Command Rate (CR) 1T on Mars Mobo

[LordIT]

I had no difference in the Everest Memory Benchmark between CR1 and CR2 as HDCHOPPER.

Then I did a memory OC:

Setting: FSB 150x10 @3-3-3-9 @ 1,89V
OC with Aegis Panel. CR1 and CR2 were the same: Blue Screen (memory crashed) @ FSB 162

Setting: FSB 150x10 @3-3-3-9 @ 2,2V
OC with Aegis Panel. CR1 and CR2 were the same: Screen frozen (CPU?) @ FSB 182

The 1T definitely isn't 1T then, you should be seeing a few whole ns less on the latency and something between 700 and 900mb/sec increase on the read with 1T, i'm pretty sure.

700-900 MB/s on the read - thats a lot!

[Xello]

I think it actually varies, maybe as low as 200-300mbps on the read and 800mbps~ total bandwidth on Sandra. It's definitely measurable, though

[HDCHOPPER]

The 1T definitely isn't 1T then, you should be seeing a few whole ns less on the latency and something between 700 and 900mb/sec increase on the read with 1T, i'm pretty sure.

I take it Memset is still showing you as 2T even when you set 1T in the bios? I'm pretty certain if memset is showing you as 2T then you really are on 2T.

ya ya memset still showed 2T during all that

was just hopeing that would help all try to figger it out

eyem with all in hopeing 1T "can" work ...if it is or not is beyond me

esp seeing how eye just got some hot rod domminators

[Xello]

I'm on 1T, and memset won't let me change to 2T either, so i'd say this is simply a setting that can't be changed once you're into Windows, any testing you do make sure you reboot and change in bios.

[LordIT]

I'm on 1T, and memset won't let me change to 2T either, so i'd say this is simply a setting that can't be changed once you're into Windows, any testing you do make sure you reboot and change in bios.

Yes, it is a chipset timing as saaya said and the Computer has to turn off that the NB can be reprogrammed. It is the same with the first main memory timing CAS Latency (tCL). Or can anybody change the number (tCL) in MemSet? I dont think so. And my next question is: Why does MemSet have these options though its has to change in BIOS?

[HDCHOPPER]

ya same here guess thats a question for felix

[LordIT]

im not sure about command rate though... i suspect its not detected properly, but im not sure... P35 ddr3 definately supports 1T, so why shouldnt the DDR2 version? ill check with the BIOS engineers

Have you found any new results, meanwhile?

[HDCHOPPER]

ya how about some real 1T action saaya ??

[saaya]

i talked to the engineers and this is actually a tough nut...
there IS 1T for DDR2 operation in the p35/x38/x48 memory controller, BUT, its a different command rate than for DDR3 operation...
its acompletely different register to set, which is weird. well let me explain it.
there arent two memory controllers in the chipset, its one that can run ddr2 and ddr3.
basically ddr3 is an upgrade to ddr2 in that the memory controller needs some extra pins and functions, so you can disable those and have a ddr2 memory controller.
this doesnt mean any memory controller that can handle ddr3 can handle ddr2 as well, but its very very easy to add ddr2 support for a ddr3 controller.

so... its weird that there is a different way of using command rate for ddr2 and ddr3.
the engineers ran some tests and there is indeed a difference between 1t and 2t, but its almost nonexistant... its actually possible that the test results are identical and just hapenned to fluctuate slightly which made the 1t look faster.

so basically... make of that what you want...
p35 doesnt support 1t in ddr2 mode
p35 supports 1t in ddr2 mode but its very slow
p35 2t for ddr2 mode is very fast, almost as fast as 1t...

if this would be our only board atm, or the board we focus on most, then id ask them to dive deeper and see if they can work something out and maybe force the ddr3 1t mode for ddr2 operation somehow...which might be possible.... but i hope you guys understand that x48 and x58 are our focus now and we use our bios engineering resources on that, especially since thats what we are short on atm.

if you guys are interested in this you can use a register mapping tool like baredit from felix88 (author of memset) and create register dumps using 1t and 2t and post it here, and ill see if i can find something myself

[HDCHOPPER]

well thanks for the feedback anyways
if thiers hardly any differance then it's too much work for yall while getting the newer chipsets cleaned up
thanks for looking into it all tho

[LordIT]

Yes, many thanks for your help, saaya.

i talked to the engineers and this is actually a tough nut...
so basically... make of that what you want...
p35 doesnt support 1t in ddr2 mode
p35 supports 1t in ddr2 mode but its very slow
p35 2t for ddr2 mode is very fast, almost as fast as 1t...

These are nice kind of views - all answers are true!

if you guys are interested in this you can use a register mapping tool like baredit from felix88 (author of memset) and create register dumps using 1t and 2t and post it here, and ill see if i can find something myself

If I knew how to do I would do it!

[saaya]

thanks for looking into it all tho

thats what im here for

Yes, many thanks for your help, saaya.
These are nice kind of views - all answers are true!
If I knew how to do I would do it!

here you go
http://www.tweakers.fr/baredit.html

whenever you change something in bios or use a tool to change a voltage or setting etc, it actually changes some register values inside the chip your tweaking. there are registers inside each cpu, gpu, chipset and memory that control how the chips is operating and enable and disable certain features. some of them are locked, like the higher multipliers in cpus for example...

with tools like baredit you can access those registers directly...
this has an advantage and a disadvantage. you can access any and all registers, but you dont know what it does or what you should set it to

back when 90% of the boards didnt have even basic memory timing options people used tools like this to change the trcd and tras settings etc and tweak the performance directly from within windows
i used a tool called wpcredit back then, a veeery old tool, i think it was written sometimes in the 80s

note:
not all registers can be accessed from within windows
not all registers can be changed from within windows
not all registers can be changed at all
many registers have no meaning whatsoever
many registers only change something in combination with other registers, when adjusted individually they do nothing or crash the system when you change them
many registers need to be set to the right setting to work, most settings will crash the system

so this is the most hardcore tweaking there is guys...
its pretty much completely poking in the dark and pure trial and error...
this is actually what bios engineers do at all mainboard manufacturers to tweak the performance of their BIOSes
well, they have some documents from intel etc which guide them, but there are very few informations in those docs that actually help.

usually only nvidia, intel, via, amd etc have the full access to the register settings internally, and they dont share those infos with anybody. which does what, and what setting works etc. but even the engineers who actually designed and work with the chipset often DONT know what combinations of settings works, and how and when it works best etc.
there are almost infitie combinations of settings you can try, and then going cold and increasing voltages changes everything again... plus, like i mentioned, many things depend on each other, so once you figure out one thing, another setting might suddenly work that didnt work before
(lower cas and tras makes lower PL more stable as a basic example)

tho if your smart you try different settings you know work (boot with 1t, boot with 2t, boot with 333 strap, boot with 266 strap) and then read the register values and see what actually changed

there are several people doing this and creating their own "maps" of the register settings, noting which does what, when this setting is set to this or that, how it impacts performance, what other settings work for this register, what side efeccts and relation to other registers it has etc etc.

I was hoping to do this and maybe get some of you to do this as well, so we can build a more complete map of register tweaks and share them between all of us and make it public. so far there are only very few people tweaking like this and their results are usually not shared.

PS, this is a useful tool to compare register dumps and highlight the differences between 2 dumps
http://www.download.com/ExamDiff/300...dlPid=10619470

[LordIT]

Thanks saaya - thats a big adventure!

I installed baredit and selected Selection By Memory Adress; pressed Mem.txt to save:


Two different register text files can be compared in ExamDiff:


The problem is that every single action in windows cause a change in register. Only a restart of baredit causes changes - partly in different lines. I guess it is easier in DOS isnt it?

I startet my computer and my first action was to a save register file. I did that three times without changing anything in BIOS- but had 6-8 differences in bareedit - partly in different lines.

Imagine we could find some register lines for CR 1T. What is the second step?

[saaya]

Thanks saaya - thats a big adventure!

yes

hmmmm well im not sure why its changing all the time...
let me check with felix88

now you can compare the register dumps with exam dif and note the differences. if i were you id open an excel table and would note down all the registers that change. from left to right, which register, then what setting and name the top column to what configuration in BIOS you had for this particular dump. then try to change those registers manually one by one and see if they affect performance.

[LordIT]

What does happen if I pressed the "write" button? Do I flash into the NB directly. Can the NB be killed by using wrong numbers? Or is it only a change in windows registry that will be deleted after restarting?

[HDCHOPPER]

careful thier ... bet it writes it into the bios

[saaya]

What does happen if I pressed the "write" button? Do I flash into the NB directly. Can the NB be killed by using wrong numbers? Or is it only a change in windows registry that will be deleted after restarting?

you directly write into the registers in the chipset afaik.
after a reboot its reset to normal... so nothing to worry about

mmhhhh it might be possible to kill the chipset by writing the wrong stuff to the wrong register... but i find that very hard to believe actually...
its really safe, i havent heard of anybody killing their board this way...

[LordIT]

hmmmm well im not sure why its changing all the time...
let me check with felix88

Does felix88 know?

[aoch88]

I can't get it working @ 1T. Same issue here, it will always go back to 2T.

[LordIT]

There is one EPP profile enabled in the Muskin memory:


I changed from 2T to 1T in EPP profile:


Adjusted the EPP settings in BIOS (multipler 4e) but couldnt get 1T:


Enabled a second EPP profile with SPD-tool which is 1T basically. But I think its not a real profile because of the funny timings:


Could I get it with a real 1T EPP profile for DDR-2? Does anybody have it?