• 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 much speed gain in faster RAM?

RonAKA

Member
On my last overclocking of an E4300 I ran the FSB at 333 to give me a CPU frequency of 3 GHz or so. It also nicely kept my 667 DDR2 right on spec with no OC using a 1:1 multiplier. I had to replace the motherboard so am rebuilding it again. RAM now seems to be quite cheap. Is it worth it to upgrade to 2 GB from 1, and go to 1066? I would think there must be some combination of CPU multiplier and RAM multiplier that would put the RAM in the right frequency range. I run XP.
 
I have not seen any noticeable gains going from DDR2-667 all the way up to DDR2-1300 in anything i've used on my PC. Synthetic benchmarks may show an improvement but real world apps may not benefit as much.
 
I'd definitely get more RAM.
2x2 GB minimum these days.

But getting faster RAM isn't really all that helpful for real world improvements; though it does help if you are maxing out your RAM's speed when OCing.

However, even your DDR2-667 will likely to OC to 800 speeds, if not more, if you are now needing to overclock.

But if you only had 1 GB, i'd definitely pick up a new kit.

Here's some DDR2-800 5-5-5 1.8v for only $37.
http://www.newegg.com/Product/...x?Item=N82E16820231207

If you really want to go with faster RAM so you can run higher than 1:1 ratios, then this 1066 stuff is good, & cheap.
http://www.newegg.com/Product/...x?Item=N82E16820231166
 
ram is so cheap i would upgrade just so I could say I have more RAM 🙂 2x2GB seems a great price point i would go with that. the whole 1066 thing is not all its cracked up to be IMO 800mhz should be fine.
 
Thanks for the advice. It looks like I may gain better prerformance by using more RAM than by faster RAM. This Gigabyte GA-EP45-UD3R board seems to have a bit of an odd way of setting the memory multiplier. If I read it right you have to select a (G) MCH Frequency Latch of 266(A), 333(B), 200(C), or 400(D). Then it gives you a choice of multipliers which depending on your choice of latch speed, can be from 2.0 to 4.0 (total of 7 speeds), and the multipliers seem to include the doubling, so 2.0 is really 1:1. However, it seems you can select any Frequency Latch regardless of where you run the FSB, and the whole thing seems to be a convoluted way to just pick the multiplier. If you really do run at the selected latch frequency, then the provided multipliers for that option do work out to the standard RAM speeds, so I guess that is why they do it that way?

Anyway all that said, I did a spreadsheet based on the assumption that all multipliers were available to use, and came up with the following options for 1066 RAM. Which ones would be the most preferrable? Is there a downside to running the actual FSB frequency higher instead of lower? Harder on the Northbridge? My target for OC is about 3.2 GHz, but it may work out to a bit lower or higher.

FSB, CPU X, CPU Freq, Mem X, Mem Freq
350, 9, 3.15, 3.2, 1120
400, 8, 3.20, 2.66, 1064
450, 7, 3.15, 2.4, 1080
525, 6, 3.15, 2.0, 1050
 
If you are running 1GB of ram, moving to 4GB which can be had for $40 these days is a great upgrade. You will surely notice lower hard disk cache usage and thus an appreciable increase in performance. As others have also stated, moving to faster ram doesn't give much in the way of actual performance, however the ability to potentially hit a higher stable CPU overclock can yield solid real-world gains.
 
Another couple of options if I go with 800 RAM would be:

FSB, CPU X, CPU Freq, Mem X, Mem Freq
350, 9, 3.15, 2.4, 840
400, 8, 3.20, 2.0, 800

Not sure I can run 4 GB with XP. Somewhere I think I read XP can only handle 3 GB, unless that was changed in SP-3.
 
Went with DDR-2 800. First tried some Corsair XMS2, but it would not POST. This Gigabyte GA-EP45-UD3R board, which I am becoming less and less pleased with would appear to be the problem.

With Kingston Value RAM I got 3.0 GHz using the 9X CPU mulitplier, 333 FSB, and the 2.4 mulitipler on the RAM to get 800 MHz. If anything this board is not as good as my old ASUS P5B DX.
 
Originally posted by: RonAKA
This Gigabyte GA-EP45-UD3R board seems to have a bit of an odd way of setting the memory multiplier. If I read it right you have to select a (G) MCH Frequency Latch of 266(A), 333(B), 200(C), or 400(D). Then it gives you a choice of multipliers which depending on your choice of latch speed, can be from 2.0 to 4.0 (total of 7 speeds), and the multipliers seem to include the doubling, so 2.0 is really 1:1. However, it seems you can select any Frequency Latch regardless of where you run the FSB, and the whole thing seems to be a convoluted way to just pick the multiplier. I

FSB, CPU X, CPU Freq, Mem X, Mem Freq
350, 9, 3.15, 3.2, 1120
400, 8, 3.20, 2.66, 1064
450, 7, 3.15, 2.4, 1080
525, 6, 3.15, 2.0, 1050

The main reason it's confusing is because Gigabyte has a retarded layout/description in their BIOS.

The MCH Frequency Latch is the NB strap...different ratios will be available for different straps...not all straps have the same ratios; actually, generally every strap has different ones.

This i stole from elsewhere, but it explains the normal strap + their corresponding ratios.

200 Strap
1:1 (DDR2-400)
2:3 (DDR2-533)
3:5 (DDR2-667)
1:2 (DDR2-800)

266 Strap
1:1 (DDR2-533)
4:5 (DDR2-667)
2:3 (DDR2-800)
1:2 (DDR2-1066)

333 Strap
1:1 (DDR2-667)
5:6 (DDR2-800)
5:8 (DDR2-1066)

400 Strap
1:1 (DDR2-800)
3:4 (DDR2-1066)


Performance differences will generally only be noticeable when running benches, but for best performance, go for highest FSB + highest RAM speed.

So of those you listed, 8x400 (3200 MHz) 3:4 aka 2.66 (DDR2-1064) will likely be the fastest.

That said, you'll likely find that different straps/ratios allow for a lower/higher tRD value, & this timing makes a huge difference for RAM performance.

tRD is called Static tRead Value (look under advanced memory options)
Generally the lower the strap is, the lower you can set this, but the higher the FSB is, the higher you need to set this.

Be aware messing with this will cause your system not to POST if it's set too tight, & if you're right on the edge of it being too low, more NB voltage can sometimes be the difference between stable/unstable or POST/no POST even.

Originally posted by: RonAKA
Went with DDR-2 800. First tried some Corsair XMS2, but it would not POST. This Gigabyte GA-EP45-UD3R board, which I am becoming less and less pleased with would appear to be the problem.

With Kingston Value RAM I got 3.0 GHz using the 9X CPU mulitplier, 333 FSB, and the 2.4 mulitipler on the RAM to get 800 MHz. If anything this board is not as good as my old ASUS P5B DX.

I assume you reset CMOS when you put the new RAM in?
Make sure you do that, or at the very least, you should be trying with everything stock.

As for the UD3R being inferior to the P5B-D, not even remotely close.

I own a P5B-D, & i owned a UD3P for a few weeks, & i'm sorry, but you cannot even compare the boards.
The P5B-D has a far more basic BIOS layout, far fewer options, so there are fewer ways to break things, & it's much simpler.

But in terms of overclockability, stability, & performance, they aren't anywhere close to being in the same league.

Don't get me wrong, i like my P5B-D; it was great for its time.
But if you get to understand how to work with a board like the UD3R/P, you'll understand why it's so well regarded.



 
When I said I thought the P5B DX was better than the Gigabyte board, I was mainly referring to the OC speed I got out of it. Using the same processor and same voltages, the P5B hit the wall at about 3.2 GHz, while the UD3R made 3.1 GHz. I though with a newer board I may have achieve more, not less. I backed the P5B off to 3.0 for stabiltiy, and hope the UD3R will also be stable with less of a back off to the same 3.0. It ran OCCT for an hour without errors, so perhaps it will be ok.

Certainly agreed that the Gigabyte board has more BIOS options. But it has had some quirks though that I did not experience with ASUS. It would not post with the memory in slots 2&4 while the manual says it is ok, and the ASUS had no problem with it. Similar it would not post the Corsair memory which needs 1.9 volts. I did not reset the CMOS but had the BIOS setting for memory all on auto. I would have expected the BIOS should have read what the memory needed and set it to 1.9 volts.

The UD circuit board with the 2 oz of copper stikes me as more hype than any real cooling or durability benefit. ASUS puts their copper in the heat sinks where I think it does have value, while Gigabyte only use aluminum. All in all, I think while there may be a few more bells and whistles on the Gigabyte, the overall quality is higher in the ASUS product. This is my first non Asus board after doing four ASUS builds in a row. In future I would look hard again at the ASUS line.
 
Back
Top