Is 1gb dual channel better than 1.5gb+ single channel 8

[werD420]

I have 2x512mb kingston hyperx plugged up dual channel in my system. i have two open slots and another stick of 512(ultra) and a 1GB ocz stick thats gone for rma at the moment but if i plug in the 512 the bios post as singlechannel. I generally use this as my dev machine so i have a lot of documents open at once and i could really use the expanded memory. should i fork out the money to get two more sticks of memory or can i make do with what i have. ie is the bandwith loss gonna make that big of an impact on what im doin in comparison to the benefit of having almost double the memory.

Athlon 64 2.4 Ghz 939
Kingston HyperX 2x512MB


MCP, .Net Solutions Development <%_%>

 

[kmcferrin]

So here's my only question: If the OP is using enough applications or large enough data sets that the system requires more than 1.5 GB of RAM, won't having 1.5 GB in single channel mode still be faster than having 1 GB in dual channel mode and having to swap to disk for anything above 1 GB?

Then I think back to when the socket 939 Athlon 64 CPUs were introduced, and I didn't think that with their dual channel capability that an equivalently specced S939 version was significantly faster than a S754 version (single channel). Sure, the theoretical bandwidth is there, but you really only come close to that in high load situations or benchmarks. Right?

 

[cdogg]

That's the magic question buddy! Hence, why I am hesitant to take sides.

Like you said it really depends on how the PC is used, particularly the software itself and the "unique" mixture you create by running several programs at once.

It would be great to get a benchmark result from werD420 if he has time to test it under both environments.

~cdogg
"Insanity: doing the same thing over and over again and expecting different results." - Albert Einstein
[tab][navy]For general rules and guidelines to get better answers, click here:[/navy] faq219-2884

 

[werD420]

It would be great to get a benchmark result from werD420 if he has time to test it under both environments.

Ask and you shall receive I adjusted my timings back to auto and added one stick of 512 Ultra Memory to my 3rd bank. so i have
3 Sticks of 512 MB. The bios post as single channel without error but it appears that you all may have been on to something as far as the 1x512 + 2x256 ie( it appears that the bios emulates additional banks??)


SiSoftware Sandra

Benchmark Results
RAM Bandwidth Int Buff'd iSSE2 : 2105 MB/s
RAM Bandwidth Float Buff'd iSSE2 : 2108 MB/s
Results Interpretation : Higher index values are better.

Int Buff'd iSSE2 (Integer STREAM) Results Breakdown
Assignment : 2078MB/s
Scaling : 2078MB/s
Addition : 2134MB/s
Triad : 2131MB/s
Data Item Size : 16 byte(s)
Buffering Used : Yes
Offset Displacement Used : Yes
Bandwidth Efficiency : 80% (estimated)

Float Buff'd iSSE2 (Float STREAM) Results Breakdown
Assignment : 2089MB/s
Scaling : 2087MB/s
Addition : 2128MB/s
Triad : 2129MB/s
Data Item Size : 16 byte(s)
Buffering Used : Yes
Offset Displacement Used : Yes
Bandwidth Efficiency : 80% (estimated)

Features
SSE Technology : Yes
SSE2 Technology : Yes
SSE3 Technology : Yes
Supplemental SSE3 Technology : No
HTT - Hyper-Threading Technology : No

Chipset 1
Model : Acer Labs Incorporated (ALi/ULi) ULi M1689 K8 Northbridge with AGP and hypertransport
Front Side Bus Speed : 2x 824MHz (1648MHz data rate)
In/Out Width : 16-bit / 8-bit
Maximum Bus Bandwidth : 4944MB/s (estimated)

Logical/Chipset 1 Memory Banks
Memory Controller in Processor : Yes

Chipset 2
Model : Advanced Micro Devices (AMD) Athlon 64 / Opteron HyperTransport Technology Configuration
Front Side Bus Speed : 2x 824MHz (1648MHz data rate)
In/Out Width : 8-bit / 16-bit
Maximum Bus Bandwidth : 4944MB/s (estimated)

Logical/Chipset 2 Memory Banks
Bank 0 : 256MB DDR-SDRAM 2.5-3-3-7 (tCL-tRCD-tRP-tRAS) CR2
Bank 1 : 256MB DDR-SDRAM 2.5-3-3-7 (tCL-tRCD-tRP-tRAS) CR2
Bank 2 : 512MB DDR-SDRAM 2.5-3-3-7 (tCL-tRCD-tRP-tRAS) CR2
Bank 4 : 256MB DDR-SDRAM 2.5-3-3-7 (tCL-tRCD-tRP-tRAS) CR2
Bank 5 : 256MB DDR-SDRAM 2.5-3-3-7 (tCL-tRCD-tRP-tRAS) CR2
Channels : 1
Speed : 2x 164MHz (328MHz data rate)

//^-- That appears to take a good hit as well as my bandwith which isnt even getting 3200 --v//

Width : 64-bit
Memory Controller in Processor : Yes
Cores per Memory Controller : 1 Unit(s)
Maximum Memory Bus Bandwidth : 2624MB/s (estimated)




The test to me would be the build times in visual studio. My current projects, unfortunately, arent all THAT huge so when it came down to it.. build time was around 3-5 seconds with both setups. Would a test like Super-Pi have a significant change based on memory bandwith or is that more processor specific? Lots of good thoughts floating around here and I appreciae the opportunity to hear all of them. Stars All Around!

Cheers,
DrewG


MCP, .Net Solutions Development <%_%>

 

[kmcferrin]

Super-Pi probably wouldn't be a good test since the code is probably small enough to fit within the cache and execute without really touching main memory.

 

[cdogg]

Try running SYSmark 2004 (patch 2) from this link:

http://www.bapco.com/updates.html

The "SE" edition isn't free, but I think the one at that link is.

~cdogg
"Insanity: doing the same thing over and over again and expecting different results." - Albert Einstein
[tab][navy]For general rules and guidelines to get better answers, click here:[/navy] faq219-2884