Looking for an upgrade for RAM

[infinitesadd]

I currently don't run very good RAM and i'm looking for an upgrade. I think my mobo can only handle up to DDr2-800. It is an epox mf570 sli. I am wondering what brand to get a couple gigs from. I currently have 2 gigs of OCZ SOE PC-667.

I also have a somewhat newb question about windows xp, i have home edition and service pack 3, what is the ram limit to windows xp, I was told it was 3.12 gigs, but microsoft's website says 4?

thank you for your help!

the rest of my system is in my sig.

[itznfb]

are you looking for OC the RAM? if not i'd go with G.Skill F2-6400CL5D-2GBNQ

WinXP Pro 32-bit RAM limit is 4gb. your pagefile takes some of the available, some cases so does video mem, virtualmem. so you end up with somewhere around 3.1 -3.6gigs.

whats wrong with the ram you have?

[Utnorris]

Here is a deal on 4gb of GSkill ram for $75 shipped from Newegg. These are rated at 1000Mhz, so its a pretty good deal and you can run them at 800Mhz with tight timings.

http://www.newegg.com/Product/Product.aspx?Item=N82E16820231145

Verify compatibility with your motherboard in the Gskill support forums.

Utnorris

[infinitesadd]

itznfb;2988879']are you looking for OC the RAM? if not i'd go with G.Skill F2-6400CL5D-2GBNQ

WinXP Pro 32-bit RAM limit is 4gb. your pagefile takes some of the available, some cases so does video mem, virtualmem. so you end up with somewhere around 3.1 -3.6gigs.

whats wrong with the ram you have?


Nothing is really wrong. I was just hoping to get a bit of a performance increase, and yes, overclock it a bit. They don't sell my current RAM anymore, so I was going to get 4 gigs of something new.

[itznfb]

the ram Utnorris linked is excellent. i have a set they run at 1066 easy.

[little_scrapper]

Rams dont make the machine anymore like they did back in the DDR400 days. Keep in mind that the responce time of DDR667 3-3-3-8, DDR800 4-4-4-12, DDR1000 5-5-5-15 is all the same!!! The newer memory has more bandwidth yes, but its not faster. It will only help if you need bandwidth and only VERY FEW apps need more bandwidth to excel.

[itznfb]

^ that makes no sense

[little_scrapper]

itznfb;2989973']^ that makes no sense

What part of it doesnt? Maybe I didnt say it clearly enough. Calculate the latency/responce time in nanoseconds of those three at those timings and tell me the difference.

All Im saying is that something like DDR1600 may have a faster clock speed but with timings like 7-7-7-20 its only about the same responce times as DDR800 4-4-4-12.

[itznfb]

well... i don't think 1000 5-5-5 = 667 3-3-3 performance wise. i could be wrong.

but for things like gaming and photoshop and video editing they are very bandwidth dependent, which are the types of apps we use as an excuse to spend thousands of dollars on new hardware.

[little_scrapper]

itznfb;2990972']well... i don't think 1000 5-5-5 = 667 3-3-3 performance wise. i could be wrong.

but for things like gaming and photoshop and video editing they are very bandwidth dependent, which are the types of apps we use as an excuse to spend thousands of dollars on new hardware.

http://www.tomshardware.com/reviews/ddr3-1333-speed-latency-shootout,1754-23.html

That link shows FPS results for DDR3 rams at 1066, 1333, and 1600. All the same.

Consider the latency ratings of the three most recent memory formats: Upper-midrange DDR-333 was rated at CAS 2; similar-market DDR2-667 was rated at CAS 4 and today's middle DDR3-1333 is often rated at CAS 8. Most people would be shocked to learn that these vastly different rated timings result in the same actual response time.

Because cycle time is the inverse of clock speed (1/2 of DDR data rates), the DDR-333 reference clock cycles every six nanoseconds, DDR2-667 every three nanoseconds and DDR3-1333 every 1.5 nanoseconds. Latency is measured in clock cycles, and two 6ns cycles occur in the same time as four 3ns cycles or eight 1.5ns cycles. which is specifically 12 nanoseconds. If you still have your doubts, do the math!

[itznfb]

do they tell you what timings for each setting were used in the benchmarks?

[little_scrapper]

http://www.tomshardware.com/reviews/ddr3-1333-speed-latency-shootout,1754-20.html

yes. They used the best "stable" timings they could achieve for each fo the 3 speeds.

[adamsleath]

it makes no 'cents' :rofl:

http://i.neoseeker.com/a/ddr2_vs_ddr3/ut2004.png
http://www.neoseeker.com/Articles/Hardware/Reviews/ddr2_vs_ddr3/11.html

[little_scrapper]

Thank you adam. That chart illistrates what I am saying very well.

Look at the top two lines, with the processors running 458x7 the difference between running DDR3 1552 and DDR2 970 is 6 fps!!! 3% you cant even FEEL 3%.

Now look at the middle two dark blue lines. At nearly identicle timings the extra 30% clock speed gains 2.5 fps, thats less then 2%.

If you can afford it and want to add +1 to your e-penis then buy the highest DDR3 you can get and then shell out another $300 for the mobo that can handle it.

If your like me, not rich, buy some "GOOD" DDR2. The price of DDR2 is dropping like a stone and you can get 4GB of nice stix for about $100. My Ballistix do DDR1200 4-4-4-12 and can be bought these days for about $50.

Dont mean to get all preachy or anything, I just think that the extra $200 can be spent on something more appropriate like hookers and beer.

[zanzabar]

Rams dont make the machine anymore like they did back in the DDR400 days. Keep in mind that the responce time of DDR667 3-3-3-8, DDR800 4-4-4-12, DDR1000 5-5-5-15 is all the same!!! The newer memory has more bandwidth yes, but its not faster. It will only help if you need bandwidth and only VERY FEW apps need more bandwidth to excel.

ur a little off 667c3 is about 800c4 and 1066c5, but that wont really matter since 667c3 no longer exists unless u make a custom order

but NS delay is the same so if u have the bandwidth from the MC to the ram u want as close as u can with a little over

[xMrBunglex]

performance memory does make a difference - just not a huge difference. don't expect your computer's overall performance to feel faster just because you drop some money on RAM. and for most common memory-hungry applications, bandwidth is more important than latency.

[little_scrapper]

performance memory does make a difference

:shakes:

No. In the real world applications it really makes such a small difference that you wont even notice it. About the only time you can even SEE and difference is by using synthetic benchmarks that have no bearing or correlation to real applications. Not until DDR3 timings are at DDR2 levels will you start to see any marked difference. Look at the two middle dark blue lines on adams chart again. Nearly identicle timings but the memory running 30% faster only gets 2.5 fps more. Hell thats probably less than std. fluxuation. 2.5 fps @ 135 fps is nothing. there the same. If you want any really marked performance increase your only going to see it with faster CPU and better Vid cards.

Now, the real arguement for high end DDR3 is that the newest chipsets that take it are roughly 6-10% more performance over my P965 chipset. That coupled with some of the best DDR3 that can reach high clocks without having to use rediculously slow timings can add another coup[le percents to the performance. So in the end you can feasable get ...well I go as far as saying 10%-15% better performance just from mobo and DDR3. But the premium for 10% performance is about 100-400% in cash money.

again Im NOT rich so my $100 chipset being within 6% performance of your $300 chipset is very easy to live with. And i can spend the other $200 on my brand new Fusion7 Dual HD tuner card and still have enough left over to take my kids to the zoo for the afternoon.

[itznfb]

for the sake of argument lets say the lower timing with lower clocks and higher clocks with higher timings even out as described by little_scrapper. i'm not implying you're right or wrong. i'd like to do more research on my own and do some tests.

but... saying you're spot on... performance memory will still make a difference as the case would be, the higher end memory you get, the higher you can get the clocks, with lower latencies. for example... getting 2x1 gb for $40 will get you what? 800 @ 5-5-5-15. using gskill F2-6400CL5D-2GBNQ as an example, i can't get cas4 @ 800 with the couple sets i have. now if you buy performance/gaming memory, spending around $150 - $190 you can find sets that will do 1066 @ 4-4-4-12.

will this not yield a significant boost in performance? i'm not saying it definitely will, i'm looking for an honest informative answer. in my opinion buying performance memory is a benefit for the reasons i stated. am i wrong?

[little_scrapper]

itznfb;2993462']for the sake of argument lets say the lower timing with lower clocks and higher clocks with higher timings even out as described by little_scrapper. i'm not implying you're right or wrong. i'd like to do more research on my own and do some tests.



Absolutely. Do the math and/or find other articles on this. At those standard speeds/timings you get a responce time of 12nanoseconds.



itznfb;2993462']but... saying you're spot on... performance memory will still make a difference as the case would be, the higher end memory you get, the higher you can get the clocks, with lower latencies. for example... getting 2x1 gb for $40 will get you what? 800 @ 5-5-5-15. using gskill F2-6400CL5D-2GBNQ as an example, i can't get cas4 @ 800 with the couple sets i have. now if you buy performance/gaming memory, spending around $150 - $190 you can find sets that will do 1066 @ 4-4-4-12.



My Ballistix will do just shy of 1200 @ 4-4-4-12. But I dont trust em there. I RUN them at DDR888 @ 3-3-3-9 and they work awsome at 2.0V. The last time I saw a price on these they were shipping for about $50-60.


itznfb;2993462']will this not yield a significant boost in performance?

In short, no.

All the atricles I have read would seem to indicate that memory alone, even uber high end DDR3 at the tightest possible timings, only relfect a small margin of performance gain. In terms of overall system performance its less then 5%. And realistically it looks like you'd be lucky to be getting 3% from ram alone.

http://www.tomshardware.com/reviews/high-end-ddr3-memory-hook,1758-11.html

look at those charts, the Corsairs dominators 1800 vs the DDR1352. EXACT SAME TIMINGS. The 1800 is running about 35% higher clock speed and you get about 3% increase in FPS. And you have to go to the synthetic benchmarks to even see 10% performance gain from 35% increase in clock speed. In the real world apps that it does matter on, it looks like the 1800 is getting a mere 5-6% performance boost.

[itznfb]

In short, no.

All the atricles I have read would seem to indicate that memory alone, even uber high end DDR3 at the tightest possible timings, only relfect a small margin of performance gain. In terms of overall system performance its less then 5%. And realistically it looks like you'd be lucky to be getting 3% from ram alone.

if i have time this weekend i'm going to run some benches. it just seems kind of ridiculous that there has been so much emphasis on ram speed if it makes such small difference.

[little_scrapper]

itznfb;2993712']if i have time this weekend i'm going to run some benches. it just seems kind of ridiculous that there has been so much emphasis on ram speed if it makes such small difference.

Yea it would seem that way. But I have read at least 6 or 8 articles that seem to do a fairly good/comprehensive comparison of different rams and they tend to all have very similar results. Just on Toms Hardware alone there is like 3 good reads in the RAM section from like the last year that have decent ram comparison charts.

But other review sites have done similar articles with similar conclusions. Im not just basing my ideas exclusively on Tom's Hardware articles.:up:

As for your benches. I definately am interested in your results. Do you by chance have a MOBO that takes DDR2 AND DDR3? If so, I would like to see a couple results of tightest timings w/ highest speeds. And, stock speeds with the twos timings set to the same thing. So basically the DDR2 would be set to very slow timings. I am always interested in more data points that I can refer to.

[adamsleath]

if i have time this weekend i'm going to run some benches. it just seems kind of ridiculous that there has been so much emphasis on ram speed if it makes such small difference.
ram speed doesnt seem to be a significant bottleneck, dats why i get the cheapest stuff now :yawn:
cpu and gcard ok, but ram is boring:lol:

but if you get 3%+, all the best...:| - about as much as doubling your cpu cache:rolleyes:

i got sick of throwing money away on expensive rams;) fortunately i was able to sell them off.

[itznfb]

so all this time i've been a whore to the marketing... lol

[xMrBunglex]

video games? i never said performance memory should be used to increase your fps in video games. people use computers for other things. the video card is the obvious bottleneck playing any modern game anyway.

[Serra]

Let me give an example to illustrate this.

DDR2 @ 800MHz CAS 4 = DDR2 @ 1000MHz CAS 5

The reason is that with CAS 4, you have to wait effectively 4 cycles to access data... at CAS 5, 5 cycles. There is some fine-tuning that you can do and a few things beyond this, but this is where the big effect is.

Now,
1 second/800MHz = 1.25ns access time for DDR2 800*
1 second/1000MHz = 1ns access time for DDR2 1000*

So to combine the above...
4x 1.25ms = 5ns access time
5x 1.00ms = 5ns access time

*Yes, I know DDR2 uses both the start and end clock so I should have doubled frequency, but I'm keeping things simple ok? The numbers still work out the same.

See the basic idea? Overclocking is good, but getting low times is crucial to actually making the numbers go zoom.

[xMrBunglex]

show me a computer with cheap RAM encoding video at 350 frames per second and i'll retract my earlier statement.

http://i21.photobucket.com/albums/b267/davedave01/compiling.jpg

memory does more than just synthetic benchmarks.

[little_scrapper]

show me a computer with cheap RAM encoding video at 350 frames per second and i'll retract my earlier statement.



http://xtreview.com/review217.htm

scroll to the bottom where they show the applications results. DDR1000 to DDR1600 the only single charts that show any real diff at all is the everest charts. and thats a synthetic I believe. The rendering and encoding charts show no diff.:up: But I do like how thaty say the DDR3 has higher performance in the encoding: then I look to see its 1 point at nearly 80 points. hehe 1.2%

http://www.tomshardware.com/reviews/high-end-ddr3-memory-hook,1758-11.html

Not sure if Xvid equates but this show that video compression, at DDR 1352 and DDR 1800, using the exact same timings, has a 6% performance difference. Still that is only one example. If you use Xvis well there you go!

http://www.tomshardware.com/reviews/ddr3-1333-speed-latency-shootout,1754-24.html

3D studio max? Not sure about this program either but I believe its a content creation software. Somewhat similar?? maybe not but its in the ballpark anyways.

http://www.xbitlabs.com/articles/mainboards/display/ga-p35c-ds3r_11.html#sect0

There is some video encoding with 1333 vs 1066. no diff.

I doubt your ram has much to do with that figure you stated above. No trying to start a fight I'm just trying to save alot of peoples wallets.

[xMrBunglex]

i'm not a believer in big dividers either. i actually run my memory underclocked @ DDR 1480 because that's my sweet spot. memory-intensive applications run faster for me at 1480 with tight timings than they do at 1780 with looser timings. but memory is part of the equation. i used to run DDR2 800 memory with this processor before i got the new motherboard and i've run tests. this RAM has allowed me to OC my CPU further and its extra bandiwdth makes my computer noticeably faster. i also have 50% higher memory bandwidth and tighter (relative) timings than i did with the DDR2 800. it has no affect on games though.

[little_scrapper]

i also have 50% higher memory bandwidth and tighter (relative) timings than i did with the DDR2 800. it has no affect on games though.


Quite correct and generally supports my first post regarding that very thing.

I win!:up: Everyone else who bought DDR3, enjoy it. Because for the price you paid, I am enjoying the exact same performance from my DDR2, AND!!! brand new FusionHDTV7 Dual Express, case of beer, bottle of patrone, and the latest issue of greasy hooters magazine!! :) and I got the wife a... ahem NM.

I guess this means I'm NOT extreme. But I'll take a night with the wife over DDR3 any day!! Hehehe

[xMrBunglex]

i'll take a night with your wife over DDR3 any day too. fortunately, i got to have both.

if you built a PC just to play video games and you think you actually were economical about it, think again. you'd have been better off just buying an X-Box 360. RAM has nothing to do with frames per second. everyone already knows that.

[little_scrapper]

i'll take a night with your wife over DDR3 any day too. fortunately, i got to have both.



You like ugly fat broads too? hehe :rofl::clap: