Originally Posted by fat4l @gupsterg
, when you were doing my mod, did u see if there's any difference between my ares 3 MC timings and normal 290X MC timings? I'm wondering if asus made som tweaks to super uber dabest Ares 3
No, as I can't tell where the MC timings are
In post 1331
I placed an image, left side my 290X AFR, right side 390X AJR.Here it is (Click to show)
Now the top red section differs as part of that is a) table length b) revision c) pointers to sections after that are 4 empty bytes, then RAM IC setup start. Next blue & pink segments are identical, green section differs as they are RAM straps/timings. Aqua section differs at the end, also see the note that if there are 2 RAM ICs what happens there. Next the purple section is identical, its even identical in a VRAM_Info with 2 RAM ICs or even 3 like this ROM
SO in my educated guess I thought 390 MC timings are in the aqua section, I copied that into my VRAM_Info table, this was what I called v1.0 390MC test. This didn't give a boost in performance or stability or cause instability.SO then I did v2.0, in spoiler is info on that plus v0.0 test, etc (Click to show)
Originally Posted by gupsterg
v0.0 circa Oct 2015 (link to posts
I copied complete 390/X VRAM_Info (Hynix AJR) into my ROM. Added stock Hynix AFR timings and only
adjusted supported memory size (80h to 40h) (not aware of density at the time). This basically was unstable at even RAM frequencies that stock factory ROM would attain, so I never checked performance. As Stilt shed no further info when I asked in linked post/thread, I abandoned project.
v1.0 circa 5 days ago (link to post
Here I only copied what I thought is 390 MC timings into my stock AFR VRAM_Info. I saw no performance gain or stability gain.
v2.0 circa 3 days ago (no link as conducted off forum)
Here I modified 390/X VRAM_Info (Hynix AJR).
a) 80h to 40h for RAM size.
b) 53h to 43h RAM density.
c) added AFR stock timings.
d) changed RAM IC ID ie text length 0B.
This I've seen consistent small performance gain, I don't think yet gained further stability, but there is no instability gaming (4hrs) or F@H (16hrs+) (1100/1525 or 1100/1575).
Due to the testing I did aka v0.0 vs v1.0 vs v2.0 MC Timings are either in the red section or red and aqua section.
Originally Posted by fat4l
What bout the frequency ? U remember my card has stock mc clock of 1350MHz instead of 1250Mhz(or something like that). Do u think this has any influence on my high stock mem clock capability ?
No influence, originally I thought it set MC frequency. As when I compared stock ROMs with factory OC'd ROMs that frequency was matched to highest RAM freq. in a ROM. Even if it was out of sync in testing there was no drop or gain in performance
, so that section in post 1 was written on basis to make like the "best practice" I was seeing in ROMs.
It was only when I was querying The Stilt about MLU PowerPlay, he revealed PowerPlay can have more states of VDDCI for RAM frequency
. At that point I learned all the frequency did was be a link to what VDDCI to use. I did a mod of 2 & 3 states VDDCI and it failed, the Stilt then presented some pointers which I had to decrypt, this taught me about PowerPlay directory structure. OP post has been updated with new info, only got to add 2 or 3 state VDDCI/Frequency mod info there now (successfully done
You can run Stilts timings at 1500MHz? I was getting artifacts at desktop after about 1440 ish IIRC.
I wouldn't know how yet, you see we could make the red section in image above to be appropriate for 2 or even 3 RAM IC support but
the aqua section I wouldn't know how to tackle (haven't really spent much time on it, as generally been busy
Cheers for share
BTW your ROM is showing as 8GB in 3dmark results, have you got a 8GB card or you didn't change supported RAM size and density on your mod of 390 MC?