hey guys.. i have hear around the net in some places about the FSB and DRAM sync ratio... whats the big deal about this thing... why so many people want them sync?????
FSB:DRAM
- Thread starter Pelu
- Start date
From what I understand, 1:1 is the best ratio because it makes your FSB and your RAM work in sync.
Other ratios will cause you to basically 'waste' clock cycles because either your FSB or your RAM has to 'wait' on data.
That is why in synthetic benchmarks, 1:1 will always yield a higher score.
PS: Welcome to the forums!!
Other ratios will cause you to basically 'waste' clock cycles because either your FSB or your RAM has to 'wait' on data.
That is why in synthetic benchmarks, 1:1 will always yield a higher score.
PS: Welcome to the forums!!
n7
Elite Member
- Jan 4, 2004
- 21,281
- 4
- 81
1:1 is easy to run, & is common when overclocking.
But it absolutely is not the "best ratio".
Aside from on 975X (old) or nForce chipsets running 1T (less common), 1:1 will generally lose to 5:6, 4:5, 2:3, 1:2, etc. in real world & synthetic benches.
It's the same with the AM2 platform.
This isn't new or news; it's well known, so i'm somewhat disappointed to see 1:1 being claimed as the best.
One can easily run their own benches to see this.
However, the disclaimer with regards to higher ratios is that the benefit from running a higher speed is extremely small.
It can be measured, but due to the extra price & voltage required, it's not usually worth it.
That doesn't mean it's not generally faster though, as it is (though by very small margins).
But it absolutely is not the "best ratio".
Aside from on 975X (old) or nForce chipsets running 1T (less common), 1:1 will generally lose to 5:6, 4:5, 2:3, 1:2, etc. in real world & synthetic benches.
It's the same with the AM2 platform.
This isn't new or news; it's well known, so i'm somewhat disappointed to see 1:1 being claimed as the best.
One can easily run their own benches to see this.
However, the disclaimer with regards to higher ratios is that the benefit from running a higher speed is extremely small.
It can be measured, but due to the extra price & voltage required, it's not usually worth it.
That doesn't mean it's not generally faster though, as it is (though by very small margins).
It does take benchmarks to 'see' the difference. A game may have 1 or 2fps increase. Photoshop or winrar may finish the 5m30s job like 12s faster. You may see small gains w/ benchmarks, but actually using 2 computers that are identical except how the ram is config'd you wouldn't be able to see/feel the difference.
Yes cheex you stand corrected. In synthetic benchmarks out of syncdoes yield better scores in many situations. The one that comes to mind as not benefiting is having a 1066FSB cpu, and when you run the ram at 533/667/800, the 667 score is lower then the 533, and the 800 is only a little better then the 533. 1:1 is definatly the easiest and unless you have a specific need/want for out of sync settings (ie Benchmark bravado or a DC program) it's the way to go.
Ahh yes, i rem that now earthwormjim.
yeah, running out of sync does not slow anything down. It's just that you get the most bang for your buck by using 1:1. In other words, you can buy the cheapest RAM available that will work 1:1 with your CPU, even if it is overclocked you should be able to do this fairly easily.
Nothing wrong with running higher ratio, it's just more expensive since you are spending more on more expensive RAM. And yes, for compression algorithms it seems to be a bit faster, but gaming is not really RAM limited, it is much more CPU / GPU limited. Increasing ratio for gaming is just not worth it, same reason no one really recommends DDR3 right now. You'd need quite a high FSB to make DDR3 worthwhile right now...
Nothing wrong with running higher ratio, it's just more expensive since you are spending more on more expensive RAM. And yes, for compression algorithms it seems to be a bit faster, but gaming is not really RAM limited, it is much more CPU / GPU limited. Increasing ratio for gaming is just not worth it, same reason no one really recommends DDR3 right now. You'd need quite a high FSB to make DDR3 worthwhile right now...
EarthwormJim
Diamond Member
- Oct 15, 2003
- 3,239
- 0
- 76
As I said in another similar thread, running higher memory than fsb ratios on an Intel platform allows you to run a lower tRD.
That's typically where the gain in performance is coming from. Not inherently from running faster memory, but being able to lower your tRD.
If your memory can do it, run it as fast as possible for the most performance.
That's typically where the gain in performance is coming from. Not inherently from running faster memory, but being able to lower your tRD.
If your memory can do it, run it as fast as possible for the most performance.
Okay, I stand corrected on my above statement.
I also have a question...
What is so special about the tRD?
I also have a question...
What is so special about the tRD?
n7
Elite Member
- Jan 4, 2004
- 21,281
- 4
- 81
Q6700 @ 10x350 = 3500 MHz 2:3 DDR2-1050 5-5-5-15 TRCD=55 - TRD=5
vs.
Q6700 @ 10x350 = 3500 MHz 2:3 DDR2-1050 5-5-5-15 TRCD=60 - TRD=8
vs.
Q6700 @ 9x389 = 3501 MHz 1:1 DDR2-778 4-4-4-12 TRCD=60 - TRD=8
Some of the the subtimings that were set to auto did tighten slightly @ 1:1, so there's a bit of an advantage to the 1:1 benching, but performance level aka TRD remained the same, at 8.
All benches were run at least three times then averaged.
Everest Memory Benchmark
Read
2:3 TRD 5 - 8410.67 MB/s
2:3 TRD 8 - 7575.33 MB/s
1:1 TRD 8 - 7314.67 MB/s
Write
2:3 TRD 5 - 6379 MB/s
2:3 TRD 8 - 6378.33 MB/s
1:1 TRD 8 - 7100 MB/s
Copy
2:3 TRD 5 - 7066 MB/s
2:3 TRD 8 - 6901 MB/s
1:1 TRD 8 - 7139 MB/s
Latency
2:3 TRD 5 - 59.67 ns
2:3 TRD 8 - 68.37 ns
1:1 TRD 8 - 70.87 ns
WinRAR bench
2:3 TRD 5 - 1583.25 KB/s
2:3 TRD 8 - 1521.75 KB/s
1:1 TRD 8 - 1429.5 KB/s
3DMark01 defaults
2:3 TRD 5 - 45823
2:3 TRD 8 - 44806
1:1 TRD 8 - 45344.33
Cinebench R10 64-bit
2:3 TRD 5 - 14327
2:3 TRD 8 - 14147.33
1:1 TRD 8 - 14245.33
vs.
Q6700 @ 10x350 = 3500 MHz 2:3 DDR2-1050 5-5-5-15 TRCD=60 - TRD=8
vs.
Q6700 @ 9x389 = 3501 MHz 1:1 DDR2-778 4-4-4-12 TRCD=60 - TRD=8
Some of the the subtimings that were set to auto did tighten slightly @ 1:1, so there's a bit of an advantage to the 1:1 benching, but performance level aka TRD remained the same, at 8.
All benches were run at least three times then averaged.
Everest Memory Benchmark
Read
2:3 TRD 5 - 8410.67 MB/s
2:3 TRD 8 - 7575.33 MB/s
1:1 TRD 8 - 7314.67 MB/s
Write
2:3 TRD 5 - 6379 MB/s
2:3 TRD 8 - 6378.33 MB/s
1:1 TRD 8 - 7100 MB/s
Copy
2:3 TRD 5 - 7066 MB/s
2:3 TRD 8 - 6901 MB/s
1:1 TRD 8 - 7139 MB/s
Latency
2:3 TRD 5 - 59.67 ns
2:3 TRD 8 - 68.37 ns
1:1 TRD 8 - 70.87 ns
WinRAR bench
2:3 TRD 5 - 1583.25 KB/s
2:3 TRD 8 - 1521.75 KB/s
1:1 TRD 8 - 1429.5 KB/s
3DMark01 defaults
2:3 TRD 5 - 45823
2:3 TRD 8 - 44806
1:1 TRD 8 - 45344.33
Cinebench R10 64-bit
2:3 TRD 5 - 14327
2:3 TRD 8 - 14147.33
1:1 TRD 8 - 14245.33
EarthwormJim
Diamond Member
- Oct 15, 2003
- 3,239
- 0
- 76
Why not lower performance level? That's really the main benefit to running a higher fsb to memory ratio.
Wow, so it is really a give and take situation.
Probably what we need to find out is the difference it makes in which type of applications.
Example: Let's say that 1:1 is best for games, and another ratio is best for encoding, whatever, etc...
Probably what we need to find out is the difference it makes in which type of applications.
Example: Let's say that 1:1 is best for games, and another ratio is best for encoding, whatever, etc...
Top 4 are purely synthetic:
Everest Memory Benchmark
Read
2:3 - 7575.33 MB/s
1:1 - 7314.67 MB/s
3.56% improvement
Write
2:3 - 6378.33 MB/s
1:1 - 7100 MB/s
11.31% improvement
Copy
2:3 - 6901 MB/s
1:1 - 7139 MB/s
3.45% improvement
Latency
2:3 - 68.37 ns
1:1 - 70.87 ns
3.66% improvement
WinRAR bench
2:3 - 1521.75 KB/s
1:1 - 1429.5 KB/s
6.45% improvement
3DMark01 defaults
2:3 - 44806
1:1 - 45344.33
1.2% improvement
Cinebench R10 64-bit
2:3 - 14147.33
1:1 - 14245.33
6.93% improvement
2:3= 3.56+3.66+6.45=13.67/3=4.56% on average
1:1=11.31+3.45+1.2=15.96/3=5.32% on average
Take that for what you will, I just wanted to work out the numbers, instead of absolutes (x won and thats that. But won by how much?..)
Everest Memory Benchmark
Read
2:3 - 7575.33 MB/s
1:1 - 7314.67 MB/s
3.56% improvement
Write
2:3 - 6378.33 MB/s
1:1 - 7100 MB/s
11.31% improvement
Copy
2:3 - 6901 MB/s
1:1 - 7139 MB/s
3.45% improvement
Latency
2:3 - 68.37 ns
1:1 - 70.87 ns
3.66% improvement
WinRAR bench
2:3 - 1521.75 KB/s
1:1 - 1429.5 KB/s
6.45% improvement
3DMark01 defaults
2:3 - 44806
1:1 - 45344.33
1.2% improvement
Cinebench R10 64-bit
2:3 - 14147.33
1:1 - 14245.33
6.93% improvement
2:3= 3.56+3.66+6.45=13.67/3=4.56% on average
1:1=11.31+3.45+1.2=15.96/3=5.32% on average
Take that for what you will, I just wanted to work out the numbers, instead of absolutes (x won and thats that. But won by how much?..)
EarthwormJim
Diamond Member
- Oct 15, 2003
- 3,239
- 0
- 76
Originally posted by: n7
That'd be for another day
I just whipped up those benches last night.
I'm also still figuring out my DFI board, since thus far, it seems to be alot smarter than i am
It's a pretty crazy board as far as available features. Have you read the anandtech review/guide? Was pretty helpful as was the xtremetech forum.
Try messing with the tRD setting, it's pretty quick to find a stable one. If your computer posts, it's stable (pretty much). I'm thinking you could do a tRD of 5 with 2:3 ratio and a tRD of 6-7 with 1:1.
EarthwormJim
Diamond Member
- Oct 15, 2003
- 3,239
- 0
- 76
http://www.anandtech.com/mb/showdoc.aspx?i=3208&p=5
Basically, lowering the tRD leads to a near linear reduction in memory latency. I'm not just basing that off the anandtech article, it's what I have found playing with my DFI board too.
CAS and the other well known memory timings have no where near as linear or direct of a decrease on memory latency.
That is interesting... for a comparison here's my everest read and latency results running 11x300 3:5 DDR2-1000Mhz 5-5-5-15-2T. I can't control tRD with my board, or at least I don't know how I could.
Memory Read 7979MB/s
Latency 58.1ms
I haven't purchased Everest so that's all I get. I also tried 2:3 DDR-900 4-4-4-12-2T
Memory Read 7746MB/s
Latency 62.5ms
I was kind of surprised that the 100Mhz leveled out the tighter timings.
n7
Elite Member
- Jan 4, 2004
- 21,281
- 4
- 81
EarthwormJim, just for you (& because i wanted to play with lower TRD eventually anyway)
Q6700 @ 10x350 = 3500 MHz 2:3 DDR2-1050 5-5-5-15 TRCD=55 - TRD=5
vs.
Q6700 @ 10x350 = 3500 MHz 2:3 DDR2-1050 5-5-5-15 TRCD=60 - TRD=8
vs.
Q6700 @ 9x389 = 3501 MHz 1:1 DDR2-778 4-4-4-12 TRCD=60 - TRD=8
Some of the the subtimings that were set to auto did tighten slightly @ 1:1, so there's a bit of an advantage to the 1:1 benching, but performance level aka TRD remained the same, at 8.
All benches were run at least three times then averaged.
Everest Memory Benchmark
Read
2:3 TRD 5 - 8410.67 MB/s
2:3 TRD 8 - 7575.33 MB/s
1:1 TRD 8 - 7314.67 MB/s
Write
2:3 TRD 5 - 6379 MB/s
2:3 TRD 8 - 6378.33 MB/s
1:1 TRD 8 - 7100 MB/s
Copy
2:3 TRD 5 - 7066 MB/s
2:3 TRD 8 - 6901 MB/s
1:1 TRD 8 - 7139 MB/s
Latency
2:3 TRD 5 - 59.67 ns
2:3 TRD 8 - 68.37 ns
1:1 TRD 8 - 70.87 ns
WinRAR bench
2:3 TRD 5 - 1583.25 KB/s
2:3 TRD 8 - 1521.75 KB/s
1:1 TRD 8 - 1429.5 KB/s
3DMark01 defaults
2:3 TRD 5 - 45823
2:3 TRD 8 - 44806
1:1 TRD 8 - 45344.33
Cinebench R10 64-bit
2:3 TRD 5 - 14327
2:3 TRD 8 - 14147.33
1:1 TRD 8 - 14245.33
Massive difference IMO
Just by tightening one timing, one that so many motherboards don't even let you set.
Q6700 @ 10x350 = 3500 MHz 2:3 DDR2-1050 5-5-5-15 TRCD=55 - TRD=5
vs.
Q6700 @ 10x350 = 3500 MHz 2:3 DDR2-1050 5-5-5-15 TRCD=60 - TRD=8
vs.
Q6700 @ 9x389 = 3501 MHz 1:1 DDR2-778 4-4-4-12 TRCD=60 - TRD=8
Some of the the subtimings that were set to auto did tighten slightly @ 1:1, so there's a bit of an advantage to the 1:1 benching, but performance level aka TRD remained the same, at 8.
All benches were run at least three times then averaged.
Everest Memory Benchmark
Read
2:3 TRD 5 - 8410.67 MB/s
2:3 TRD 8 - 7575.33 MB/s
1:1 TRD 8 - 7314.67 MB/s
Write
2:3 TRD 5 - 6379 MB/s
2:3 TRD 8 - 6378.33 MB/s
1:1 TRD 8 - 7100 MB/s
Copy
2:3 TRD 5 - 7066 MB/s
2:3 TRD 8 - 6901 MB/s
1:1 TRD 8 - 7139 MB/s
Latency
2:3 TRD 5 - 59.67 ns
2:3 TRD 8 - 68.37 ns
1:1 TRD 8 - 70.87 ns
WinRAR bench
2:3 TRD 5 - 1583.25 KB/s
2:3 TRD 8 - 1521.75 KB/s
1:1 TRD 8 - 1429.5 KB/s
3DMark01 defaults
2:3 TRD 5 - 45823
2:3 TRD 8 - 44806
1:1 TRD 8 - 45344.33
Cinebench R10 64-bit
2:3 TRD 5 - 14327
2:3 TRD 8 - 14147.33
1:1 TRD 8 - 14245.33
Massive difference IMO
Just by tightening one timing, one that so many motherboards don't even let you set.
n7
Elite Member
- Jan 4, 2004
- 21,281
- 4
- 81
Scoop, get MemSet (if you don't already have it) & tell us what "Performance Level" is set to.
DL here: http://www.xtremesystems.org/f...showthread.php?t=92190
Performance Level is a fancy name for TRD
DL here: http://www.xtremesystems.org/f...showthread.php?t=92190
Performance Level is a fancy name for TRD
EarthwormJim
Diamond Member
- Oct 15, 2003
- 3,239
- 0
- 76
N7, try playing with the tRD phase adjustments now. You may be able to run a tRD of 4 on some phases, if not all.
In that menu, setting each option to enabled does a -1 to your current tRD. Do each one individually so you can find out which ones you can and can't enable.
If you don't see that option in memory timings, try a newer bios my original bios didn't have it.
On my DFI board (same as yours) I can't post at a tRD of 6. However with the phase adjustments, I can set my tRD to 7 and with all phases enabled, I'm really running a tRD of 6.
In that menu, setting each option to enabled does a -1 to your current tRD. Do each one individually so you can find out which ones you can and can't enable.
If you don't see that option in memory timings, try a newer bios my original bios didn't have it.
On my DFI board (same as yours) I can't post at a tRD of 6. However with the phase adjustments, I can set my tRD to 7 and with all phases enabled, I'm really running a tRD of 6.
n7
Elite Member
- Jan 4, 2004
- 21,281
- 4
- 81
I have lots of questions for you regarding this mobo...
It's been actually really frustrating to work with, since there's almost too many options, & maintaining stability seems to be insanely hard.
Truth is, i haven't been able to find a single combination of settings that can pass HCI Memtest.
I know it's because of having 8 GB, but i would have thought that if i cranked NB voltage i could get it to at least make it to 100% thru four HCI instances, but nope.
Yet i can pass P95 custom set to stress all 8 GB easily.
It seems HCI puts too much stress on the NB...board just can't hack it.
I don't know why you wouldn't be able to post at TRD 6...likely you need more vNB...
I mean, if i can get it running with four 2 GB dimms raping the board, i woulda thought anyone can.
FWIW, i'm using the 266/800 ratio if you hadn't already done the math to figure that out.
The 333/667 & 333/800 ones seem to hate me unless i pour in vNB...
Anyway, maybe we need to make another thread here for discussion on this board.
Originally posted by: n7
Scoop, get MemSet (if you don't already have it) & tell us what "Performance Level" is set to.
DL here: http://www.xtremesystems.org/f...showthread.php?t=92190
Performance Level is a fancy name for TRD
Yeah, checked it, it's set to 4.
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 24K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes Discussion Threads
- Started by Tigerick
- Replies: 20K
-
-
-
Facebook
Twitter