I'm not following you.
Real 1:1 divider (DDR3-400) could actually be useful. It's proven many times before that the high memory frequency is holding back the HTT clock.
Printable View
I'm not following you.
Real 1:1 divider (DDR3-400) could actually be useful. It's proven many times before that the high memory frequency is holding back the HTT clock.
I think if you choose 400 on ddr III it shows as 2/1?Quote:
Originally Posted by massman
Not good with ratios. But at 350 HT and lowest ram divider 1400 ram is not what I would call high memory frequency.
The problem is related to memory IMO but not high mem frequency unless you consider 1400 ish "high"
Try the lowest ram divider you will see what i'm talking about.
Internal latencies are changing is my guess with the lowest ram divider on ddr 3 as well as some other things.
Work in progress not saying it can be done but it can't be done unless we try :)
Brian, you are absolutely correct. When using the 400 divider, 1600 Mem and 400HTT, CPUz would show memory at 1/2. Massmans mention of 1/1 would mean that you'd have to run 800Mhz HTT on the 400 divider to be at 1/1 with the memory at 1600Mhz. That won't be happening anytime soon without a miracle:DQuote:
Originally Posted by chew*
Edit: Confusion may be that 400HT is DDR, effectively 800Mhz ..... HT 800 Mem 1600 - Memory is twice HT speed
-------------------------------------------------------------------- 1-----/----2
You are definitely pressing in the right direction my friend and you have many diligantly watching your progress.
best regards,
Randi:D
I grew up on cheap chips randi etc the ones that were locked becasue thats what I could afford. I'd like the to see those with locked chips have boat loads of fun as well Plus i find dealing with a fixxed multi more challenging. Old habits die hard :p:Quote:
Originally Posted by MrIcee
Lowest divider for the DDR3 is indeed 1:2, or "400". Set 400HTT and you get DDR3-1600 speeds.
The problem is the combination of high HTT and high memory. DDR3-1400 isn't high, no, but then again the limitation isn't at 350HTT either. Think about trying to reach HTT frequencies of 450+ ... with DDR2 you can just run 450MHz memory, but with DDR3 it's 900MHz.
We've seen this on the previous generation of hardware ... the board (memory type) limits the HTT overclockability of the CPU.
Yah and thats what we are trying to get beyond.Quote:
Originally Posted by massman
Depends on board but try it "booting" no software....most are knocking on a wall right at 1400ish ram, 349,350, 351 in that area for HT.
d5 post code hang is not HT/HTT related.
Hi guys,
Yesterday I flashed 0801 bios. Since I have 1,5Mhz swing HT speed. Somebody, know how can we solve this? or I need back 0707 bios..
await a new bios mate... downflashing is not an option...
Care to share some settings/voltages with us? :up: would be fun to test it out...Quote:
Originally Posted by bingo13
any one try downgrade bios with afudos 236ES
afudos (bios.rom) /p /b /n /c
damn it :( I would like to bench with LN2, because I need back 1090T tomorrow.Quote:
Originally Posted by Leeghoofd
So, if I use 26-27x ratio, cpu swing is 40Mhz :S
I also want to know.Quote:
Originally Posted by Hazzan
Well, that's exactly what's happening...Quote:
Originally Posted by Leeghoofd
http://www.tech-caffe.com/gfx/cpu_nb.jpg
I had it set to 1.200000 and it set itself to 1.052 on the first boot. Meanwhile i did what *chew described (save profile/load defaults/reload profile), hope somehow it fixes it.
Have you got Auto Load Lane calibration Enabled or Disabled in that tab NaMcO?
I haven't had this issue on my system, but I'm running it Enabled since day 1.
Interesting stuff ...
No go. It can't be downgradedQuote:
Originally Posted by Hazzan
Both LLCs are enabled :shrug:Quote:
Originally Posted by Lightman
Ok, now I know what you're talking about.Quote:
Originally Posted by chew*
I'm not sure if the inability to boot at 350HTT and the lower max HTT on DDR3-based motherboards are different problems or have a shared underlying cause. What I do notice is that changes in the memory setup (single or dual, different memory slots) change the maximum HTT frequency I can post at.
What do you have in mind as cause of the problem?
I was having HT ref swings, not able to raise mulit, well it would raise then in windows, regaurdless of which one I loaded into, it would read as if I didn't change it and some other issues. While in the bios, I hit that refresh button on the back near the PS2 ports. I have been running 13.5x305 HT Ref for over 24hrs now with no swing on HT Ref. I can raise multi now and everything seems to be normal again.
Regarding the HTT swing. I only experience this in idle (365MHz to 368MHz) and only see this with CPU-Z. Leeghoofd send me the THG_clock application and that shows no swing at all.
Question is if it's a hardware issue or a CPU-Z issue.
Sorry to be off topic,but what ram slots are you guys using,ie blue or black,reason i'm asking is with my old GD70,it preffered the blue slots,furthest away from the cpu socket.Just curious.:)
I have same effect on AOD. I think bios issue, because I experienced while I flashed 0801.Quote:
Originally Posted by massman
Blue?? You mean red or black. :D I use red and unganged mode.Quote:
Originally Posted by moorhen2
Yes sorry,i meant red/black.Thanks.:)Quote:
Originally Posted by over@locker886
MRL tables at very low ram dividers and or internal latencies changing running things very tight.Quote:
Originally Posted by massman
Very similar to why we see better results on some boards with 1333 divider at 270htt versus 1600 at 225htt. aka 1800 ram but diff methods to get there.
No reason since I doubt most are running 2/1 ram divider default for 24/7 with that type of ram why some things can't be slacked a tad. It has to be memory issues imo becasue I peg D5 and the easiest way for me to peg D5 is tick off my ram.I can raise HT to some retarded speed and it will never peg d5 on port 80. I can also crank it higher that 350 with 1066 divider but once again i hit a wall there to depending on quality of ram and or "IC's".....but ram has to be uber. I want to make high HT clocking for the "AVG" guys without uber 2000 ram easy, so 2/1 divider is the one to tune.
This is also going to cause a validation issue to say the least.Quote:
Originally Posted by massman
With the swings i've seen 7 gig on ln2 with a chip that "really" does 6936 on lhe is possible now ;) No diff than intel floating BCLK atm.
I've seen larger swings than 2mhz HT the higher i'm clocked and the less stable i am. So based on this are we to accept "bugged" cpu-z valids?
Certain people have been able to valid lets say exactly 3ht over what there prior max was right on the buitton ( exactly what you saw )...I did not attempt to as I figured it was a bug seeing as I have been using the THG tool all along.
Not only with CPUz Brian, but I've had 2 of my best 3DMark scores not successfully upload because of the FSB/HT fluctuation. When it checks your system between tests, if there is any change of parameters whatsoever it will invalidate your score and it won't submit, I'm sure you've probably had that happen over the years with other boards and setups.Quote:
Originally Posted by chew*
Best regards,
Randi:D