DDR - 1T or 2T

Associate
Joined
13 Apr 2007
Posts
172
Location
Wolverhampton
Hi,

I've got 4 x 512MB of Geil ram currently running at 480MHz (240MHz) with timings of 2.5-3-3-5 @ 2T (MemTest86 stable)

The ram is not stable at those settings at 1T (tried it but instantly failed MemTest86)

Is it worth me starting my overclock from scratch with 1T and seeing what timings & speed I can stably get, or will the performance difference be negligable?

I'm guessing that my choices are something like the following:

1) 480MHz 2.5-3-3-5 (2T)
or
2) ~400MHz 2-3-3-5 (1T)

Thanks,

Banjo ;_;
 
You could try, but I think you'll be hard pushed with x4 dimms getting 1T to run. At stock speed, yes it may run 1T, however the memory controller usually has to do 400Mhz 2T on AMD systems.

It used to be, not sure if it still applies,

333Mhz 1T
or
400Mhz 2T

^ that depend on the core of the chip as well. Later revision AMD CPU's would revert to the 2T command with x4 dimms.

Although for the 40Mhz you'd drop to *try* and get 1T I doubt it'd be worth the difference.
 
Id go for which ever gives you more bandwidth, look at the top left when you are running memtest
id put money on T1 having more bandwidth than overclocked T2
 
Thanks Firegod & ASH1982,

Will try the various settings and post results of the bandwidth I'm getting :)

Banjo ;_;
 
Is it Geil DDR550 Ultra-X you have (I'm guessing from the timings you're using)?

I read the DFI NF4 Expert could do 1T at DDR400 but I've tried it on my Expert with 4 x Geil DDR550 Ultra-X and wasn't able to get it stable although I didn't try too hard. IIRC your board is very similar to the expert in design so maybe it'll be possible.

If it is Geil DDR550 Ultra-X you have did you manage to get 2 x 512MB stable at stock (DDR550) speeds as I can't get anywhere near that with mine and so far Geil haven't responded to any emails I've sent them regarding it.
 
Hey str,

Actually, (surprisingly) I've got two pairs of bog standard Geil PC3200.
One rated at 2-3-3-6 and the other at 2-4-4-8.
I know I can get the ram to 500MHz but I need to relax the latencies a bit more so I don't know whether I'll get a performance boost, I'm going to check out the bandwidth figures on Memtest86 later on :)

I've got no experience with Geil DDR550 Ultra-X, but sometimes with high-end ram, to get that speed you need to increase the voltage to it, and if you're lucky then the manufacturer will allow for this without voiding the warranty.

(Think about Corsair Dominator DDR2 RAM running at 1,111MHz, it needs to be over-volted to be stable but overvolting it (to an extent) doesn't void the warranty)

Banjo ;_;
 
Well I've read it's not a good idea to run 2.9v+ on TCCD memory as it gets too hot and eventually starts to fail. The highest I've tried is 2.85v and still can't get it stable beyond DDR500. The best I can get is DDR500 with 2.5-4-4-7 settings and one set works fine with MAL at 8 and RP at 6 but the other needs MAL at 10 and RP at 8. Both sets together work fine at DDR500 with 2.5-4-4-7 in my Ultra-D but aren't stable in the Expert at that speed so I suspect they're close to their limits at DDR500 speeds and the Ultra-D runs TCCD more stable than the Expert which is supposedly tuned for UTT and the like.
 
In the testing I have done in the past, in actual usage, 1T is not any better than 2T unless you are purely looking to have a good benchmark for memory bandwidth. So, IMO, if you cannot easily get it stable with 1T, I would not lose much sleep over it. Go for a higher OC and more CPU cycles.
 
@Yellowbeard - lots of people are suddenly reporting being able to run 1T with 4 sticks of RAM - how is this possible? Is DDR2 addressed as just one bank even if you have 4 sticks or something?
 
WJA96 said:
@Yellowbeard - lots of people are suddenly reporting being able to run 1T with 4 sticks of RAM - how is this possible? Is DDR2 addressed as just one bank even if you have 4 sticks or something?
No clue and I have not seen any reports of this. But, I am not out looking. If you have links, I'll take a look.

If I had to guess, I would guess 1 of 2 things or possibly a combination of both.
1. Since command rate is a chipset function, perhaps chipsets are improving to the point that they are able to overcome the previous limits.
2. Higher latency memory is less load and is thus allowing the chipsets to run a 1T command rate at certain FSB straps.
 
Hi everyone, thanks for the fast replies :)

Got some results for you (read-outs from MemTest86)

Speed: -- Latencies: Cmd Rate: Bandwidth:
480MHz - 2.5-3-3-5 - 2T ----- 2344MB/s
515MHz -- 3-3-3-5 -- 2T ----- 2577MB/s

396MHz - 2.5-3-3-5 - 1T ----- 2715MB/s
410MHz -- 2-3-3-5 -- 1T ----- 2834MB/s
427MHz -- 2-3-3-5 -- 1T ----- 2925MB/s
445MHz - 2.5-3-3-5 - 1T ----- 2993MB/s
454MHz -- 3-3-3-5 -- 1T ----- 3000MB/s

Thanks again for all the info, basically it looks promising as raw data, but bear in mind that those latencies are too tight for the ram to function 100% stably in 1T.

I've done those tests to show what sort of numbers I can expect if I do spend the time fine-tuning the (many) ram timings on my LANParty CFX3200-DR.

To be honest str, those are some very nice timings for DDR @ 500MHz, I don't really think you're missing out on much by not being able to hit 550MHz effective, although it's worth querying Geil as to why the ram won't/can't run at it's rated speed.

Thanks YellowBeard, to be honest I'm happier with my RAM at 515MHz and 2T because it sounds so much more impressive to brag about :D
Also, I'm going to push my CPU further (maybe to ~2.9GHz) if the chip can cope with that at ~1.5V for when I save up enough money to buy a high end GPU such as the HD 2900XT. (I know a current high end GPU will run fine on a 2.6GHz A64 but for games like Supreme Commander which is very CPU intensive, etc.)
 
Banjo said:
To be honest str, those are some very nice timings for DDR @ 500MHz, I don't really think you're missing out on much by not being able to hit 550MHz effective, although it's worth querying Geil as to why the ram won't/can't run at it's rated speed.
The thing is it's rated to do 2.5-3-3-5 at DDR500 but errors are occuring. The system is 24 hours Orthos stable at DDR500 2.5-4-4-7 but the last couple of days it's blue screened 3 times and I suspect it's the memory causing the problem. I also recently tested at DDR400 2-2-2-5-2T and with 4 modules it was failing Orthos. One set is about a year old but the other is only a month or so old so it would seem something is up with Geil's QA.

I got an email back from Geil yesterday so hopefully they'll help me get it sorted.
 
WJA96 said:
All this is pretty odd to me - one of those 'everyone knows' things is being challenged :confused:

It was a limitation of the earlier A64s memory controller that stopped you using 4 sticks at 1T. I'm not sure if it was ever fixed in the later cores.

Jokester
 
It wasn't generally possible due to the A64 memory controller but then DFI tweaked their eXpert board's memory circuitry to give DDR400 and 1T with 4 sticks as long as the memory controller played along and there are reports of some getting it working stable.
 
str said:
The thing is it's rated to do 2.5-3-3-5 at DDR500 but errors are occuring. The system is 24 hours Orthos stable at DDR500 2.5-4-4-7 but the last couple of days it's blue screened 3 times and I suspect it's the memory causing the problem. I also recently tested at DDR400 2-2-2-5-2T and with 4 modules it was failing Orthos.

Have you tried using MemTest86, even if the RAM isn't stable, it'll still give you a read-out of how much bandwidth the current set-up is giving, you should do what I did and compare your results at 400MHz, 500MHz and 550MHz with the various latencies. Just to see what you're missing out on.
Glad to hear Geil are getting in touch with you :)

Also, thanks for that bit of knowledge about DFI's boards, I wasn't sure whether It was possible to run 4 sticks at 1T. Actually, in the manual it says when 4 sticks are inserted it defaults to 333MHz, but I'd not realised why.
Might see if I can get it running stable at ~400Mz with 1T, just so that I can say that I did it :p

Banjo ;_;
 
Back
Top Bottom