What's your memory assessment speed?

[Swimaf]

Okay i am back to report how it works with 4Gb instead of 2Gb, here are the results of the winsat mem test 'before' and 'after':

RAM frequency = 400Mhz (1:1), 5-5-5-18, 1.8v:

With 2Gb: 4800Mb/s
With 4Gb: 7200Mb/s

RAM frequency = 533Mhz (3:4), 5-6-6-16, 2.1v:

With 2Gb: 6400Mb/s
With 4Gb: 8000Mb/s

We have our answer cabal06ca! Indeed the quantity of installed RAM plays a great role on the overall performance
It seems quite logical since you have 8Gb and get more than 16000Mb/s! But the problem is that I do not multiply by two my old score (with 2Gb)... There must be an issue, probably with the quality of the Motherboard, actually mine is not a good horse at all so it may explain that sum.

 

[cabal06ba]

No, your board is running fine! You can't expect a linear performance boot with more RAM installed, it doesn't work like 2GB: 6400 --> 8GB: 6400x4=25600Mb/s...just having more RAM doesn't mean that the system can use it 100% efficiently, there are certain limits to everything a chipset can do and as I said before there are many other factors influencing the results.

Your results are good, don't worry about it anymore!

For comparison, with the intel core2duo mobile platform in my laptop (T7700 2x2.4 Ghz) I get 3650Mb/s with 4GB 667Mhz 5-5-5-18.

 

[Swimaf]

Your results are good, don't worry about it anymore!

Sounds difficult when you are a perfectionist like I am ^^

Anyway, I am glad to hear that my system is running fine and without important issues. Last question: according to you should I stay at 400Mhz or continue in the Enhanced Performance Profile which is 533Mhz and means the higher voltage for the modules? Should I worry about that or let it go too?

PS: WEI RAM score before: 5.5
WEI RAM score after: 7.0 !!

 

[cabal06ba]

Sounds difficult when you are a perfectionist like I am ^^

I know, it's just too much fun to squeeze every single last drop out of your system so you can go to bed with the pleasant feeling that you have done EVERYTHING possible to make your PC perform as good as possible.

...but once in a while one should keep in mind that many of those optimizations don't even have a noticeable impact on real usage experience; in many cases it's just about numbers in benchmarks we don't even know what they really mean.

Since your results show a significant difference I would stick to the EPP settings, 2.1V is not really high for quality RAM (mine is certified to run at 2.3V) and those +0.3mV will cost you about 50 cents on the power bill for an entire year of day and night PC usage.

bottom line: I only recommend going for lower frequencies if that allows you to run really tight timings (e.g. 4-4-4-12) stable, otherwise it's just a waste of performance.

 

[Swimaf]

You are certainly right, I'll let the settings like that (till the next crash, what I'd prefer avoid ^^).
Perhaps change just the timing and see wheither lower values may work; however 5-6-6-16 2.1v are the recommended ones in EPP and show an increase of more than 10% campared to the 1:1 400Mhz 'simple' mode.

I can feel that I will have a good rest after all these frivolities

 

[cabal06ba]

If you wanna make sure there won't be any future crashes, let Memtest86+ run over night while you enjoy your well-deserved rest: Memtest86+ - Advanced Memory Diagnostic Tool

 

[Swimaf]

I had tried to run Memtest86+ with a bootable USB key but failed...
Now I try with a floppy (unforgetable piece of plastic!), I will post the results as soon as I get them

 

[cabal06ba]

No need to post any results, this program is quite simple:
When you wake up and see lots of 'red' on your screen, you (might) have a problem, if there's no red you're good to go!

 

[ZaLiTH]

We have our answer cabal06ca! Indeed the quantity of installed RAM plays a great role on the overall performance
It seems quite logical since you have 8Gb and get more than 16000Mb/s! But the problem is that I do not multiply by two my old score (with 2Gb)... There must be an issue, probably with the quality of the Motherboard, actually mine is not a good horse at all so it may explain that sum.

I've been reading through your posts starting on the previous page here, very interesting read actually.. First off; Swimaf, glad you managed to get your RAM speeds up, it's always a nice feeling when you push your PC that little bit harder and it works.

But just to clarify, it wouldn't be the quantity of RAM installed that affects your speed that much, but rather the number of idenical RAM modules. If you install two RAM modules of the same size and speed in a dual-channel configuration (it gets enabled automatically by 90% of motherboards), then you effectively double the available bandwidth from your memory controller to the RAM modules themselves. If you want to read up more on dual-channel, click here.

 

[Swimaf]

If you install two RAM modules of the same size and speed in a dual-channel configuration (it gets enabled automatically by 90% of motherboards), then you effectively double the available bandwidth from your memory controller to the RAM modules themselves

CPU-Z tells me that the two modules actually work in dual channel, however I do not double my score with 4Gb DC instead of 2Gb SC.. Could it be a problem or is it normal as cabal06ca said?

However I found it when I was reading the wiki link you gave me:

Tom's Hardware found little significant difference between single-channel and dual-channel configurations in synthetic and gaming benchmarks (using a "modern" system setup). In their tests, dual channel gave at best a 5% speed increase in memory-intensive tasks.

Hopefully for me the gain is quite higher than 5%! I know there is nothing else to do for now but I cannot be 100% satisfied with my score...
I've also read (I have got big ears) that the difference between the performance you can get with DDR2 and DDR3 are very small but wheither it is wrong, means that DRR3 would be much more efficient than its old brother, then I'd understand certain values some people here manage to get with their 8Gb DDR3 modules.

No need to post any results, this program is quite simple:
When you wake up and see lots of 'red' on your screen, you (might) have a problem, if there's no red you're good to go!

I get a nice error when I try to run Memtest86+ in order to make a floppy, even if I already read almost everything about it: the program does not recognize the files in the 'boot' folder (I run Install64 though...). I will retry later.

I've been reading through your posts starting on the previous page here, very interesting read actually

Thank you!
Actually I think it can be useful for other people on this board (and not only btw) to know how all of that works

 

[ZaLiTH]

If you install two RAM modules of the same size and speed in a dual-channel configuration (it gets enabled automatically by 90% of motherboards), then you effectively double the available bandwidth from your memory controller to the RAM modules themselves

CPU-Z tells me that the two modules actually work in dual channel, however I do not double my score with 4Gb DC instead of 2Gb SC.. Could it be a problem or is it normal as cabal06ca said?

Ok, so as I suspected it is working in dual-channel after adding the second module. It won't double your actual access speeds, computer hardware is limited by the slowest relative component; therefore if the physical memory modules are capable of processing memory only slightly faster than the available bandwidth allows, dual-channel won't make much of a difference. If the physical memory modules (more noticable with high RAM speeds and / or tighter timings) can process much faster than bandwidth allows, dual- or even tri-channel will make a more noticable difference.

It's not a 'problem' as such with any components, it's more to do with each individual component's own limitations and capabilities. At the end of the day, it comes down to how well each component works along side the next.

However I found it when I was reading the wiki link you gave me:

Tom's Hardware found little significant difference between single-channel and dual-channel configurations in synthetic and gaming benchmarks (using a "modern" system setup). In their tests, dual channel gave at best a 5% speed increase in memory-intensive tasks.

Hopefully for me the gain is quite higher than 5%! I know there is nothing else to do for now but I cannot be 100% satisfied with my score...
I've also read (I have got big ears) that the difference between the performance you can get with DDR2 and DDR3 are very small but wheither it is wrong, means that DRR3 would be much more efficient than its old brother, then I'd understand certain values some people here manage to get with their 8Gb DDR3 modules.

I must admit that can be true in certain situations, however (as I mentioned above), it depends on many different components in your build. Certain combinations of components simply work better together than others...

As a prime example, here's a post of DDR3 RAM not performing at it's peak (link), and two examples of DDR2 clearly outperforming the DDR3, first with overclocking (link) and also without (link). Makes you wonder, doesn't it..

Edit: By the way Swimaf, I like your GPU... When I first went looking for an 8800GTS or GT I wanted the 512mb G92, and when it arrived I realised it was only the 320mb G80. You got yourself a nice card there...

 

[smarteyeball]

I've only glossed over the last few posts, so excuse the intrusion - but as Zalith has pointed out, your RAM is indeed running in dual-channel and for those 5-6-6-16 timings, pretty much right on the money for speed with your current settings there Swimaf.

Outside of the difference between single/dual/triple channel, (which was reflected by the difference between 2GB single and 4GB dual) ,the biggest impact on overall bandwidth is raw mhz speeds.

Simple rule of thumb, the higher the mhz, the higher the overall bandwidth.

Tighter timings also have an impact too, particularly on the latency response times, but not as much between the difference of single vs dual channel or running with a higher clock speed.

So unless you are getting errors, or chasing a few MB extra + a few nansoeconds quicker, you're good to go

 

[Swimaf]

It's not a 'problem' as such with any components, it's more to do with each individual component's own limitations and capabilities. At the end of the day, it comes down to how well each component works along side the next.

I suddenly feel like owned by my computer
Anyway, let's stop crying about it, that doesn't count very much but for benchmarking, cabal's right.

As a prime example, here's a post of DDR3 RAM not performing at it's peak (link), and two examples of DDR2 clearly outperforming the DDR3, first with overclocking (link) and also without (link). Makes you wonder, doesn't it..

It's an euphemism! How the hell may the difference be so important between two systems? :shock:

By the way Swimaf, I like your GPU... When I first went looking for an 8800GTS or GT I wanted the 512mb G92, and when it arrived I realised it was only the 320mb G80. You got yourself a nice card there...

I love it too! 8800GT is (was) the most common GPU among the world of computer gamers and addicts Purchased long long time ago and still making great new games run! 'She' is the only one 'who' has never disappointed me

EDIT:

So unless you are getting errors, or chasing a few MB extra + a few nansoeconds quicker, you're good to go

It reinforces what I've finally started to think, thank you

so excuse the intrusion

I find no reason to blame you, it's an open thread, if there is one person who is intruding it can be only me!

 

[cabal06ba]

1.) performance boost with more RAM and single- vs. dual-channel always depends on what you're testing and how you're testing it. For example, if Tom's Hardware used Futuremark's 3D Mark and PC Mark ("synthetic and gaming benchmarks") you have to keep in mind that those benchmarks test overall system performance so you can expect the results to reflect something like this (exemplary)
- GPU performance 35%
- CPU performance 30%
- RAM performance 20%
- chipset, HDD and everything else 15%
So even if you boost your RAM performance significantly it will not show that much impact on the overall score since other components determine most of the scores.

2.) DDR2 vs. DDR3: until today, the performance difference between modules with the same speed (DDR2-1066 vs. DDR3-1066) is marginal, you can see that when you compare the results in this thread. However, also here you'll see significant differences in dual- vs. triple-channel mode. That DDR3 works more efficiently simply means that it works with lower voltages and therefore consumes less power and produces less heat for the same performance.
bottom line: at the moment you will hardly feel a difference in every-day-usage - and even in gaming experience - between DDR2 and DDR3 but this will change in the future with more demanding software and games, so it's still ok to invest in good DDR2 RAM until the time has come to go for a totally new system (mainboard + CPU + RAM) which in my case will not happen for the next two years and by then DDR2 will be a thing of the past anyways.

3.) the same applies to 4GB vs. 8GB: basically right now there is no point in having 8GB of RAM unless you do some hardcore video and graphics editing. I actually only got the extra 4GB for my system because I got it dirt-cheap on ebay and hope that it will get me through the next two years without any further upgrades.

EDIT: wow, you guys are quick! (or I'm too slow)
Just wanted to add that you should burn the MemTest86+ image on a CD if you have the chance, I don't think it's supposed to run from floppy.

 

[Swimaf]

According to you everything would be a question of tests, their nature and how they are used to put a mark over different hardware devices, not only the devices themselves and their performance. It sounds logical indeed.

 

[cabal06ba]

maybe this the teacher's syndrome in my case: Does a student have high language skills if he gets an A+ in an English vocabulary test? :sarc:

 

[pacinitaly]

thanks for the tool.

mine is 49secs on my win7 laptop and 55sec on my vista laptop


edit: please delete post, I thought it was for the reboot VB script

 

[Swimaf]

maybe this the teacher's syndrome in my case: Does a student have high language skills if he gets an A+ in an English vocabulary test? :sarc:

Hey, I am this student! You're quite right, the fact that you get an A instead of a B does not depend only on the natural skills (a large part does however) but on many other environmental features which, when they are put together, make a whole entity.
Btw your example is a very good one, I did not realize that I was a module of RAM before :shock:

 

[cabal06ba]

@pacinitaly: even though you hit the wrong thread, those results are still impressive, especially on a laptop!

 

[cabal06ba]

@ swimaf:

the fact that you get an A instead of a B does not depend only on the natural skills (a large part does however) but on many other environmental features

That's certainly true and additionally a lot also depends on criterial validity (does a test actually measure what it's supposed to find out): you can perform quite well in a vocab test when you study hard, but that doesn't mean you can actually produce one single correct sentence in English!

I did not realize that I was a module of RAM before

you are even better than RAM cause upgrading only costs time and no money!
...and you'll never have a problem with incompatibility due to new socket standards.


EDIT: now that we are all settled and satisfied with our RAM settings, here are some other results to talk about in the respective threads:
ReBoot Time
hard drive performance (@Swimaf: got me beat there! )
WEI
Reliability index score