CAS 1.5 is doubt to work on A64 based systems......

[HARDCORECLOCKER]

According to this thread post #87 I feel shocked:

http://www.xtremesystems.org/forums/...t=66721&page=4

CAS 1.5 is not working on a64 based systems ?

Can't believe it - if it is a fake each CAS2 RAM would run 1.5......

So guys what do You think about it?

[HousERaT]

Well technically AMD themself says their cpus don't support CAS 1.5. All I can say is supported or not, CAS 1.5 seems to work fine on my system. Some people claim that they get better system performance & stability with CAS 1.5. I will neither confirm nor deny this statement.

[HARDCORECLOCKER]

Well technically AMD themself says their cpus don't support CAS 1.5. All I can say is supported or not, CAS 1.5 seems to work fine on my system. Some people claim that they get better system performance & stability with CAS 1.5. I will neither confirm nor deny this statement.

Nice statement - let's see what other experts have to say........

[bigjohns97]

It comes from a good source

In a way it shows that a64tweaker just reads what the bios sets.

I guess what the memory controller intrepets is cas2.

You also notice it knocks out a bunch of values for other timings

CAS# Latency (Tcl)—Bits 2–0. Specifies the CAS#-to-read-data-valid.
000b = reserved
001b = CL=2
010b = CL=3
011b = reserved
100b = reserved
101b = CL=2.5
110b = reserved
111b = reserved

Row Cycle Time (Trc)—Bits 7–4. RAS#-active to RAS#-active or auto refresh of the same bank.
Typically ~70 ns. These bits are encoded as follows (only 7–13 are valid; all others are
reserved):
0000b = 7 bus clocks
0001b = 8 bus clocks
. . .
1110b = 21 bus clocks
1111b = 22 bus clocks

Row Refresh Cycle Time (Trfc)—Bits 11–8. Auto-refresh-active to RAS#-active or RAS# autorefresh.
0000b = 9 bus clocks
0001b = 10 bus clocks
...
1110b = 23 bus clocks
1111b = 24 bus clocks

RAS#-active to CAS#-read/write Delay (Trcd)—Bits 14–12. Specifies the RAS#-active to CAS#-
read/write delay to the same bank.
000b = reserved
001b = reserved
010b = 2 bus clocks
011b = 3 bus clocks
100b = 4 bus clocks
101b = 5 bus clocks
110b = 6 bus clocks
111b = reserved

[Kunaak]

cas 1.5 is a placebo.
it doesn't work, but you think it does.
truth is, A64's don't support latencies below 2.
its just creative bios tricks on Oskars part, that can give a sense of what to expect from cas 1.5 or so... kinda like a "simulated 1.5" haha...

I've known this for sometime from a friend online, however, this is the first time I really looked at actual data to support that. interesting.

[guess2098]

i hate to tell you guys about this
but in my taiwanese forums, we already found out that CL1.5 is slower than CL2

and even CL1.5 works find, doesn't mean CL2 work in same speed

[HARDCORECLOCKER]

i hate to tell you guys about this
but in my taiwanese forums, we already found out that CL1.5 is slower than CL2

and even CL1.5 works find, doesn't mean CL2 work in same speed

Do You have any screenies for compario ?

[bigjohns97]

If this is true and the memory controller on the chip does have to convert values from the bios to the "supported" memory timings this should be a sticky of what memory timings to set in the bios.

Could even edit the bios before loading to to only allow these values

[guess2098]

Do You have any screenies for compario ?

cuz i am not the one who post the pic, i can't just copy and post in here
but i can give you the URL of that thread

http://www.coolaler.com/ipb/index.php?showtopic=55053

[HARDCORECLOCKER]

O.K. - this is gettin' hot - can someone make it a sticky ?

[dnottis]

I tried Cas 1.5 and found while it was Memtest stable, I had issues with games - CTD, reboots... and Cas 1.5 in Sandra wasn't any faster.

I too feel it's probably a placebo, not working - and probably makes things unstable anyways...

[guess2098]

I tried Cas 1.5 and found while it was Memtest stable, I had issues with games - CTD, reboots... and Cas 1.5 in Sandra wasn't any faster.

I too feel it's probably a placebo, not working - and probably makes things unstable anyways...

some how i think 1.5 is more stable than 2 under same speed^^"

[Agent-JCDenton]

I posted this when I came across Geil's upcoming Ultra-X memory on Anandtech - Rated at DDR400 1.5-2-2-5:

"The new DFI nForce4 boards support CAS 1.5 as a BIOS choice, and possibly a few other high-end boards. CAS 1.5 sounds impressive, but AMD tells us that they do not support CAS 1.5 in their on-chip memory controller." - Anandtech

So CAS 1.5 is not of much use , not for the time being anyway on AMD, but I can't help wondering if/when they'll support it. I sure hope Geil know what they are doing tho.

[G H Z]

Hmm from my early testing 1.5 was faster than 2. I'll have to revisit this and verify that. But yes according to the AMD .pdf, 1.5 is nothing.



Still, I doubt that Oskar would option that into a BIOS for "simulated 1.5"

[guess2098]

well~~ i am thinking that.... if K8 can;t pick up 1.5... then what value does k8 use when we place at 1.5.....

[HARDCORECLOCKER]

So maybe OSKAR himself knows somethin' about it.........?

[TheMeatFrog]

When I use 1.5 it's a more stable for me. So I use just 2 unless I'm really pushing my ram to the limits.

[Eldonko]

Geil uses the lower CAS values as a marketing gimmick, leading people to believe they have come up with some new tech that runs at cas1.5. Supposedly these sticks can even run at cas1.0, maybe they should read some white papers. Link

[Sucka]

CAS 1.5 sure does look a lot cooler than CAS 2

I've found 1.5 actually does nothing for me, good or bad.

[s7e9h3n]

HCC - Why don't you just run a test to put this issue to rest. Your Corsair can run cas 1.5, so how about a comparison of your own?

[Tony]

Cas1.5 is not supported, its not setting cas1.5 and its no faster than cas2. OCZ ran tests months ago to see if it was worth releasing VX at cas1.5, upon advice we decided not to as its not actually cas1.5 being set.

[haPpydUde]

Some vx-4000 I bought of opp a little while ago prefers cas 1.5 by over 10mhz, on my 3700+ sandy. So cas 1.5 is no myth in my mind.

[TheMeatFrog]

Just did a little test run using SuperPI 1M and Sandra Bandwidth.

CPU @ 2.75GHz
Mem @ 250Mhz

CAS 2.0
SuperPI 1M
1st Run - 30.922
2nd Run - 30.965
3rd Run - 30.797
Sandra Bandwidth
Int - 7218
Float - 7142

CAS 1.5
SuperPI 1M
1st Run - 30.921
2nd Run - 30.950
3rd Run - 30.797
Sandra Bandwidth
Int - 7218
Float - 7140

Going by these results imo they are the same as far as performance goes. I still think that in some ram 1.5 adds stability some how.

[s7e9h3n]

Some vx-4000 I bought of opp a little while ago prefers cas 1.5 by over 10mhz, on my 3700+ sandy. So cas 1.5 is no myth in my mind.

But if cas 1.5 was truly a reality, your memory should clock lower than cl2

[haPpydUde]

Mabye but some ram just likes low latts. And I to noticed that the preformance was the same pretty much saw some improvments in some 3d programs. Nothing major, but the fact that I can clock 10mhz highier because of it is why I like it.