2 gigs or 3 gigs of RAM

see specs below, do you think I would do better for gaming with 3 gigs?It will be a 70 dollar difference which isnt that much but still. Will it help a lot?

 

yes it will, certainly for gaming

 

You will get a much bigger performance increase if you upgrade your CPU.
Spend your money there.

 

ugh okay, so whats the difference between a 4mb cache and a 2mb cache?

 

On Vista basic 2GB is already above the reccomended and should do fine. The suggestion to upgrade the CPU is a good tip in general as harder to do later than either RAM or HDD. But it is not needed and with your config, the GPU is going to be the limiting factor and you can't upgrade that so for gamming well balanced system right now.

I read some tests 4MB vs 2MB L2 about 10% improvement at same clock on CPU intensive tasks so I would estimate about 30% to 35% 1.6 vs 2.0 but only on the absolute most demanding CPU tasks not gamming.

 

for some reason 3GB gives a substantial increase when gaming compared to 2GB. Getting a new CPU won't increase it much, except in Supreme commander,

 

Is that true on home basic just like premium?

 

No differance in performance between the different versions.

 

3 GB > 2 GB
but I did't look at the CPU: you might want to upgrade the CPU to a T7300 (with 4 MB cache) ... that'll be overall more helpfull (while the extra RAM will primarely be noted in games)

 

Answer NO 2X1GB is fine all who have opinions have just that! Your computer is fine and balanced so unless you want a different one get this!

 

Your config is fine as it is. Only games like supreme commander will benefit from a faster cpu, most it will barely make any difference and won't be worth the upgrade costs. 3GB ram is probably unecessary right now for 32 bit gaming, 2GB is fine.

Your setup is great as it is.

 

Ya I guess I can always upgrade If I really need it

 

2 x 1 GB IS fine
but 2 + 1 GB is better
there's no real discussion about that

as soon as you use a program that's a bit heavier than firefox, you'll notice a difference (small or big, depends on the program you're using)

 

For those who are worried that they loose RAM bandwidth with3GB, I have just made some measurements on my Zepto 6024W (Intel 965GM) using the SiSoftware Sandra. I have tested 3 scenarios:

2 x 1GB PC5300 read/ write = 4012 / 4021 MB/s
1 x 2GB PC5300 read/ write = 3685 / 3671 MB/s
1GB + 2GB PC5300 read/ write = 3919 / 3879 MB/s

These are the best of several tests. which can vary by around 5%.

It can be seen that having only one module reduces the bandwidth by about 9% but having two modules results in speed reduction of around 3% compared to having two identical modules.

Also included for comparison is the result I measured on the Samsung Q70 (Intel 965PM = dedicated graphics) which scored, for 2 x 1GB, 4321 / 4318 MB/s. It is possible that this additional speed (~7.5%) is the result of not having the GPU sharing the system memory. Or maybe the computer was better tuned (the CPU gave a very good SuperPi score).

John

 

I would get 3gb, also get Vista Home Premium.

 

Mr. Ratsey outstanding presentation and you motivated me to update my copy of SiSoft. Kind of hard to argue with numbers. So you've got me considering that it really does not matter speed wise running 3GB and you have 50% more RAM than 2GB. I have to see 1 test run, the cache and memory benchmark. In your test I am troubled by the fact that the results are below the theoretical bandwidth of the RAM running 1 stick on a 64 bit bandwidth let alone in dual channel using 128 bit. Is there any chance you could redo your test using the the Cache and Memory benchmark? I get 7365MB on that running PC4300 (533) 2X1GB. If you run 1X2GB and it caps at 4300MB or less then we know we have a better test. And just a side note running my 533 RAM I score about 3950MB most recent SiSoft on the test you ran so about 50MB diff from your 667 RAM? That is a 1% difference against RAM clocked 20% faster, and us AMD guys have to enjoy any benchmark we do well on. I can not do a credible test like you because I only have 2 1GB sticks.

 

bumped for John

 

Yes, I've seen your request.

Does the attached make sense to you? If so, then later on will swap the RAM and do the other tests.

However, as I understand it, the Cache memory benchmark is measuring the CPU cache speed. I have the T7300 so you will see the step change in performance once the block size exceeds the 4MB on-board cache.

John

 

I read it all very throughly and put mine next to it and compared. When you ask do I understand? All of the words in and of themselves and what they refer to yes sort of, but do I grasp and comprehend, not completely and all the implications no way. I concede this test may not be the correct for what I am trying to determine but thought and think it might in fact tell us something. The warning you have in the bottom “Warning 2544 : System/Video shared memory greatly reduces performance. Use external video card.” Only peaks some of my suspicions that this test might be useful for RAM. But I admit I may not have a clue, that is why I wanted to test. The text files of your tests would be interesting to compare more info than graph alone.

If you do run tests maybe redo 1st w/1X1GB? Just to see, shouldn't matter as test only used 512MB but once again just to see.

Would like to see this in the 1GB+2GB:
Logical/Chipset 1 Memory Banks
Bank 0 : 512MB DDR2-SDRAM 5.0-5-5-15 (tCL-tRCD-tRP-tRAS) CR1
Bank 1 : 512MB DDR2-SDRAM 5.0-5-5-15 (tCL-tRCD-tRP-tRAS) CR1
Bank 4 : 512MB DDR2-SDRAM 5.0-5-5-15 (tCL-tRCD-tRP-tRAS) CR1
Bank 5 : 512MB DDR2-SDRAM 5.0-5-5-15 (tCL-tRCD-tRP-tRAS) CR1
Shared Memory : 8MB
Channels : 2
Memory Bus Speed : 4x 333MHz (1332MHz data rate)
Width : 64-bit
Performance Acceleration Technology : No
Memory Controller in Processor : No
Maximum Memory Bus Bandwidth : 21312MB/s (estimated)

 

I, too, would be more concerned with the CPU than with RAM in this case. 2 Gb of ram is a reasonable amount (I'm running XP still tho, so your millage may vary). OTOH, More CPU is often a better return on investment. As for your earlier cache question, which I didn't see any answer to (skimmed - sorry if I'm maligning anyone). If that's a dual core (looks like it is) then you want 4 mb cache (effectively 2 for each).

Back in the day 512Kb cache (over 256) was always traded for extra ram (512 instead of a gig, usually). Now a days, it may not make as much difference, but old habits...

 

id put some money on a 17 inch screen if ur looking for heavy gaming... how can u play tommorows games on a 15inch screen!

 

rofl, that's really no problem
atm I gaming on a 15,4" @ 1900x1200 ... my next laptop will be a 17", but 15" works just fine (I just want bigger )

 

OK, I attach the Cache Memory results for 1GB, 1x 2GB and 3GB. They are all similar except that 1GB is the slowest and 3GB is the fastest (but not by much).

Also, for completeness, the memory bandwidth result for 1 x 1GB. Vista was very sluggish with 1GB!

John

 

answer: plug it in a 37" hdtv. It's what i use with my lil' laptop while i play CIV IV

 

CIV IV on a 37"?
could you post a pic of that?
i'd love to see how that looks
can't imagine what the benefit of that game on a big screen would be though

 

OK john thank you, could you give the text files from you bandwidth test? I have done them on mine w/1X1GB and 1X2GB very interesting. AMD has about 30% diff so want to see yours as the Intel bench and you have more configs also.

 

I didn't save text files for those tests - I will have to redo.

John

 

On the bench results it says a couple of interesting things on both John's and mine, gives bus, when it says 128 bit that is dual channel. When I remove 1 Dimm drops to 64 bit not dual. Also when I bench I utilize 46% of bandwidth w/dual 75% w/out. The “cache and memory” bench does show a diff but not dramatic like I hoped. The bandwidth never goes beyond the theoretical limit of a 64 bit bus so even though it does better dual it is not pushing the theoretical limits of even single. As I have read and these bench marks confirm dual does less for Intel's than AMD. As I stated b/4 my RAM bandwidth is the same as John's (edit: not correct I can not double Intels FSB to get bandwidth) but I run 533Mhz and he run 667Mhz, good to know. W/cache mem test Intel very far ahead w/block sizes 4MB and less much less so above than AMD and in fact above 16MB my 1.6Ghz beats John's. I have no idea what the block size means real world but can tell if pulling 64MB and higher blocks mine is faster.

 

Herewith the detailed results for four RAM combinations. Also a summary chart showing the bandwidths for the four combinations (results may be slightly different from the previous because each test gives slightly different numbers).

They all say 64 bit bus width whether one channel or two.

A big difference between AMD and Intel architecture is that AMD has the memory controller built into the CPU while with Intel it is in the northbridge.

John