WHere does 1t make a difference over 2t?

[Stoneburner]

Can't find the thread but I remember a while ago in this forum somebody showed that 1t barely has any advantage over 2t. HOwever, I can run at 296 HTT with my settings at 2-2-2-6 and 2t versus 289 HTT at 2.5 3 3 7 at 1t. Which one is superior?

 

[the Chase]

Hmm...that might be a close one-Run some Super PI runs at each setting and let us know!

 

[nerp]

I think Sandra shows a huge difference. In games the difference is maybe 2-10 FPS at most. It's been pretty well proven that 2T does NOT make your machine significantly slower.

 

[themusgrat]

For A64s, the difference between 1T and 2T isn't huge in the big picture, but it is relative to other RAM timings. It does take precedence over latencies, so 2.5-3-3-7 1T is better than 2-2-2-6 2T. But always leave tras at 7, since it helps the other timings, and AT's reviews have shown it to be best for bandwidth.

 

[5t3v0]

Originally posted by: nerp
I think Sandra shows a huge difference. In games the difference is maybe 2-10 FPS at most. It's been pretty well proven that 2T does NOT make your machine significantly slower.

Nerp is right, I drop a whopping 1000MB/s on both Integer & Floating point tests in Sandra but only 14 3dmarks in 3dmark05 when I switched to 2T. So bugg3r-all difference in real world use.

 

[JohnAn2112]

Originally posted by: Stoneburner
Can't find the thread but I remember a while ago in this forum somebody showed that 1t barely has any advantage over 2t. HOwever, I can run at 296 HTT with my settings at 2-2-2-6 and 2t versus 289 HTT at 2.5 3 3 7 at 1t. Which one is superior?

What kind of RAM do you have that can run 2-2-2-6 2T at 296HTT?

 

[IdaGno]

http://www.hardforum.com/showthread.php?t=917047

In 10x250, we lose 3% Aquamark performance when switching to 2T.
In 10x227, we lose 4.5% Aquamark performance when switching to 2T. Not a very big deal if getting 2 gigs means having to use 2T.

 

[letdown427]

I remember that thread, where is it now? I was searching for it and came across this one, has anyone dug up that old thread? It was good reading.

 

[htmlmasterdave]

Originally posted by: themusgrat
For A64s, the difference between 1T and 2T isn't huge in the big picture, but it is relative to other RAM timings. It does take precedence over latencies, so 2.5-3-3-7 1T is better than 2-2-2-6 2T. But always leave tras at 7, since it helps the other timings, and AT's reviews have shown it to be best for bandwidth.

Could you give the link to that review? I was recently wondering about this all myself... I have my 2800+ s754 @ 2.25ghz 1:1 2.5-4-3-7 2T using OCZ PC4000EL(2x512 kit), and I'm trying to figure out how to squeeze some better timings out of it. Also does anyone have experience with that specific ram? Thanks!

P.S. I get about 1950 meg/s using memtest86... is that decent?

 

[Diasper]

On single-core CPUs you won't lose much - dual-core is much more up for debate since when both cores are under load they have shown gains from additional bandwidth - and 2T inhibits bandwidth significantly.

In an anandtech review I think dual-core when playing Doom 3 and encoding at the same time showed as much as a 10% speed boost from 200->250mhz. That 10% might even be more significant as it might bely an increased minimum FPS much greater than 10%.

*edited*

 

[letdown427]

We could just do some testing and post the results? Recreate that old threads greatness lol.

I.E people just run a benchmark say 3 times and take average. All settings stock(or at least equal) and do it at 1t, and then at 2t. and ten post the results and % difference. Although we can probably ignore sandra as a bench, as it loves 1t.

just a thought.

 

[Diasper]

Sure though until dual-core games come out it'll be hard to tell - you'll just need to run Game eg Half Life 2 (prob more representative of future CPU loads) + FRAPS to measure min FPS and then on the other core do some encoding in a way that you can measure min/average FPS encoded.

Then both of these need to be compared when just run on their own as presumably a single CPU process with 200/250mhz is much less likely to be bandwidth straved and thus be able to provide the constant value of proper potential if not bandwidth limited.

Anyone?

 

[letdown427]

I meant more in response to the 1t v 2t question. Do benchmarks as normal with 1t, then wit 2t, compare, discuss.

Measuring how much of an issue 2t is for dual cores is difficult yes. SuperPI on each core? meh, i dunno.

 

[Diasper]

Also, depending what you're benchmarking I think best testing procedure would be to take the 2nd and 3rd runs ignoring the first.

Ah well, anyone up for it? PM Zebo?

 

[letdown427]

I'm definately up for it, i'll do some aquamark3 benches,3dmark01 and superpi. why discount first run? Trying to discount any possible cache advantages? I'll probably do four runs and take average of 2nd to 4th run then

wasn't it zebo who had that thread in the first place? Loads of people posted results, it was v.good.