The Inquirer-Home

Testing Vista's different memory configurations

First INQpression Vista 32-bit with 512, 1024, 2048 and 4096MB memory
Mon Dec 04 2006, 13:13
WE WANTED to find out about what happens when you upgrade your memory amount with Vista.

Everyone knows that Vista needs more memory and we tried a few scenarios to discover the difference between certain memory setups.

This is just the first part of the memory testing that we plan to conduct in the few coming days.

We used a high-end machine only, powered with Core 2 Duo 6700 CPU and EVGA's latest greatest 680i SLI Premium motherboard.

We ran tests to determine what happens when you upgrade from 512MB to 1GB from 1GB to 2 GB and so on.

We used 32-bit Vista and ran into the memory allocation problem. Once you move to 4GB, this board only sees 2814MB of memory. The graphic card and some other devices ate more than a gigabyte of system memory. We are currently testing the 2GB and 4GB setup under Vista 64 but the Vole is not being cooperative on this subject.

alt='8500_corsair_2x512gb-f'

Corsair TWIN2X1024-8500 two times 512 MB kit
We used one or two 512MB Corsair TWIN2X1024-8500 800MHz at 800MHz as the board didn't want to set a single module at 1066. Whatever we tried with the single module you end up at 800MHz. The same thing happened with a single 1024MB Corsair Dominator TWIN2x2048-8500CAS5 and it stays at 800MHz only. We think it is a BIOS issue but Corsair didn't bother that much about it as no one will use a single module. EVGA is alerted on the issue and is working hard to fix it.

alt='8500_corsair_2x1gb-f'

When you pair the memory, no matter whether are we talking about 2x512 MB, 2x1024 or 4x1024 it worked at the desired 1066 mode. We used the 5-5-5-15 memory settings and didn't have any stability issues at this speed.

The 2x2 GB memory chips should be available in the months to come, probably shortly before the Vista announcement.

For the first batch of testing we decided to run all 3Dmarks with the latest patch which delayed our review, PCmark05, Sandra 2007 and two games FEAR and Battlefield 2142 well know for its spyware abilities. The synthetic tests showed the difference but we had a hard time to show the difference in games. Checking the load times, we did find a difference.

alt='vista680i'

Benchmarketing

We used:
EVGA nForce 680i SLI board
Intel Core 2 Duo E6700, 2x 2.67GHz, 266MHz FSB, 4MB Cache
Sapphire Radeon X1950XTX - Catalyst Vista RTM
Seagate Barracuda 7200.9 400GB SATA NCQ hard drive
Akasa EVO AK 922 Blue Athlon 64/X2/FX cooler and Intel CPU's
OCZ 700W GameXstream power supply

alt='512mb-fs'
A single 512 MB under Vista

the big picture

alt='2x512=1024mb-fs'
Two times 512 MB under Vista

the big picture

alt='1024mb-fs'
A single Dominator 8500CAS5 under Vista

the big picture

alt='2x1042=2048mb-fs'
Two times Dominator TWIN2x2048-8500CAS5 under Vista

the big picture

Before I even start I have to say that we never made it to 4096 GB. We tested 2814 MB as there was not more physical memory available. In 3Dmark 4x1024 GB were even slower by 1000 marks from the 2x1024 GB. Then we really saw the difference. Running a single 1 GB module is not a good idea. You will lose 300 marks and two frames in game one and one frame in game three. This is due the slower timings and you can also see this as DDR 2 1066 versus DDR 2 800 benchmark. It is not much but you can see the difference.

Composite Figures 3Dmark 05
3DMark 05
Game1
Game2
Game2
EVGA nForce 680i SLI Premium - 4096MB Corsair Dominator TWIN2x2048-8500CAS5 1066MHz
12606
55.8
34.4
66.8
EVGA nForce 680i SLI Premium - 2X1024MB Corsair Dominator TWIN2x2048-8500CAS5 1066MHz
12701
55.5
35.3
66.9
EVGA nForce 680i SLI Premium - 1024MB Corsair Dominator TWIN2x2048-8500CAS5 800MHz
12407
53.5
34.9
65.4
EVGA nForce 680i SLI Premium - 2x512MB Corsair TWIN2X1024-8500 1066MHz
12608
55.0
34.9
66.7
EVGA nForce 680i SLI Premium - 512MB Corsair TWIN2X1024-8500 800MHz
12251
52.2
34.5
65.4
? ? ? ? ?
CPU Test
CPU Score ?
CPU Test 1 ?
CPU Test 2
EVGA nForce 680i SLI Premium - 4096MB Corsair Dominator TWIN2x2048-8500CAS5 1066MHz
9763
6.5
6.6
64.8
EVGA nForce 680i SLI Premium - 2X1024MB Corsair Dominator TWIN2x2048-8500CAS5 1066MHz
9522
6.2
6.5
30.8
EVGA nForce 680i SLI Premium - 1024MB Corsair Dominator TWIN2x2048-8500CAS5 800MHz
8652
5.8
5.8
30.8
EVGA nForce 680i SLI Premium - 2x512MB Corsair TWIN2X1024-8500 1066MHz
9213
6.1
6.2
30.8
EVGA nForce 680i SLI Premium - 512MB Corsair TWIN2X1024-8500 800MHz
8532
5.7
5.7
30.8
? ? ? ? ?
Composite Figures 3Dmark 05
Single Texturinng
Multi Textur.
Pixel Shader

VS/VS
complex

EVGA nForce 680i SLI Premium - 4096MB Corsair Dominator TWIN2x2048-8500CAS5 1066MHz
5094.5
8390.3
360.3
136.5/51.3
EVGA nForce 680i SLI Premium - 2X1024MB Corsair Dominator TWIN2x2048-8500CAS5 1066MHz
5171.4
8297.3
360.9
136.6/51.3
EVGA nForce 680i SLI Premium - 1024MB Corsair Dominator TWIN2x2048-8500CAS5 800MHz
5107.1
8264.5
360.1
134.1/51.3
EVGA nForce 680i SLI Premium - 2x512MB Corsair TWIN2X1024-8500 1066MHz
5161.8
8345.0
360.0
136.3/51.3
EVGA nForce 680i SLI Premium - 512MB Corsair TWIN2X1024-8500 800MHz
4108.5
6935.4
309.2
127.5/55.5
? ? ? ? ?
Composite Figures 3Dmark 06
3DMark 06
? ?
EVGA nForce 680i SLI Premium - 4096MB Corsair Dominator TWIN2x2048-8500CAS5 1066MHz
6417
? ? ?
EVGA nForce 680i SLI Premium - 2X1024MB Corsair Dominator TWIN2x2048-8500CAS5 1066MHz
6400
? ? ?
EVGA nForce 680i SLI Premium - 1024MB Corsair Dominator TWIN2x2048-8500CAS5 800MHz
6384
? ? ?
EVGA nForce 680i SLI Premium - 2x512MB Corsair TWIN2X1024-8500 1066MHz
6399
? ? ?
EVGA nForce 680i SLI Premium - 512MB Corsair TWIN2X1024-8500 800MHz
6201
? ? ?
? ? ? ? ?
SM2.0 Test
Score
GT1
GT2


EVGA nForce 680i SLI Premium - 4096MB Corsair Dominator TWIN2x2048-8500CAS5 1066MHz
2522
19.748
22.280
?
EVGA nForce 680i SLI Premium - 2X1024MB Corsair Dominator TWIN2x2048-8500CAS5 1066MHz
2517
19.699
22.258
?
EVGA nForce 680i SLI Premium - 1024MB Corsair Dominator TWIN2x2048-8500CAS5 800MHz
2518
19.695
22.279
?
EVGA nForce 680i SLI Premium - 2x512MB Corsair TWIN2X1024-8500 1066MHz
2523
19.742
22.305
EVGA nForce 680i SLI Premium - 512MB Corsair TWIN2X1024-8500 800MHz
2512
19.662
22.206
? ? ? ? ?
HDR/SM3.0 Test
Score
HDR1
HDR2


EVGA nForce 680i SLI Premium - 4096MB Corsair Dominator TWIN2x2048-8500CAS5 1066MHz
2754
26.605
28.477
?
EVGA nForce 680i SLI Premium - 2X1024MB Corsair Dominator TWIN2x2048-8500CAS5 1066MHz
2751
26.578
28.449
?
EVGA nForce 680i SLI Premium - 1024MB Corsair Dominator TWIN2x2048-8500CAS5 800MHz
2743
26.521
28.344
?
EVGA nForce 680i SLI Premium - 2x512MB Corsair TWIN2X1024-8500 1066MHz
2744
26.471
28.416
EVGA nForce 680i SLI Premium - 512MB Corsair TWIN2X1024-8500 800MHz
2620
24.165
28.229
? ? ? ? ?
CPU Test
Score
CPU1
CPU2


EVGA nForce 680i SLI Premium - 4096MB Corsair Dominator TWIN2x2048-8500CAS5 1066MHz
2227
0.690
1.149
?
EVGA nForce 680i SLI Premium - 2X1024MB Corsair Dominator TWIN2x2048-8500CAS5 1066MHz
2206
0.687
1.133
?
EVGA nForce 680i SLI Premium - 1024MB Corsair Dominator TWIN2x2048-8500CAS5 800MHz
2198
0.682
1.125
?
EVGA nForce 680i SLI Premium - 2x512MB Corsair TWIN2X1024-8500 1066MHz
2210
0.686
1.140
EVGA nForce 680i SLI Premium - 512MB Corsair TWIN2X1024-8500 800MHz
2086
0.657
1.060
? ? ? ? ?
Sandra Lite XI
CPU
Memory Bandw.
Memory Latency
Cache&Memory
EVGA nForce 680i SLI Premium - 4096MB Corsair Dominator TWIN2x2048-8500CAS5 1066MHz
24681/16917
5627/5625
72 ns / 62.3
22693 MBs / 71.0
EVGA nForce 680i SLI Premium - 2X1024MB Corsair Dominator TWIN2x2048-8500CAS5 1066MHz
24598/16919
5833/5722
75 ns / 64.4
22434 MBs / 71.6
EVGA nForce 680i SLI Premium - 1024MB Corsair Dominator TWIN2x2048-8500CAS5 800MHz
24535/16901
5187/5155
84 ns / 72.0
19421 MBs / 100.7
EVGA nForce 680i SLI Premium - 2x512MB Corsair TWIN2X1024-8500 1066MHz
24537/16908
5671/5648
75 ns / 66.0
19930 MBs / 94.1
EVGA nForce 680i SLI Premium - 512MB Corsair TWIN2X1024-8500 800MHz
24351/16896
5099/5048
85 ns /73.6
18321 MBs / 115.5
? ? ? ? ?

PC Mark 2005

CPU
Memory
HDD
Total
EVGA nForce 680i SLI Premium - 4096MB Corsair Dominator TWIN2x2048-8500CAS5 1066MHz
6765
5685
5428
6858
EVGA nForce 680i SLI Premium - 2X1024MB Corsair Dominator TWIN2x2048-8500CAS5 1066MHz
6764
5695
5453
6873
EVGA nForce 680i SLI Premium - 1024MB Corsair Dominator TWIN2x2048-8500CAS5 800MHz
6474
5424
5270
6751
EVGA nForce 680i SLI Premium - 2x512MB Corsair TWIN2X1024-8500 1066MHz
6736
5677
5190
6757
EVGA nForce 680i SLI Premium - 512MB Corsair TWIN2X1024-8500 800MHz
0
0
0
0
? ? ? ? ?
FEAR
1024x768
1280x960
1600x1200
2048x1536
EVGA nForce 680i SLI Premium - 4096MB Corsair Dominator TWIN2x2048-8500CAS5 1066MHz
148
115
85
57
EVGA nForce 680i SLI Premium - 2X1024MB Corsair Dominator TWIN2x2048-8500CAS5 1066MHz
146
114
84
56
EVGA nForce 680i SLI Premium - 1024MB Corsair Dominator TWIN2x2048-8500CAS5 800MHz
143
113
84
54
EVGA nForce 680i SLI Premium - 2x512MB Corsair TWIN2X1024-8500 1066MHz
145
113
84
56
EVGA nForce 680i SLI Premium - 512MB Corsair TWIN2X1024-8500 800MHz
102
84
57
36
? ? ? ? ?
FEAR High Quality FSAA 4X + Aniso 16X
1024x768
1280x960
1600x1200
2048x1536
EVGA nForce 680i SLI Premium - 4096MB Corsair Dominator TWIN2x2048-8500CAS5 1066MHz
110
80
57
37
EVGA nForce 680i SLI Premium - 2X1024MB Corsair Dominator TWIN2x2048-8500CAS5 1066MHz
109
81
57
37
EVGA nForce 680i SLI Premium - 1024MB Corsair Dominator TWIN2x2048-8500CAS5 800MHz
107
79
56
36
EVGA nForce 680i SLI Premium - 2x512MB Corsair TWIN2X1024-8500 1066MHz
108
80
56
36
EVGA nForce 680i SLI Premium - 512MB Corsair TWIN2X1024-8500 800MHz
86
64
42
25
? ? ? ? ?
Battlefield 2142 High Quality FSAA 4X + Aniso 8X
2048x1536
Loading first time
Loading game second time
Loading time other map
EVGA nForce 680i SLI Premium - 4096MB Corsair Dominator TWIN2x2048-8500CAS5 1066MHz
54.45
26-30sec
16-18sec
5-7sec
EVGA nForce 680i SLI Premium - 2X1024MB Corsair Dominator TWIN2x2048-8500CAS5 1066MHz
54.22
27-30sec
17-19sec
5-7sec
EVGA nForce 680i SLI Premium - 1024MB Corsair Dominator TWIN2x2048-8500CAS5 800MHz
52.12
35-40sec
29-32sec
7-15sec
EVGA nForce 680i SLI Premium - 2x512MB Corsair TWIN2X1024-8500 1066MHz
54.13
34-37sec
27-30sec
6-9sec
EVGA nForce 680i SLI Premium - 512MB Corsair TWIN2X1024-8500 800MHz
Not playable at first load, sec load 40FPS avg minimum 0 FPS with stuttering
270-280sec
235-240sec
70 --95sec
? ? ? ? ?

alt='4096mb-fs'

The big picture

You can clearly see that running the 2x512MB Corsair TWIN2X1024-8500 at 1066MHz gets the performance back and you get the same performance as with 4x1024 or slightly slower than 2x1024.

Then we brought it to its knees. Vista and gaming with 512MB are not a good idea. You end up 500 marks slower than the best score and almost 400 slower from the 2x512MB. The CPU score really suffers when you use a single module. You end up 1200 marks or about fifteen percent slower than the top score. 4x1024 will get you to the top score and it is faster than 2x1024MB memory. You lose performance in single and muti-texturing with 512MB so you need to have at least 2x512MB to feel good with this test.

3Dmark06 really likes 4096MB memory and runs a few points faster than 2048MB memory. 2x512MB are still good for this test but once you go to single 1024 or 512MB you lose performance. The CPU score is again the fastest with 4x1024MB with 512MB coming in last place, as expected.

Then we moved to Sandra XI lite for Vista. The CPU scores were virtually identical so we won't bother with them. In memory testing 2X1024MB Corsair Dominator wins the crown and can pull out 200 MB more than 4x1 GB or 2x512 MB. You always lose some speed when you plug 4x1GB versus 2x1 GB that is a logical thing to expect but it was not that much. Running a single module is not a good idea as it will penalise the performance and will score 500 to 700 MB/s less than with two modules.

Interestingly enough, memory latency test is the fastest on 4x1GB as the smaller is better in this test. It is not much but it beats 2x1 GB. You can see the difference when you run a single module stacked at 800MHz and you drop from 72 to 85 ns or in second test from 72 to 66 ns.

Once again 4x1024MB is the winner combination. It runs 250MB in cache test than the 2x1GB. 2x512MB is faster than a single 1GB but it also means that 1GB of total memory means up to 3200MB/s slower. A single 512MB is 4300MB/s slower than the winning 4x1024 MB. 4x1024 again has the best memory score and beats 4x1024 by 0.6 points. In this test smaller is better. a single 512MB gets you the worst score and is almost sixty percent slower than the 4x1 GB. 1GB and 2x512MB are again significantly slower in this test.

PC mark05 had almost identical CPU and HDD scores so we wont mention them. the HDD does runs a bit slower at 2x512 MB and 1GB but that's about it. 2X1024MB Corsair Dominator TWIN2x2048-8500CAS5 1066MHz is slightly faster than 4x1024 MB of this memory. 2x512MB Corsair TWIN2X1024-8500 1066MHz is just slightly slower while a single 1024MB Corsair Dominator TWIN2x2048-8500CAS5 800MHz scores 260MB slower.

Then we moved on FEAR and Battlefield 2142 as we certainly knew it will work. FEAR runs one to two frames faster with 4x1024 MB it is an advantage but not a big one just a bit over one percent. 2x512MB can be up to two frames faster than 1 GB by almost four percent. A single 1GB at DDR 2 800 speed runs three to five frames slower or less than four per cent.

But you really see the difference with 512 MB. You instantly lose fifty to fifty-eight per cent of performance as the system runs up to 50 FPS slower. And the load times increases a lot. FEAR loads faster with 2GB memory than with one or 512MB and then we got an idea about Battlefield.

FEAR FSAA and Aniso scores showed the same difference and 512MB means a significant drop in frames. It makes the game hard to play and the load times are enormous. We measured the load times in battlefield 2142.

At first we wanted to test the game at 2148x1536 with Antialiasing and learned that 4x1024MB is slightly faster than 2GB, than 2GB is faster than 1GB and 2x512 are faster than 1GB. 512MB is not enough for Vista and this game. At first load it is not playable at all. You have to load the game again to able to play it and even then it is stuttering. FRAPS averages the game to 40 FPS but sometimes it drops to a few frames only.

Then we saw a light, as we wanted to test the load times. And that was a full Monty. 4x1GB means you will end up slightly faster to load than with 2x1GB. It takes 26 second versus 27 with 2GB. Once you move to 1GB the load time extends to 35 to 40 seconds while 2x512MB gets you to 34-37 seconds. The 512MB loading is a nightmare. It takes 270-280sec to load this game. Two numbers are the average from the latest load to the slowest one as we tried this dozen times.

We figure out that second time the map loads much faster. this even occurs when you go out of the Battlefield and start it again. It is 16-18 sec with 4GB, 17-19sec with 2GB, 29-32 sec with 1GB, slightly faster at 27-30 sec with 2x512MB. Again, it takes 235-240sec or more than thirteen times more to load the map with the 512MB. We again spotted the difference when loading a second map. It takes five to seven seconds to load it with 2x1 GB or 4x1 GB and it takes up to twice that much 7-15 sec to load it with 1GB. 2x512MB will load the second map for 6 to 9 seconds and this is how long it takes from the Italian coast to a down town Belgrade. Loading times with 512MB are 70-95 sec again thirteen and a half times slower. It is simply nightmare if you have 512MB and run it under Vista.

In Short
You simply cannot play games on Vista with 512MB o memory, at least not the games we tested. You will sometimes get playable scores but even then the experience and the load time will be ridiculous.

Anything more than 512MB will help you a lot, with 2x1GB being the best price performance buy.

4x1GB is the best choice for best raw performance. It is hard to prove it in every test, but if you have 2x1GB and you load as much in memory, the system will become endessly slow.

Slipping in 4x1024MB modules will overcome this problem and will help you out running more than 2048MB memory but it does come at a saucy price. Battlefield 2142 and Vista takes up to 1500MB of memory, if you have two or more GB and it would not be hard to waste another 500MB and put the two gigs in trouble.

In syntactical scores you can see that 2GB or higher is the way to go and, except with latency, 4x1GB is fast. We can easily say that 2GB is the one that we can recommend to just about anyone who wants to play games on Vista. One gig will work but not that smoothly all the time. ?ber geeks should go for 4x1GB or alternately wait for the birth of 2x2048GB modules. ?

Reviewed and tested by Sanjin Rados and Fuad Abazovic

 

Share this:

blog comments powered by Disqus
Advertisement
Subscribe to INQ newsletters

Sign up for INQbot – a weekly roundup of the best from the INQ

Advertisement
INQ Poll

Heartbleed bug discovered in OpenSSL

Have you reacted to Heartbleed?