Do the new X2s still have 2T timing for 4 dimms?

[tompson]

Hey guys. Just bought a new 4800+ and I must say I love it. I could never recommend the FX57... the smoothness of the overall computing experiance is awsome. Anyway, thats the topic for another thread.



I have 1GB now and I was wondering if I have to take the 2T hit when moving up to 2GB. I know with my old FX53 I had to, which really hurt my memory bandwidth... So much so, I sold the extra sticks and stuck (sorry about the pun) with 1GB.

I've heard rumors but does anyone know for sure if they've fixed the memory controllers in the new X2s?

 

[Avalon]

Yes, all AMD cpus are stuck with 2T if you fill your DIMM slots.

 

[tompson]

So when they said... we're putting together a new memory controller for the new core revisions, and it'll solve all the old problems... they were joking?

Thats annoying.

Excuse my Skepticism, but does anyone who actually has an X2 know for sure that it has the 2T limitation????

 

[xenolith]

2T isn't really a problem with A64.

A64 is not bandwidth limited because of the on-die mem controller, therefore you will experience very little if any real-world performance loss with CPC set to 2T.

 

[seanp789]

I have an X2 4400+ I can say that it is an improvement of the the memory controller of the winchester cores(D) but if you have 4 dimms then you will have to run at 2T even with the (E) cores. 4dimms are jsut too many dimms to query in one cycle. although i did notice my stable memory overclock with tighter timings did improve with the (E) revision by a slight margin.

 

[gunblade]

Originally posted by: seanp789
I have an X2 4400+ I can say that it is an improvement of the the memory controller of the winchester cores(D) but if you have 4 dimms then you will have to run at 2T even with the (E) cores. 4dimms are jsut too many dimms to query in one cycle. although i did notice my stable memory overclock with tighter timings did improve with the (E) revision by a slight margin.

It isn't about how many dimms to query since the MMU will have to translate the address anyway regardless of how many dimms are plugged in. It is more about electrical timing issues since dram are mostly capacitive and more dimms will present a bigger capacitive load and thus slow down the access time.

 

[Geomagick]

I use 4 Dimms and the advantage of RAM quantity more than outweighs the change to a 2T command rate.

One of the improvements to the memory controller is that it allows 4 dimms to be used at DDR400 rather than DDR333.

 

[Markbnj]

I haven't noticed any benchmark difference from 2T to 1T and back.

 

[imported_michaelpatrick33]

Originally posted by: tompson
So when they said... we're putting together a new memory controller for the new core revisions, and it'll solve all the old problems... they were joking?

Thats annoying.

Excuse my Skepticism, but does anyone who actually has an X2 know for sure that it has the 2T limitation????

Uhh, the new memory controller as another posted stated allows for 4dimms to run at DDR400 instead downclocking to DDR333. You might want to do a search for a 1t versus 2t thread that showed that the 2t timings outside of Sisandra and other synthetic benchmarks don't show a huge difference in bandwidth. There are a few where there is a relative dropoff but most are under three percent, if that.

Edit: Yes, I have a X2 but I have two 1gig dimms so I run at 1t timings and no, any E revision processor single or dual will have the same memory controller limitation. (At least every E revision processor as of today 10/04/05, LOL)

 

[imported_Soldier]

Originally posted by: xenolith
2T isn't really a problem with A64.

A64 is not bandwidth limited because of the on-die mem controller, therefore you will experience very little if any real-world performance loss with CPC set to 2T.

Simply not true. Anand's own articles show clear performance drops associated with the 2T command rate, sometimes as much as 15%. Im not sure where this incorrect data about the on-die memory controller erasing the 2T command rate impact started, but its pure hogwash and people should stop spreading it around. A64 loves tight timings, always has and always will.

"The 2T timing at 9x300 really kills performance, coming in 15% slower than 9x300 OCZ RAM. If you opt for the Venice 3000+ chip, you might want to spend a bit extra for 2-3-2 RAM as opposed to 2.5-3-3 RAM. The extra $15 or so should bring up performance at 9x300 substantially"

These results are repeated with different games and show clearly that anyone who games should be settling for nothing less than 1T. If your not a gamer the the losses are smaller, but still there.

http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2548&p=18

 

[imported_Kensai]

You do lose performance by using 2T but more ram overcomes that in some cases but in others, it's just a performance lost.
Personally, I like to use 1T on my main and work rigs, reserving 2T only for A64 servers (They're cheap!).

 

[Bona Fide]

I believe the "improvement" is that you can now run 4 sticks at DDR400. Still have to run em at 2T though.

 

[seanp789]

Originally posted by: Soldier

Originally posted by: xenolith
2T isn't really a problem with A64.

A64 is not bandwidth limited because of the on-die mem controller, therefore you will experience very little if any real-world performance loss with CPC set to 2T.

Simply not true. Anand's own articles show clear performance drops associated with the 2T command rate, sometimes as much as 15%. Im not sure where this incorrect data about the on-die memory controller erasing the 2T command rate impact started, but its pure hogwash and people should stop spreading it around. A64 loves tight timings, always has and always will.

"The 2T timing at 9x300 really kills performance, coming in 15% slower than 9x300 OCZ RAM. If you opt for the Venice 3000+ chip, you might want to spend a bit extra for 2-3-2 RAM as opposed to 2.5-3-3 RAM. The extra $15 or so should bring up performance at 9x300 substantially"

These results are repeated with different games and show clearly that anyone who games should be settling for nothing less than 1T. If your not a gamer the the losses are smaller, but still there.

http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2548&p=18

I would like to point out in that article that the higher the resolution you go the less memory bandwidth is an issue. The anandtech article only tested up to 1024x768 which reduced the difference to 5% fps. at 1600x1200 and higher the difference between 2T and 1T will be less than 1% because ram is not longer a bottleneck. If you are going to be buying high end ass ram i would hope you would be running higher than 1024x768.

 

[Hacp]

AMD can still run 4 dimms at 1T. What It can't run is more than 4 sides at 1T.

 

[formulav8]

Yeps, if you have 4x Single Sided dimms it will run 400mhz@1T, otherwise, it will run at 400mhz@2T. Thats the latest I have heard anyways.


JAson

 

[CheesePoofs]

Originally posted by: Soldier

Originally posted by: xenolith
2T isn't really a problem with A64.

A64 is not bandwidth limited because of the on-die mem controller, therefore you will experience very little if any real-world performance loss with CPC set to 2T.

Simply not true. Anand's own articles show clear performance drops associated with the 2T command rate, sometimes as much as 15%. Im not sure where this incorrect data about the on-die memory controller erasing the 2T command rate impact started, but its pure hogwash and people should stop spreading it around. A64 loves tight timings, always has and always will.

"The 2T timing at 9x300 really kills performance, coming in 15% slower than 9x300 OCZ RAM. If you opt for the Venice 3000+ chip, you might want to spend a bit extra for 2-3-2 RAM as opposed to 2.5-3-3 RAM. The extra $15 or so should bring up performance at 9x300 substantially"

These results are repeated with different games and show clearly that anyone who games should be settling for nothing less than 1T. If your not a gamer the the losses are smaller, but still there.

http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2548&p=18

then why do everyone else's proove otherwise?
http://forums.anandtech.com/messageview...atid=28&threadid=1577496&enterthread=y

 

[seanp789]

UPDATE!!!

Apparently the new DFI board based on the ATI chipset can do 4 dimms at 1T

 

[Markbnj]

Fwiw, I saw about a 1500 point difference in Crystalmark between 1T and 2T timings, on an overall score of ~52k it just didn't seem that big a deal.

 

[KoolDrew]

Originally posted by: Avalon
Yes, all AMD cpus are stuck with 2T if you fill your DIMM slots.

This is not true. If the sticks are single-sided it can be ran at 1T. If they are double-sided you would be correct though.

 

[Zim]

Originally posted by: KoolDrew

Originally posted by: Avalon
Yes, all AMD cpus are stuck with 2T if you fill your DIMM slots.

This is not true. If the sticks are single-sided it can be ran at 1T. If they are double-sided you would be correct though.

Can someone please explain how to determine if a dimm is single sided?

 

[TGS]

If the memory modules are only installed on a single bank, typically done on only one side of the PCB. With either an x4 or x8 memory module layout.

Stolen from Xbit:

AMD engineers promised to eliminate this problem in the new Venice core and they did keep their word. Athlon 64 processors based on Venice can work with four single-side DDR400 SDRAM modules without any limitations, and if there are double-side DDR400 SDRAM DIMMs installed, they can work at 400MHz with 2T timing.