A-data DDR600 test report by socket939 fx53

JEDEC???

Yeah but i meant like when SD went to DDR... Some boards had DDR and SD slots...

Isnt it the same case with the new socket??? Cause DDR2 has the most insane timings...

hm, yes, but afaik there isn't a Dual-RAM-board.....dunno if someone is gonna make one :confused:

but, u can forget DDR2 until DDR2-667 cames out....the speed will compensate the lazy timings....

Quote:

---

Originally Posted by qk4722

http://61.220.123.253/cpg/albums/use...10004/3003.jpg

---

No DC@300MHz at all...look at the multi. ;)

But 6600 still is a bit low, indeed...

Greetz, Flox

but a low multi doesnt matter on a64's afaik, no?

pazifist, ther ARE dual ram boards that have ddr and ddr2 slots.

and mini, its not about speed, its about memory bandwidth. to have memory that runs 1:1 with the fsb at 266 is nice, but intel chipsets have great memory dividers and dont need 1:1, what the P4 need is enough memory bandwidth, not high mhz speeds on the memory.

ddr2 is intel at its best again, trying to f00l customers once again with the mhz myth, this time not with cpu mhz but with memory mhz. :stick:

@ Flox: wie meinste das?

wegen dem ungeraden mutli oder wie?

Quote:

---

Originally Posted by saaya

what the P4 need is enough memory bandwidth, not high mhz speeds on the memory.

---

high mhz results in high bandwith...

Quote:

---

Originally Posted by maxx-p

high mhz results in high bandwith...

---

thats just as wrong as intels old slogan "high mhz results in high performence"

:P

eb at 2.5-2-2 and 200mhz beats 250mhz with 3.0-4-4 :) mhz isnt everyhting, same that goes fro cpus goes for memory and also videocards :)

performence=work/time

so if you have a really fast speed time will get really low, but if the work the hardware does at one time interval is low the result will still be bad.

Quote:

---

Originally Posted by saaya

but a low multi doesnt matter on a64's afaik, no?

pazifist, ther ARE dual ram boards that have ddr and ddr2 slots.

and mini, its not about speed, its about memory bandwidth. to have memory that runs 1:1 with the fsb at 266 is nice, but intel chipsets have great memory dividers and dont need 1:1, what the P4 need is enough memory bandwidth, not high mhz speeds on the memory.

ddr2 is intel at its best again, trying to f00l customers once again with the mhz myth, this time not with cpu mhz but with memory mhz. :stick:

---

A the lower the multi the lower the mem bandwidth on A64

Set 200 htt and keep it their, then run sandra @ 7x..then 8x...then 9x.....ect ect. You'll notice a difference every multi bump

hm....kk, i'll check it with ClockGen and 250MHz.....will set all multis from 5x to 10x....w/ .5x-multis too :)

so, just w8 30mins, after i recorded my Minidisc ;)

Quote:

---

Originally Posted by saaya

thats just as wrong as intels old slogan "high mhz results in high performence"

:P

eb at 2.5-2-2 and 200mhz beats 250mhz with 3.0-4-4 :) mhz isnt everyhting, same that goes fro cpus goes for memory and also videocards :)

performence=work/time

so if you have a really fast speed time will get really low, but if the work the hardware does at one time interval is low the result will still be bad.

---

What??? So you are saying that high Mhz doesnt mean high bandwidth??? Comeon...

Sandra gives 250Mhz 3-4-4-8 more than 200Mhz 2-2-2-5... Same goes for games...

The only thing ive seen where timings matter are benchmarks???

What program is the best to test memmory bandwidth???

Timings dont mean bandwidth... They mean response time... It gives very little bandwidth... While Mhz gives bandwidth... But DDR2 just sux...

Or am i just wrong???

blue, the memory badwith gets bigger because the cpu is clocked higher afaik. we are talking about buffered memory bandwidth wich depedns on the cpu speed (speed of the mxx and see units)

unbuffered there should be almost no difference in memory bandwidth with different multi and resulting different cpu speed.

and even if the 6600 have to do with the low multi, its not going to be 7500 with 250mhz higher cpu speed and a higher multi :P

and yes mini, memory bandwidth depends on the timings as well. correct me if im wrong.

ok, here we go:

http://www.8ung.at/dresign/5x250.jpg

http://www.8ung.at/dresign/5.5x250.jpg

http://www.8ung.at/dresign/6x250.jpg

http://www.8ung.at/dresign/6.5x250.jpg

http://www.8ung.at/dresign/7x250.jpg

http://www.8ung.at/dresign/7.5x250.jpg

http://www.8ung.at/dresign/8x250.jpg

http://www.8ung.at/dresign/8.5x250.jpg

http://www.8ung.at/dresign/9x250.jpg

http://www.8ung.at/dresign/9.5x250.jpg

http://www.8ung.at/dresign/10x250.jpg



so, NEVER use .5-half-multis
qk4722 give us full-multi-benchies :)

Wow interesting :)

With Soc939 in dual channel, every 1x I drop I loose 200 or so points in Sandra mem.

Quote:

---

Originally Posted by saaya

blue, the memory badwith gets bigger because the cpu is clocked higher afaik. we are talking about buffered memory bandwidth wich depedns on the cpu speed (speed of the mxx and see units)

unbuffered there should be almost no difference in memory bandwidth with different multi and resulting different cpu speed.

and even if the 6600 have to do with the low multi, its not going to be 7500 with 250mhz higher cpu speed and a higher multi :P

and yes mini, memory bandwidth depends on the timings as well. correct me if im wrong.

---

Hummm...heres how it looks on soc939
Mem set to cas2-2-2-5-1t

200x8 = 4859/4799
200x9 = 5317/5253
200x10 = 5645/5582
200x11 = 6013/5951

And then @ 260htt
Mem set to cas 2.5-3-3-7-1t

260x10 = 7219/7134
260x11 = 7639/7535

so, NEVER use .5-half-multis
qk4722 give us full-multi-benchies :)[/QUOTE]

ok ! thx (p4z1f1st)

with a multi x.5 the divider for the mem is the next bigger

http://www.pc-tweaker.ch/Pictures/A6... Frequency.JPG

yah its because with a half multi, the mem divider at 1:1 goes to the next hight number, so with a 5.5 multi, the mem clock is CPU/6, which would make it (HTT x 5.5)/6 which is (5.5/6) times the HTT clock, thats lower than 6/6 or 5/5 which are 1

@ Blue078: could u try half-dividers? :)

for example: try 7.0 to 9.0 with all dividers :) (total 5 dividers) :)

.5 multis dont work on my board with this BIOS :)

lol?

right then ^^

Quote:

---

Originally Posted by saaya

hey qk4722 :wave:




10-2-2-2.0 max fsb and 10-3-3-2.5 max fsb please :D
thank you! :p:

and high vdimm! :cord:
hehehehe :D

or is
2.85v 2.5-3-3-5 260x10 1T 4X the maximum?
can you try 2.85v 2.5-3-3-10 1T 3x/2x max fsb?

thx! :D

looks like nice memory! :D

and look codered! they are using your a64 tweaker! :D

---

ues 2.5V vdimm :banana:
http://61.220.123.253/cpg/albums/use...%20prime95.JPG

even fsb300Mhz OK :toast:
http://61.220.123.253/cpg/albums/use...4G%20300x8.JPG

4X,1T 6-3-3-2.5@DDR600
http://61.220.123.253/cpg/albums/use...%201t%204x.JPG

http://61.220.123.253/cpg/albums/use...04x%207369.JPG

johnnyliu3377

Quote:

---

Originally Posted by johnnyliu3377

4X,1T 6-3-3-2.5@DDR600

johnnyliu3377

---

So guys...PLZ!! Don´t you get the point?

There is no "1:1" with half multis, as A64 only supports integer memory dividers.

When running your CPU@300x8.5, your Memory will run at 2550/9=283MHz!

Greetz, Flox

@pazifist: das Gleiche meinte ich auch oben...und blinky hat es ja auch schon angesprochen...

Quote:

---

Originally Posted by Gogeta

I disagree fully until I see some proof :D

DDR600 is sooo sweet :toast:

---

he's right unfortunately. Trcd and Trp at 3 cost you big points not to mention 4 and indeed it's slower than 2-2-2-5 at 240mhz.