New Memory Tweaker for Intel Chipsets

Quote:

---

Originally Posted by FELIX

...yes I'm very busy at this time (and probably for a long time :) )
but I try to take some time for make a SD update, simply the
Intel datasheet report nothing about memory timings and other setting. :mad:
and I can't do nothing without it.

---

Since we all know how it is to be busy, and we all appreciate the what you've provided us (regardless of if they express it)!!

I think this is the Datasheet you want (fullll of registers info!):
http://download.intel.com/design/pro...hts/324642.pdf (~4.8mb, 290 pages)
EDIT: Found an update as well which might be of relevance... One change in the update PDF states for one of the many Erratas listed: "Changing the Memory Type for an In-Use Page Translation May Lead to Memory-
Ordering Violations"
ftp://download.intel.com/design/proc...pdt/324643.pdf

I found this one from google, and I assume it is the same one you looked at:
http://download.intel.com/design/pro...hts/324641.pdf

THEN I found ftp://download.intel.com and went to the "datashts" directory, sorted by date, and checked out the first one. It had a ton of info, but was for different chip :( So I checked the one below it, which was ONE PDF number different, and it had to do with the same CPU... So I figured "why not, I'll see what "324642.pdf" comes up with in the link. Sure enough!

Quote:

---

2nd Generation Intel® Core™
Processor Family Desktop
Datasheet – Volume 2
Supporting Intel® Core™ i7, i5 and i3 Desktop Processor Series

---

I don't have a Sandy Bridge, and no current plans for one, but maybe my dreams for AMD changes will come true :D

So yea, hopefully that's the one you're after! :up:

I dont know if this has ben asked before, I've searched in the last pages and found nothing:

Is there any plans for atom support?

I'm planning in purchasing a D510MO from intel and would love to be able to tweak and play arround with it.

Is memset out for sandybridge yet does anyone know!

Quote:

---

Originally Posted by FELIX

...yes I'm very busy at this time (and probably for a long time :) )
but I try to take some time for make a SD update, simply the
Intel datasheet report nothing about memory timings and other setting. :mad:
and I can't do nothing without it.

---

Not yet.. He is currently very busy ;)

hey guys, is there any update for the cpu tweaker that supports SNB?

I don't know about the other boards but on ASUS P8P67 EVO the timing registers appear locked by BIOS so that having access to them isn't going to help. :(

Hi,

FELIX could you take a look at this? There seem to be a problem with XMP profile2 in the SPD.
Timings set by bios correctly at post #13.
CPU Tweaker SPD at post #16.
CPU-Z SPD(also has a bug?) at post#18.
Link to CPU-Z test dump at post #20.
Thaiphoon XMP profiles at post #21.

EDIT: apparently these modules conform to XMP Rev. 1.2, which i can't find the specs for.
EDIT2: as far as I could find, bytes 210 and 245 defibe QPI voltage and the rules are the same as for VDD(bytes 185 and 220) in case you decide to add this to the SPD window.

WTB CPU-Tweaker for AMD Fusion APU :( Even if only for the memory timings, and not overclocking!

I see Asus are now allowing changing of mem timings by software for the MaximusIV. Hope the other SNB Asus boards are going to follow.

As you can see in the SS, H67 timings are locked (greyed out). Probably be the same for other Asus P67 too.

Thanks for these ones!

I first want to say thank you for a nice piece of software!

I would like to ask for Arrandale support on multi changing, when you have time to add it to cpu tweaker.

Quote:

---

Originally Posted by some_one

I see Asus are now allowing changing of mem timings by software for the MaximusIV. Hope the other SNB Asus boards are going to follow.

As you can see in the SS, H67 timings are locked (greyed out). Probably be the same for other Asus P67 too.

---

will the timing changes revert back after reboot? hmm.. gotta try this

hello everyone,
I've a compaq presario V5009EA AMD Mobile Semprom 3000+ with 512 Mo ram DDR,i run windows xp pro.
I've used MemSet to configure my ram like this:
http://i40.servimg.com/u/f40/13/70/09/73/memset10.jpg
is this good or somebody can suggest me a better configuration?
Thanks

Quote:

---

Originally Posted by magikstar29

hello everyone,
I've a compaq presario V5009EA AMD Mobile Semprom 3000+ with 512 Mo de ram DDR,i run windows xp pro.
I've used MemSet to configure my ram like this:
http://i40.servimg.com/u/f40/13/70/09/73/memset10.jpg
is this good or somebody can suggest me a better configuration?
Thanks

---

Those primary timings (top group) are pretty dang good (depending on what brand a model of DDR you have). If that Sempron has a K8 memory controller, then dropping that Command Rate to 1T should REALLY boost your scores. My Socket 939 K8 I was getting 7000mb/s in Sandra, which decimated my friend's AM2 system with DDR2 lol

TEST EACH SETTING ONE AT A TIME. Use something like Sandra Memory Bandwidth or MaxxMEM2 and run it BEFORE any changes for a baseline, then use it by comparing to find out if there are any performance gains after each change. If you want to really make sure things work then I recommend getting IntelBurnTest and set it to use as much memory as you can (might have to do Custom and enter just below what is "free), to stress your system for stability.
[Caution: IntelBurnTest, or any other stress test, will also greatly load up your CPU. This can cause it to get quite hot (more than typical use), so it might be good to take the side panel off, if it's a desktop.]
My suggestions:

1. Lower Command Rate to 1T
2. Rule of thumb is tRP + tRAS = tRC. So following that, your tRC of 7 should become 4.
3. If you can drop that tRFC down too, that'd help.
4. Refresh Mode Select set to either 7.8us or 15.6us (caution: systems I've tested newer than Soc939 may not like 15.6us)
5. Try changing your Idle Cycle Limit to 16, even 32
6. Read/Write Queue Bypass, crank it up as high as it can go (8x if I'm remembering?)
7. Queue Bypass Max, also run it at the highest setting

When setting the tRC and tRFC, make sure to drop it one number at a time (testing each time), otherwise if you jump from 7 to 4 it may crash. So be aware that most of these can (and will) likely cause a system crash. I'm sure you've experienced them already through testing, but if not, just don't be surprised lol Biggest thing that gave me performance (again on Socket 939 K8s) was the Command Rate, so I'd focus on that. May need a bump in voltage if you haven't already :)

Good luck!

original (default) settings:

CAS# Latency (CL) = 2,5
RAS# to CAS# Delay (tRCD) = 3
RAS# Precharge (tRP) = 3
Precharge Delay (tRAS) = 7
Command Rate = 2T
Row Cycle Time (tRC) = 10
Refresh Cycle Time (tRFC) = 12
Ras# to Ras# Delay (tRRD) = 2
Refresh Mode Select = 7,8us
Idle Cycle Limit = 16
Write Recovery time (tWR) = 3
Write to Read Delay (tWTR) = 1
Read to Write Delay (tRTW) = 3
Read/Write Queue Bypass = 16x
Read Preamble = 6,0ns
Max. Asynchrone Latency = 6ns
Queue Bypass Max = 7x


tRC min value is 7
tRFC min value is 9
i can't set under this values

Refresh Mode Select -> 15.6 work it's the highest value
Idle Cycle Limit -> 16 or 32 work (i also have 0, 4, 8, 64, 128 or 256 value to set)
Read/Write Queue Bypass -> 16x by default (and is the highest value)
Queue Bypass Max -> 7x by default (and is the highest value)

you said tRP + tRAS = tRC it's bad if not?

i can change the settings and test settings with Sandra Memory Bandwidt without rebooting or i must reboot before test?

For Idle Cycle Limit i've read http://www.techarp.com/showfreebog.a...ng=0&bogno=440 it said :
"For general desktop use, it is recommended that you set this feature to 8T. It is important to keep the pages open for some time, to improve the chance of page hits. Yet, they should not be kept open too long as temporal locality dictates that the longer a page is kept idle, the less likely the next data request will require data from it.

For applications (i.e. servers) that perform a lot of random accesses, it is advisable that you select 0T as subsequent data requests would most likely be fulfilled by pages other than the ones currently open. Closing those open pages will force the bank to precharge earlier, allowing faster accesses to the other pages for the next data request. There's also the added benefit of increased data integrity due to more frequent refreshes."

-----------------------------------------------------------------------------------------------------------------------------

edit:

Sandra Memory Bandwidt is not reliable, for the same values​​, I have never the same results

Man. I don't know why I didn't think about it before, but your best option is to use A64Info. If you're using a S754 or S939 Sempron, that will work great (though intended for S939), in desktops or laptops! That is the only Athlon 64 tweaking utility I'll use since it does what MemSet can, but multiple things it also can not. :)

To address your questions anyways:

Quote:

---

Originally Posted by magikstar29

tRC min value is 7
tRFC min value is 9
i can't set under this values
~~~~
you said tRP + tRAS = tRC it's bad if not?

---

If you can't drop it lower than 7 and 9, then that's alright. It's not "bad", but it may hurt performance a little. How much? Not sure, sorry :\

Quote:

---

Refresh Mode Select -> 15.6 work it's the highest value
Idle Cycle Limit -> 16 or 32 work (i also have 0, 4, 8, 64, 128 or 256 value to set)
Read/Write Queue Bypass -> 16x by default (and is the highest value)
Queue Bypass Max -> 7x by default (and is the highest value)

---

I remember testing the different Refresh Mode Select and Idle Cycle Limit settings, and 15.6us seemed to work the best. Whether 3.9, 7.8 or 15.6 is the optimal choice is dependent on your memory speed, and you'll just have to test and compare since I don't remember the formula to find that out (sorry). Idle Cycle limit seemed to vary, but 16 or 32 usually worked best. The Queue Bypass options I always lost performance setting them any lower, so they were never changed on my system.

If you end up using A64Info, there is a setting that replaces Refresh Mode Select, and offers a lot finer tuning (which is good). If you end up using that I'll try and get the numbers to try, as there are a lot in the list :P

Quote:

---

i can change the settings and test settings with Sandra Memory Bandwidt without rebooting or i must reboot before test?

---

If you restart the computer, the settings default back to the BIOS configured settings. So you need to Apply them, then save the configuration to a file, and tell MemTest to apply them at startup. Be aware though, I have encountered a time or two where my computer didn't like having them ALL changed at once :P I think it was Command Rate that needed to be done by itself, but it's been such a long time :\

Quote:

---

Sandra Memory Bandwidt is not reliable, for the same values​​, I have never the same results

---

It is reliable ;) Most programs that test memory in real time and report fairly detailed results will have a margin of error. Sandra and MaxxMEM both have it. For Sandra it's usually ±10-20MB/s but can be upwards of 100MB/s, and MaxxMEM seems to be pretty much the same as well. If you run them after a fresh restart of Windows, and give the system like 5minutes to completely settle down (have everything loaded), then you should hit more consistent numbers. They'll still be different, but the variance shouldn't be as high. But yes, it can prove to be a difficult thing when trying to figure out if it was a gain in performance or just margin of error. I usually didn't consider it a worthwhile gain unless it was 100MB/s or more, both because it let me know it was likely above the margin of error, and also due to me not feeling 15MB/s increase being worth the system becoming potentially unstable :P


Hope that helps more, and again, I recommend giving A64Info a try from my link above (uploaded myself to my personal webspace, since that latest version isn't as easy to find).

Some time ago, while I was playing with R3E, I thought CPU Tweaker was broken for not showing right value for tWR.
No matter which value below 12 is set, it showed 12.

Now, with OC Orange, I see I was wrong. It shows changes in tWR from 14 to 12 to 10 and even to 8.

Obviously R3E wouldn't let me set tWR lower then 12.

I hope this helps other R3E owners.

My apologies to FELIX, for thinking CPU Tweaker is broken.

Long time no news from Felix.
Are the application EOL or....?

I hope not. It was my main tool since golden days of Athlon64, when the fun started.

But he never claimed any copyright on CPU Tweaker and Asus made their own memory tweaker, which is widely used.

Maybe FELIX is just indisposed. If it helps, I would use CPU-Tweaker just for being the first and cause it reminds me of "good old days"

Quote:

---

Originally Posted by donmarkoni

Asus made their own memory tweaker, which is widely used.

---

:confused: AFAIK MemTweakIt was made to only allow adjustments on the Asus flagship boards, for other boards it's just a display not a tweaker or is that not the case?

Any new versions?

Hoping for update @ X79 now!

Pm coming, hopefully you have a little spare time to look at a CPU-z dump and see if X79 is unlocked, and if so hope you have a little time to push out an X79 update sometime soon.

Hi Xtreme Forumers !
Where can I download the last Memset version that will run on Windows 98 SE please ? This is on an old DELL Dimension 4100, i815 chipset. I could find links only to dnld the latest and (?)greatest, but that is not for me :-(

CPUID has replacement for X79 and Ivy Brigde with CPUID Mem TweakIt 1.01.6:
http://www.mediafire.com/?enyauvt2fe75v1n

Quote:

---

Originally Posted by stasio

CPUID has replacement for X79 and Ivy Brigde with CPUID Mem TweakIt 1.01.6:
http://www.mediafire.com/?enyauvt2fe75v1n

---

thanks man :up: