good gain. As far as DDR333+ modules are concerned, I never found that behavior on any other comparison between any of the timings.

I've posted screens in this thread where RTC2 gained 1 sec in pifast vs RTC 3. How much would you say BH-5 would gain then?

[Endre]

Originally posted by Fewture
I've posted screens in this thread where RTC2 gained 1 sec in pifast vs RTC 3. How much would you say BH-5 would gain then?
http://users.tellas.gr/~dawn/BH-5.jpg
http://users.tellas.gr/~dawn/CH-5.jpg
http://users.tellas.gr/~dawn/CH-5rtc3.jpg
http://users.tellas.gr/~dawn/BH-5rtc3.jpg

[Fewture]

Originally posted by Endre
http://users.tellas.gr/~dawn/BH-5.jpg
http://users.tellas.gr/~dawn/CH-5.jpg
http://users.tellas.gr/~dawn/CH-5rtc3.jpg
http://users.tellas.gr/~dawn/BH-5rtc3.jpg

Ok..now I feel stupid :D Hehe

[STEvil]

heh...

We should get someone with some of those old samsung modules to give this a run, too.. see how close they are to BH-5.

[nutcase]

Ok, I have to reply..;)

Here is what I have been able to do with my Kingston HyperX PC3500 with ch-5 chips...

All benchmarks are at 2-2-2-11 timings and at 1:1 ratio. all Vdimm voltages between 3.2-3.3V.. all are benchmark stable.

Abit Nf7 2.0 and Xp1700+ DLT3C - 235 FSB (chipset gives out higher)

DFI Ultra Infinity and XP1700+ DLT3C - 257 FSB (might go higher but have not tried yet)

Shuttle AN50R and AMD64 3200+ - 243 FSB (chipset acting up)

ASUS P4C800 DLX and P4 2.4C - 240 FSB

_______________________________

Note: None of these systems will run over 240 FSB with my HyperX PC3500 with BH-5.....

Another thing, Since BH-5 just cannot be found anymore, Why do we keep Reccomending it.. we are stuck with CH-5 or BH-6 :mad:

[STEvil]

We arent debating that CH-5 cant do high FSB's with tight timings, we are debating that BH-5/6 actually USE the tight timings where-as with CH-5 it isnt as effective, which means the timings arent taking full effect on the CH chips.

Another thing, Since BH-5 just cannot be found anymore, Why do we keep Reccomending it.. we are stuck with CH-5 or BH-6

Whats wrong with BH-6? Mine are doing 236 2-2-2-5-1T 3.1v on my 8RDA+ right now...

[nutcase]

Originally posted by STEvil
We arent debating that CH-5 cant do high FSB's with tight timings, we are debating that BH-5/6 actually USE the tight timings where-as with CH-5 it isnt as effective, which means the timings arent taking full effect on the CH chips.


Whats wrong with BH-6? Mine are doing 236 2-2-2-5-1T 3.1v on my 8RDA+ right now...

That might be true BUT if I can push my CH-5 17Mhz faster FSB than my BH-5, that overcomes the timing issue by a LOT... (on the DFI)...

BUT, It could be My BH-5 Is just JUNK. :eek:

As for BH-6.. Nothing yet.. When I can find some, I will Buy it in a Heart Beat..:p:

Forgot.. Somebody stated you cannot run CH-5 on an Intel Chipset and hold 2-2-2-X timings. well, I posted because I have been able to do it up to 240 1:1..

[Anarki]

My Hyper-X CH-5 runs at 240MHz on my NF7-S.

Timings are 2-3-3-7.

tbh im very happy with that, voltage is at 2.9V.

I notice a slight diff if i run them at 2-3-3-6 but i cant get the higher MHz out of it.

I can run them at 250 but Win2k starts closing everything down and writes out numerous error logs :smileysex:

[stevenmh]

Originally posted by Grace
To that, when comparing modules of the same specifications with different chips (let them be anything that can be compared to each other), the timing changes affect the speed about the same. Only on the Nforce motherboards and the CH-5 the particular timing affects nearly nothing while BH-5/6 (no other can do that at this speed to compare with) do give a good gain. As far as DDR333+ modules are concerned, I never found that behavior on any other comparison between any of the timings.

After reading this and looking at the benchmark jpegs I decided to play around with my system a little. I found that I could get an additional 1 MHz on my FSB by changing RTC from 2 to 3. (I don't know that it's actually stable... I've done P95 for 12 hrs at 227, and it always craps in a minute or two at 228. Setting TRC from 2 to 3 it ran for about 15 minutes and I stopped it to do benchmarks.) Here are the results on my system at both timings:

228 x 11, 11-3-2-2
Sandra 3408/3215
3DMark01 18659
3DMark03 6400

227 x 11, 11-2-2-2
Sandra 3410/3224
3DMark01 18831
3DMark03 6407

Not only does RTC2 give me better results than RTC3, it does so in spite of going up against an additional 1 MHz FSB and 11 MHz clock speed. I could bench 227 x 11, 11-3-2-2, and I will if anybody would like, in order to isolate the difference due only to the timing change. But I didn't bother doing it for my own benefit, because the above results tell me all I need to know regarding which settings are best for my system.

[STEvil]

2-2-2-5 will be faster for 3dmark.

2-2-2-11 is just fastest in sandra.

[stevenmh]

Originally posted by STEvil
2-2-2-5 will be faster for 3dmark.

2-2-2-11 is just fastest in sandra.

Perhaps on your system, but not on mine. I just ran a couple more rounds of benchmarking...


To address the 5 vs 11 blanket statement:

227 x 11, 5-2-2-2
Sandra 3376/3198
3DMark01 18811
3DMark03 6404

Compare to 227 x 11, 11-2-2-2 above. Lower scores across the board. Marginally, I'll admit, but still consistently lower.



Since I was back to it anyway, I did another to address the RTC 3 vs 2 on nF2 boards:

227 x 11, 11-3-2-2
Sandra 3389/3199
3DMark01 18588
3DMark03 6395

Compare to 227 x 11, 11-2-2-2 above. Lower scores across the board. Almost 250 points in 3DMark01 just by changing RTC.


So, on my system, my worst-to-best settings were:

227 x 11, 11-3-2-2
228 x 11, 11-3-2-2
227 x 11, 5-2-2-2
227 x 11, 11-2-2-2

[Endre]

Originally posted by STEvil
2-2-2-5 will be faster for 3dmark.

2-2-2-11 is just fastest in sandra.
That varies on different systems and FSBs. For me Tras of 10 is the fastest (tested with 225-230 MHz). ~150 3DMarks over 11 and even more over 5.

[wierdAl]

i've got my twinmos running prime at 2-2-2-11@2.7V@210... memtest stable... ran it over night...

but this is SC, the wierd thing is that the twinmos will only do those timings in the SLOT that is most distant from the CPU... i think that's Dimm1... (epox rda 2.1)

[Hallowed]

Originally posted by Endre
That varies on different systems and FSBs. For me Tras of 10 is the fastest (tested with 225-230 MHz). ~150 3DMarks over 11 and even more over 5.

Yeah, varies per system. And the Infinity does NOT perform at all if Tras is at 10 or 11.

[Fewture]

Originally posted by Hallowed
Yeah, varies per system. And the Infinity does NOT perform at all if Tras is at 10 or 11.
??
I'll check that out. My mem gets unstable when disabling CPC. But when I disable it the system takes a huge git :( Argh.. Can't run prime 240MHz with CPC on.

[Hallowed]

I found something odd about the Infinity. It may be that I need to sink a few critical components but somehow I doubt it would have the same effect.

After seeing some results of an Infinity on either 12-10 or 1-31, I concluded it was stronger-per-clock than the NF7-S. But I did a bit more testing... on the same BIOS, and found that the Infinity gets noticeably weaker per clock the higher it scales. My NF7-S is stronger at 11x223 than the Infinity at 10x242, and I dont mean sandra... which is pretty useless sometimes, but in 3DMark.

The NF7-S of mine has nudged 111 FPS in car high at 11x231, which as far as I know has only been passed by Infinity boards at substantially higher clocks / FSB (Like S.A.E's results).

I'll retest with 12/10 in order to check, but I did find it strange.

[STEvil]

How'd you manage 20k at those clock on your 9700 pro??

432/346 only nets me 19k, and i'm only 100mhz behind you on my cpu.

[Soulburner]

I thought the 10 multi and lower were known to be a little "weaker"?

[STEvil]

They are, on anything but the Infinity/Lanparty B.

10.5 is fine, though.

[Endre]

Originally posted by STEvil
How'd you manage 20k at those clock on your 9700 pro??

432/346 only nets me 19k, and i'm only 100mhz behind you on my cpu.
The CPU is the bottleneck. You need some speed!

[STEvil]

Originally posted by Endre
The CPU is the bottleneck. You need some speed!

100mhz for 1000pts?

I dont think so. :stick:

[LikwidKool]

hmmm I am interested in the 2-3-2-6 between the bh-5 and ch-5. My p4c800-E does not like the R2C 2 very much above 230 DC on any of my bh-5. My OCZ Plat LE can do it to 235 but that it. So basically I end up running 2-3-2-6 usually. I am think some ch-5 might be nice to try since I did not see much of a hit between the 2 at those timings in grace's screenies. Just a few tenths of a second. I am almost scared to try though after what I have paid for some of my bh-5 in the last few weeks:eek: Plus I see the ch-5 seems to be very strong in the 2x512 department. My Mushky L2 and XMS pc3200 2x512 give up around 235 1:1 and 230 5:4. Burn in is helping those numbers a bit, but very dissapointing to say the least.

I am eager to see what this ch-5 will do on an 875P. Looks like I am going to be spending some more money in the next week!

[unrealneo]

Originally posted by STEvil
I think we need a 230mhz @ 2-6-2-2 2300mhz direct comparison between BH-5 and CH-5. (I just picked 230mhz/2300mhz out of the air).

Same mobo, same cpu, same everything.. just swap stick of ram for stick of ram and rebench.

/me puts hand up

I can try that! :D (with a Tbred)

I have some XMS 3500 1.1 on loan ATM. :)

also have a 9800np, but unfortunately, something is making it overvolt by 1V on the vcore:eek:

[Endre]

Originally posted by STEvil
100mhz for 1000pts?

I dont think so. :stick:
You have Tbred, he has Barton. And Tras 5?

[ricjax99]

Originally posted by STEvil
100mhz for 1000pts?

I dont think so. :stick:

Erm.... :stick:

1. Your running XP :rolleyes:
2. Your running Cat 3.10
3. He's better at tweaking (no offence :) )
4. You have a Tbred not a barton
5. Hes using better memory timings
6. He has 100 MHz extra CPU.

[wierdAl]

stevil: "Mine are doing 236 2-2-2-5-1T 3.1v on my 8RDA+ right now..."
where did you see the 1T setting on the rda+?

[mind_warrior]

Originally posted by unrealneo
/me puts hand up

I can try that! :D (with a Tbred)

I have some XMS 3500 1.1 on loan ATM. :)



Someone did that already it seems in this very thread.:D
Tip : 200MHz (P4) gave me 900 points in 3dmark with a good 9800pro card overclocked to the limit, 480/400 I think.


I would like to ask the original poster, is he happy of the performance of the infinity? I have the first revision 8rda+ for my second pc and I consider the upgrade between this motherboard and the nf7-s. I'd like his opinion of the motherboard, too many nf7-s owners and too few dfi around where I am to ask.

[nutcase]

Originally posted by mind_warrior
I would like to ask the original poster, is he happy of the performance of the infinity? I have the first revision 8rda+ for my second pc and I consider the upgrade between this motherboard and the nf7-s. I'd like his opinion of the motherboard, too many nf7-s owners and too few dfi around where I am to ask.

I will throw in my 2 cents since I just switched from an NF7 2.0 to a DFI Infinity...

Best I could do with my NF7 was 235 FSB @100% Stable even with the VDD and Vdimm Mods....

Right Now, The DFI is fully stock and I am running My 1700+ DLT3C at 250x10 Fully stable. Can get it up to 257 FSB and benchmark fine. It has it's annoyances though but almost all Mobo's have some quirk..

forgot to say.. That is air cooled also on both..

[Fewture]

Originally posted by mind_warrior
Someone did that already it seems in this very thread.:D
Tip : 200MHz (P4) gave me 900 points in 3dmark with a good 9800pro card overclocked to the limit, 480/400 I think.


I would like to ask the original poster, is he happy of the performance of the infinity? I have the first revision 8rda+ for my second pc and I consider the upgrade between this motherboard and the nf7-s. I'd like his opinion of the motherboard, too many nf7-s owners and too few dfi around where I am to ask.

Well, the DFI is lots of fun and per Mhz a little faster than the NF7. But I cant get my CPU to run good on this board and that really, really sucks.. My CPU can do 2655MHz in pifast. With a lot of Vcore that is.. So with 2.1 or so 2580/2600 must be possible. Now I cant get it to run pifast @ 2500MHz.. The board is useless now.. I have a prommy coming but I doubt it will really help.

Imagine what 10 x 260 in pifast would do..but I cant run it :( 260 is no problem, but 2600 is..argh.. My cpu also doesnt heat up. Something is funny here.

[All4AMD]

since my turbo cooled psu is crapping i cant hit higher than 220 on the fsb so soon getting power supply and puting better nb and sb cooling

[Hallowed]

Originally posted by STEvil
How'd you manage 20k at those clock on your 9700 pro??

432/346 only nets me 19k, and i'm only 100mhz behind you on my cpu.

Mad system power. ;)

I have a knack for getting nice results from less than stellar clockspeeds.

You could do yourself a favor and get Win2k though for a bench partition or drive.

[STEvil]

Originally posted by ricjax99
Erm.... :stick:

1. Your running XP :rolleyes:
2. Your running Cat 3.10
3. He's better at tweaking (no offence :) )
4. You have a Tbred not a barton
5. Hes using better memory timings
6. He has 100 MHz extra CPU.

1 - 100pts diff max.
2 - 4.10, actually. He has 3.1's probably.
3 - What tweaks did he do that I didnt, since you know all the tweaks we both performed?
4 - I was blind yesterday...........
5 - Depends on the board, memory, etc.
6 - uh...



I feel really bright right now... :rolleyes:


Ram precharge is set to auto, which is 1T on the 8rda+, IIRC. If its not.. i'm a dumbass.. :slobber:


As to other points - Picked up a SATA/IDE PCI card today so I can put my 120gb in at the same time as all the rest of my HDD's... maybe i'll stick in a small hdd and do some 2K testing..

[Hallowed]

Originally posted by STEvil
1 - 100pts diff max.
2 - 4.10, actually. He has 3.1's probably.
3 - What tweaks did he do that I didnt, since you know all the tweaks we both performed?
4 - I was blind yesterday...........
5 - Depends on the board, memory, etc.
6 - uh...



I feel really bright right now... :rolleyes:


Ram precharge is set to auto, which is 1T on the 8rda+, IIRC. If its not.. i'm a dumbass.. :slobber:


As to other points - Picked up a SATA/IDE PCI card today so I can put my 120gb in at the same time as all the rest of my HDD's... maybe i'll stick in a small hdd and do some 2K testing..

1. 350 points with the right setup
2. 3.7's and DX 9.0b (slower than 3.5's and 8.1a)
3. He doesn't know all of em. ;)
4. 500+ with extra 256k (check lobbys)
5. NF7-S 2.0 is way stronger per clock than the 8RDA
6. Obvious

:)

I didn't do the impossible to get this score. And that was a bad run, Car Low cost me 200 points alone. :mad:

[STEvil]

Darn NF7 users!!

;)

[ricjax99]

Originally posted by Hallowed
1. 350 points with the right setup
2. 3.7's and DX 9.0b (slower than 3.5's and 8.1a)
3. He doesn't know all of em. ;)
4. 500+ with extra 256k (check lobbys)
5. NF7-S 2.0 is way stronger per clock than the 8RDA
6. Obvious

:)

I didn't do the impossible to get this score. And that was a bad run, Car Low cost me 200 points alone. :mad:

Hallowed, ive tweaked 2K to the max and can't possibly think could have missed anything i could list my tweaks in a pm if you wish? you dont have to tell me what it is, just whether or not im already doing it ;)

You never know you might learn something ;)

The only things i didnt do on my last run was:

Change the running order and disable startup apps (norton, quicktime, msn, ss etc).

I'm also having bandwidth problems since trying out some d21/22 bios variants, since flashing back to d10 and any other "good" bios ive lost 40~60MB/s in sandra :(

I reinstall will solve it from what ive heard :)

STEvil, your 2nd to last post was so off the mark im not even going to bother replying :rolleyes:

[unrealneo]

ric, I'll give ya one over MSN tonight, you will go far with that barton of yours! :slobber:

Spliffs gave me it while on the 1st page Ti200 hunt. :D

[G H Z]

Originally posted by unrealneo
Also have a 9800np, but unfortunately, something is making it overvolt by 1V on the vcore:eek:

And since when is that a problem;)

Hey Fewture, nice one with the CH-5. I couldn't manage anything past 220 @ 2.9v on an NF7. In fact more voltage only made it worse, it really liked 1.7v better.

[Soulburner]

Its a big problem when your card fries itself running 2.7v default I would say.

[Fewture]

Originally posted by G H Z
And since when is that a problem;)

Hey Fewture, nice one with the CH-5. I couldn't manage anything past 220 @ 2.9v on an NF7. In fact more voltage only made it worse, it really liked 1.7v better.
Tnx. The CH-5 sticks don't like superhigh voltage though. My BH-5 had no problem with 3.65V and pulled a sandra bench @ 265 :eek: The CH-5 hits a wall at 255..stil nice sticks though and not very expensive :)

[unrealneo]

Originally posted by Soulburner
Its a big problem when your card fries itself running 2.7v default I would say.

yeah, especially when it's not yours :p: :(

doing some funky stuff. It drops to around 2.05 @ load.:confused:

[ricjax99]

Neo YGPM

[G H Z]

Originally posted by Soulburner
Its a big problem when your card fries itself running 2.7v default I would say.

2.7v is perfect for benching;) ;) ;) ;) ;) ;) ;) ;) ;) ;) ;) ;) ;) ;) -can you see me now?