• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Is there any gain from running RAM 1:1 with your CPU

E

entropy1982

Golden Member
When i say this i mean if say you had a 3200+ venice that was at 270X10 and a cpu that was 300X9 (both at 2700mhz) and your ram ran at 270. Would you gain from the fact that your ram is running 1:1 with cpu?

Thanks
 
negligible.

pure processor speed makes the difference on AMD platforms. thus when you have to decide between a higher CPU cap with a divider limiting the ram, versus a ram cap but running 1:1, choose higher CPU.
 
http://www.xbitlabs.com/articles/memory/display/ocz-ddr600_4.html

In summary, cryptonomicon has no idea what he is talking about. Blada blada blada, More FSB makes no difference. Look at the results. An increase of 100 FSB on the ram makes alot of difference. Compare 3-4-4-8 DDR 400 to 3-4-4-10 DDR 600.

12% in Farcry, 9% in Halflife 2, 6% in Doom 3 is a big difference.I would not call those percentages "negligable" I doubt that even an increase in clock speed of 200 MHZ would let you catch up.

Comparing Value ram timings to the DDR600, the value ram gets a little better. The difference is negated to around 2% in doom 3, 7% in farcry, and 6% in halflife 2.

As you can see from that article, High FSB gives a noticable performance increase when compared to ram with the same timings. As you can also see, tight timings matter also as they can have a good effect on performance overall. The Ideal combanation would be medium FSB (230-250) with super tight timings. Going "1:1" would require either TCCD or Hynix (shutter) .

TCCD can do around 270 at 2.5-3-3-8 if you have half decent chips, but I think 2-2-2-5 Winbound UTT running at 245 would be better. Using the ram divider has 0 effect (litterally 1% or less than 1%) on performance as you can see in the xbitlabs article. I would reccomend UTT with a divider to get to your optimal clock speed.
 
Wo wo wo, comparing DDR400 @3-4-4-8 versus DDR600 @3-4-4-10 isn't a comparison of equal ram. If your ram will run DDR600 @3-4-4-10 it will run DDR400 @2-3-3-7, that's a big difference compared to 3-4-4-8. What you need to keep in mind is that the latency is the limiting factor in ram(usually), so if your cas latency is 5ns it can run at 2cycles/400mhz or 3cycles/600mhz.

So if your comparing equal ram (with the same latency) but different bandwidth (DDR speed) the performance is negligable (on AMD systems).
 
Originally posted by: spiritwalker2223
Wo wo wo, comparing DDR400 @3-4-4-8 versus DDR600 @3-4-4-10 isn't a comparison of equal ram. If your ram will run DDR600 @3-4-4-10 it will run DDR400 @2-3-3-7, that's a big difference compared to 3-4-4-8. What you need to keep in mind is that the latency is the limiting factor in ram(usually), so if your cas latency is 5ns it can run at 2cycles/400mhz or 3cycles/600mhz.

So if your comparing equal ram (with the same latency) but different bandwidth (DDR speed) the performance is negligable (on AMD systems).
Click to expand...

:thumbsup:

I agree that comparison is just dumb. I tested my OCZ TCCD and found,

It runs at 200mhz(DDR400) with 2-2-2-5 timings

or 238mhz(DDR476) with 2-3-2-7 which gives the best memory scores of about 2.5-3% higher than stock

or 258mhz (DDR516) with 2.5-3-3-7 which gives about 2% higher than stock in memory benchmarks

or 287mhz (DDR574) with 3-4-4-8 which gives about the sames scores as DDR516


Bottom line is low latency is more important than speed on the A64 platform, and wether you run a divider or not makes no difference what so ever.

The absolute best memory scores I have achieved with the TCCD where at

232mhz with 1.5-2-2-7, but I had to give it 3.0v and I won't run it like that on a daily basis.
 
Thanks for proving my point. AMD systems don't need bandwidth (makes minimal difference), what they like is low latency. So your ram at 232mhz with 1.5-2-2-7 has your cas latency at 3.2ns while your ram setting at 287mhz with 3-4-4-8 has a cas latency of 5.2ns, and your ram at 238mhz with 2-3-2-7 has cas latency of 4.2ns. So what you've said is that the lower the cas latency (sorry for the over simplification of just using cas latency timings) the better your performance. So 3.2ns > 4.2ns > 5.2ns.

And you mention that the divider doesn't matter? Well that's the same as saying the bandwidth doesn't matter! As DDR speed is proportionate to bandwidth period.
 
Why not compare 3-4-4-8 to 3-4-4-10? If its a question of wether pure bandwith makes a difference or not, those are the two to compare......... Shows just how much of a difference bandwith makes.
 
Originally posted by: GuitarDaddy
Originally posted by: spiritwalker2223


The absolute best memory scores I have achieved with the TCCD where at

232mhz with 1.5-2-2-7, but I had to give it 3.0v and I won't run it like that on a daily basis.
Click to expand...

That is just insane............You have some pretty awesome sticks right there....
Click to expand...
 
HACP, well 3-4-4-8 @ DDR400 and 3-4-4-10 @DDR600 (I assume from your post), not only is the DDR in this example have higher bandwith but it also has lower latencies (33% lower to be exact, except for the tras). So if the latecies are the same and bandwidth changed there isn't much of a difference.

i.e.
DDR400 at 2-2-2-6 versus DDR600 at 3-3-3-9, they both have (I'll just look at cas) cas latency of 5.0ns, but the bandwidth of the DDR600 is higher.

So just cause the cycle number is the same doesn't mean the latency is the same. Got it??
 
Originally posted by: Hacp

That is just insane............You have some pretty awesome sticks right there....
Click to expand...

Yep, they are the older platinum rev2's with TCCD and brain power, not the newer TCC5.
And they are some monster sticks. The only sad part is back last year when I bought them they were damn near $300 for 2x512🙁
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top