Sign-in / Sign-up
Your question

CPU and RAM ratio

Tags:
  • CPUs
  • RAM
Last response: in CPUs
April 16, 2009 12:54:03 AM

i have a q6600 OC at 3Ghz

FSB now 1333

my ram is OCZ 800 Mhz for a 1:1 ratio the ram needs to run at 667

(1333/4) = (667/2)

should i use the ram at 667 an have a 1:1 ratio of ram to cpu

or up the ram as high as it can go (to around 900)

for gaming i cant really see a differnce thought u guys may know what option will give best prefermance

so clock speed and forget ratio or go for lower speed and 1:1 ratio

btw i cant raise the fsb past around 1450 anyways and i wana keep my cpu at around 3ghz

More about : cpu ram ratio

a c 536 à CPUs
April 16, 2009 3:30:32 AM

If it requires lower timings at 800 than 900, then you should run it at the rated speed. You can't really see a difference between 667 and 800 or 900 because benchmarks are required to measure it.
April 16, 2009 3:46:21 AM

I prefer to keep the 1:1 going on, normally lets you get tighter timings so at least you have the satisfaction of knowing your system's running a couple of nanoseconds faster than if it was slightly outside the 1:1.
Related resources
Can't find your answer ? Ask !
April 16, 2009 10:35:43 AM

From what i know.. Running fsb to ram in a ratio of 1:1 would put lesser stress on the chipset.. Not to sure about that though.. :??: 
a c 294 à CPUs
April 16, 2009 1:20:01 PM

From the following thread - http://www.tomshardware.com/forum/247906-29-ratio#t1781...

Ah, but wait... there's more! :pt1cable: Let's consider some of the additional variables involved in the memory big picture: :o

A ratio of 1:1 provides the best level of stability, since the memory controller, which is an integral part of the northbridge chipset for Intel Core 2 processors, does not need to translate data flow across the FSB between the memory modules and the processor(s). Also, since memory and processor FSB clocks are synchronous at 1:1, (400:400 or DDR 800), there is no additional latency introduced.

If a minimal ratio of 4:5 (400:500 or DDR 1000) is used, then the resulting increase in memory frequency is effectivey cancelled out by the latency introduced in translation across the FSB between memory and processor clocks, and no increase in memory performance can be noticably detected in benchmarks. Also, asynchronous or mismatched clocks create an element of potential instability within the memory controller, so depending on the chipset, an increase in northbridge and memory voltage is required for stability, which results in more heat, and less FSB overclock ceiling.

If a more aggressive ratio of 2:3 (400:600 or DDR 1200) is used, then the increase in memory frequency can marginally overcome the latency introduced in translation across the FSB between memory and processor clocks, resulting in a marginal increase in memory performance, which typically yields an increase in memory benchmarks of 2 to 3%, and is relatively negligible in terms of overall system performance.

In the case of DDR3, where a ratio of 1:2 (400:800 or DDR 1600) or 2:5 (400:1000 or DDR 2000) is used, even with the tightest timings, an increase in memory benchmarks of only 3 to 4% is yielded over DDR2 800, which once again, is relatively negligible in terms of overall system performance. It's also noteworthy to consider that DDR2 memory timings of 4-4-4-12 compared to 5-5-5-15, will yield an increase in memory benchmarks of less than 2%. However, for those of us who have the need for speed, we'll take whatever we can tweak.

Additionally, at equal specifications, 4 memory modules offer less FSB overclock ceiling than 2 modules, since more northbridge and memory voltage is required to maintain stability, and 4 slots require twice the current of 2 slots, again resulting in more heat, which typically is pulled into the CPU cooler, where it can increase processor temperatures by a few unwanted degrees.

I hope this helps to provides a greater degree of perspective (no pun intended). :D

Comp :sol: said:
Ah, but wait... there's more! :pt1cable:  Let's consider some of the additional variables involved in the memory big picture: :o 

A ratio of 1:1 provides the best level of stability, since the memory controller, which is an integral part of the northbridge chipset for Intel Core 2 processors, does not need to translate data flow across the FSB between the memory modules and the processor(s). Also, since memory and processor FSB clocks are synchronous at 1:1, (400:400 or DDR 800), there is no additional latency introduced.

If a minimal ratio of 4:5 (400:500 or DDR 1000) is used, then the resulting increase in memory frequency is effectivey cancelled out by the latency introduced in translation across the FSB between memory and processor clocks, and no increase in memory performance can be noticably detected in benchmarks. Also, asynchronous or mismatched clocks create an element of potential instability within the memory controller, so depending on the chipset, an increase in northbridge and memory voltage is required for stability, which results in more heat, and less FSB overclock ceiling.

If a more aggressive ratio of 2:3 (400:600 or DDR 1200) is used, then the increase in memory frequency can marginally overcome the latency introduced in translation across the FSB between memory and processor clocks, resulting in a marginal increase in memory performance, which typically yields an increase in memory benchmarks of 2 to 3%, and is relatively negligible in terms of overall system performance.

In the case of DDR3, where a ratio of 1:2 (400:800 or DDR 1600) or 2:5 (400:1000 or DDR 2000) is used, even with the tightest timings, an increase in memory benchmarks of only 3 to 4% is yielded over DDR2 800, which once again, is relatively negligible in terms of overall system performance. It's also noteworthy to consider that DDR2 memory timings of 4-4-4-12 compared to 5-5-5-15, will yield an increase in memory benchmarks of less than 2%. However, for those of us who have the need for speed, we'll take whatever we can tweak.

Additionally, at equal specifications, 4 memory modules offer less FSB overclock ceiling than 2 modules, since more northbridge and memory voltage is required to maintain stability, and 4 slots require twice the current of 2 slots, again resulting in more heat, which typically is pulled into the CPU cooler, where it can increase processor temperatures by a few unwanted degrees.

I hope this helps to provides a greater degree of perspective (no pun intended). :D 

Comp :sol: 
April 17, 2009 7:54:28 AM

Quote:
Additionally, at equal specifications, 4 memory modules offer less FSB overclock ceiling than 2 modules, since more northbridge and memory voltage is required to maintain stability, and 4 slots require twice the current of 2 slots, again resulting in more heat, which typically is pulled into the CPU cooler, where it can increase processor temperatures by a few unwanted degrees.

This sentence bothers me.. Does it really affect cpu temps and overclock by using 4 dimms instead of 2 dimms..? :??: 
a b à CPUs
April 17, 2009 12:22:39 PM

Well, I can tell you from my experience, yes it does make a big difference if you run 2 or 4 sticks of memory. It has always been easier for me to overclock, tighten latencies up, run at faster memory buss speeds, lower voltages, with only 2 sticks.
In other words, yeah I think that adding more sticks does create a lot of "issues" or "circumstances" to deal with. And they get progressively worse with the faster memory you purchase.
I would also have to agree that more sticks, more heat. More heat in the case is more heat....that is all there is to that.
And yes, I would also agree from my experience that there is no noticable difference in 667, 800, 1066 memory. The faster memory you buy looking at the buss speed, if you notice the latencies get longer, which pretty much cancels out benefit of the faster buss. I also agree that about the best you can do for your money is 800mhz memory, with good tight latencies. No need for anything else.
I use 1066 memory for only 1 reason. It was on sale for $29 for a 2X2gig matched set, the day I was at Fry's picking out my parts, it was less than any 2x2gig 800mhz set they had in stock.
April 17, 2009 12:23:50 PM

I can verify that, with 2 sticks my ram can reach ~1150 MHz before failing, but with 4 they only get to ~1090, they run a little bit warmer too, 2-3 degrees difference only.
April 17, 2009 12:27:26 PM

I guess it would be the added stress on the northbridge when running 4 dimms.. Heat is definitely an important issue for me since i'm living in a fairly hot climate where a few degrees would affect system stability.. As it affects that much i might be reconsidering on going on to 8gb of ram.. :(  Might have to just get 4gb of ram and see how it performs..
a c 294 à CPUs
April 17, 2009 12:51:17 PM

Ravenica said:
This sentence bothers me.. Does it really affect cpu temps and overclock by using 4 dimms instead of 2 dimms..? :??: 
Ravenica,

Thank you for asking. The answer to your question is yes. As the author of the Core i7 and Core 2 Temperature Guide - http://www.tomshardware.com/forum/221745-29-core-core-t... - I am well acquainted with the thermal behavior of processors and CPU coolers, and the airflow charateristics of computer cases and fan specifications, as well as the effects of overclocking relevant to various Northbridge memory controllers and memory module configurations.

My primary goal here at Tom's Forums is to provide the most up to date Intel processor temperature information to Tom's readers. As such, I am not in the habit of spreading FUD or making false or misleading statements. I would not offer this information if it was not verified and true.

Ravenica said:
... I might be reconsidering on going on to 8gb of ram.. :(  Might have to just get 4gb of ram and see how it performs..
In order to bring a solid perspective to the issue of how much memory is enough, please refer to the following recent and excellent Tom's article: Do You Really Need More Than 6 GB Of RAM? - http://www.tomshardware.com/reviews/memory-module-upgra...

Don't get stuck on the fact that the article pertains to the i7 platform, but instead focus on the fact that unless it's necessary to run a 64 bit OS due to the memory requirements of a specific application, or if you're running a server, or you're into consistently heavy multitasking, then a 32 bit OS with 3Gb (4GB for Core 2 platforms) is more than adequate for the vast majority of typical users and gamers.

CompuTronix :sol: 
April 17, 2009 1:05:27 PM

Thank you CompuTronix for answering my question.. I've read about the article.. There was a performance increase in games from 4gb to 8gb ram that's what currently bothering me.. :??: 
a c 294 à CPUs
April 17, 2009 1:22:38 PM

Do you have a specific game in mind? Are you certain that you're reading the graphs correctly? I've just reviewed the article once again to be certain that I haven't missed anything. The article does not compare 4GB versus 8GB, it compares 3GB versus 6GB versus 12GB. If you look carefully at page 3 and page 6, I think you'll see that there's no appreciable difference for gaming. The differences in frame rates are confined to resolutions, not to quantity of memory.
April 17, 2009 1:26:09 PM

I believe I read an article on Tom's before about 8gb RAM, and they showed performance gains for one reason: they could turn off the page file, I don't believe they did that in this most recent article. Personally I'm happy with 8gb, some of what I do benefits from it, but the main thing is that things load much quicker due to lack of caching to disk.
April 17, 2009 1:29:48 PM

CompuTronix the article you gave me was comparing between 3gb 6gb and 12gb and they all are triple-channel kits.. The one i was referring to was most likely what Kraynor mentioned i happened to come across an article that showed performance gains when upgrading to 8gb ram from 4gb.. Although i could be wrong.. Btw Kraynor currently your not using any pagefile..?
April 17, 2009 1:34:29 PM

None at all, except when I want to play DoW2 because that requires a 1.5GB page file to load, and when working on huge rendering projects.

I haven't run into any issues with it, except for that one time in an intense 3DSMax session that went on far too long and got far too big... got an out of memory blue screen. Learned my lesson that day all right, but for general use having no page file hasn't had any negative impact and I think loading times are quicker, though I might just be trying to convince myself they are... I can't be bothered benchmarking the two to find out.
April 17, 2009 1:37:38 PM

Hmm.. If lets say you choose to use pagefile what would you set the size..? 4gb or 8gb? :) 
a c 294 à CPUs
April 17, 2009 1:40:46 PM

3GB is plenty. Also, the i7 platform shows almost no difference in performance between tripple channel or dual channel operation, or in other words, with 3 modules installed versus 2. Additionally, load times are discussed on page 7, however, load times are much more affected by fast drives in RAID 0, such as with my 3 VelociRaptor 300's, or the Intel SSD's.
April 17, 2009 1:44:22 PM

Hmm.. So am i right to say that upgrading from dual channel 4gb vs triple channel 6gb yields not much benefits..? Heh.. Can't be 3 raptors in RAID 0.. So technically 8gb of ram is most likely an overkill eh..
a c 294 à CPUs
April 17, 2009 1:49:52 PM

Correct. I'm sorry to say that after all my extensive research and testing, I've reached the conclusion that memory sales propaganda is almost a conspiracy of epic proportions. Huge piles of money is being made on the unsuspecting computer consumer.
April 17, 2009 1:53:21 PM

Heh.. Well it's okay atleast i'm able to learn something.. I have another question.. If i were to use 3gb of ram setting a pagefile size of 6gb would be dumb right as i might hinder performance more then help..?
a c 294 à CPUs
April 17, 2009 1:56:01 PM

6GB pagefile is simply waisting 3GB of hard drive space.
April 17, 2009 1:58:37 PM

Hmm.. So setting to 3gb should be plenty..? :) 
a c 294 à CPUs
April 17, 2009 2:01:33 PM

Correct.
April 17, 2009 2:05:45 PM

Thanks for all your help CompuTronix..! :D  Learned a lot today.. Since 8gb won't yield much.. I guess 2x2gb sticks would do for now.. Since my current rig uses close to 2gb.. A bit of topic here have you ever encountered problems of running prime95 blend test in vista 64bit with 2gb ram..?
a c 294 à CPUs
April 17, 2009 2:52:53 PM

Prime95 has a 32 bit and 64 bit version. Are you running the correct version?

Prime95 Homepage - latest version 25.9 - http://mersenne.org/freesoft/#newusers

Unfortunately, many computer enthusisasts don't understand that unless it's necessary to run Windows 64 due to the memory requirements of a specific application, or unless you're running a server, or you're into consistently heavy multitasking, there's no advantage to running Windows 64 instead of Windows 32, except that Windows 64 has inherently slightly better security.

In fact, many users are running Windows 64 simply because it sounds "cool", when in many instances, the drawbacks outweigh the benefits due to driver issues and problems with software compatibility, so it often present more headaches than it's worth.

Also remember that 64 bit games and applications remain extremely scarce, so when you run a 32 bit application in Windows 64 bit, it's actually is running in "WoW" mode, or "Windows on Windows". This means that your 32 bit application is running in a 32 bit shell within Windows 64, which slightly reduces system performance by an overall negligible amount of approximately 1%.
April 17, 2009 2:58:04 PM

Oh god.. Totally forgot that when i changed from 32bit xp to 64bit vista.. My bad totally forgot to recheck my stressing programs.. Hmm you say 64bit has driver issues.. That could be true a few years back when vista was first launched..? Now it seems that 64bit drivers are popping out everywhere.. Reason for going to 64bit for me was basically dx10 and more then 3gb of ram that 32bit os allowed.. I'm more into image quality is a priority over higher fps offered by dx9.. :)  I'll try the 64bit version of prime95 and hope it doesn't use up all the ram..
April 17, 2009 3:01:15 PM

I've had no issues with drivers in the last 8 months, and even then it was a small company's bluetooth usb dongle thing, a quick e-mail and 2 weeks later there was a 64-bit driver.
a c 294 à CPUs
April 17, 2009 3:01:34 PM

Vista is DirectX 10, regardless of whether you're running 32 or 64 bit. There are still driver issues with Windows 64, as well as software compatibility issues.
April 17, 2009 3:03:47 PM

Okay.. So basically my only reason left for going to 64-bit would be the 4gb of ram eh.. :) 
a c 294 à CPUs
April 17, 2009 3:07:39 PM

Exactly, however, dual channel 4GB on the Core 2 platform versus tripple channel 3GB on the i7 platform is negligible in terms of performance.
April 17, 2009 3:09:36 PM

Heh.. i7 platform itself is a force to be reckon with.. Either ways i would need the 4gb of ram.. 2gb is just isn't sufficient man..
April 17, 2009 3:14:46 PM

Okay just tested even using the prime95 64-bit still uses up 100% of my ram.. Guess i'll have to wait for 4gb of ram..
a c 294 à CPUs
April 17, 2009 3:29:41 PM

For Windows 32 with 4GB, Task Manager will typically show soemwhere between 3.0GB and 3.5GB. Since Windows 32 can map up to a total of 4GB, subtract your graphics card(s) memory from 4GB, and that's what remains. If you're running SLI or CrossFire with more than 1GB total graphics memory, then Windows 64 is preferred.

For example, I've built my rig around the highest possible performance for Flight Simulator X, which requires massive amounts of multi-threaded-overclocked CPU horsepower. FSX actually runs slower with multiple GPU's due to the additional clock cycles required to crunch frames for rendering on multiple cards, so in this instance, frame rate is higher on a single card.

Upgrading from XP DX9 to Vista DX10 will result in a 3% reduction in frame rate, and running Vista 64 instead of Vista 32 will result in a further reduction in frame rate of 1%.

FSX is also a memory pig, I'm running Vista 32, and when I tested Vista 64, it was using 2.0GB during the simulation, while Vista 32 was using 1.8GB. The difference of 0.2GB was being used for crunching Vista 64's extended code.
April 18, 2009 4:51:23 AM

Hmm.. So if i were to run an SLI setup with 4gb of ram installed on a 32bit os.. I would roughly left around 3gb of ram or maybe lesser if the graphics card have higher video memory..?

Regarding the XP DX9 to Vista DX10 the performance reduction is negligible like you stated.. But i have read that DX10.1 have shown to increase game performance.. It's a shame that Nvidia cards still do not support DX10.1 and the upcoming DX11..

So it seems that 64bit os has more overhead as it uses more ram for it's codes..? But in terms of system efficiency i feel that 64bit os seems to run faster.. I might be wrong though..