Phenom II X6 OCing

[Mr Bill]

I don't fold. I put my computer to sleep when I'm not using it. Less power used (ergo less money), less fans wear, less hdd wear, less dust in the case. I've got nothing against folding, but it's not for me.

++
When I quite doing SETI my electric bill dropped from 1200KWh/month to 800KWh/month.

[mmmmmdonuts21]

I don't fold. I put my computer to sleep when I'm not using it. Less power used (ergo less money), less fans wear, less hdd wear, less dust in the case. I've got nothing against folding, but it's not for me.

++
When I quite doing SETI my electric bill dropped from 1200KWh/month to 800KWh/month.

I typically only fold or do BOINC for half 4-5 months during the winter. I live in a basement and its stays around 45-50 degrees in the winter without a heater on. I have two choices:

1) Use a portable heater I have
2) DC projects on the computer

So I choose the DC projects to keep me semi warm. Uses about the same amount of electricity because the electric heater pulls way more power.

[Mr Bill]

Here is a link to Anandtech that benches an X6 with different NB frequencies at the same clock. http://www.anandtech.com/show/3877/asrock-890fx-deluxe-full-review-and-an-investigation-of-thuban-performance-scaling/7

An interesting article. I would have liked it more if he had benched the NB and DDR3 frequency while holding one or the other constant. Say, NB at different frequencies while DDR3 at 1600MHz.

[Mr Bill]

Here are my SiSoftware Lite 2011c-1730 results (Windows XP64)

<<< Memory Bandwidth >>>
--------------------------------------------------------------------------------
< Benchmark Results >
Aggregate Memory Performance: 14.75GB/s
Integer B/F iSSE2 Memory Bandw:14.76GB/s
Float B/F iSSE2 Memory Bandwid:14.74GB/s

<<< Processor Arithmetic >>>
--------------------------------------------------------------------------------
< Benchmark Results >
Aggregate Arithmetic Performan:40.55GOPS
Dhrystone ALU: 44.79GIPS
Whetstone iSSE3: 36.71GFLOPS

<<< Processor Multi-Media >>>
--------------------------------------------------------------------------------
< Benchmark Results >
Aggregate Multi-Media Performa:70.31MPix/s
Multi-Media Integer x16 iSSE2: 54.48MPix/s
Multi-Media Float x8 iSSE2: 90.75MPix/s
Multi-Media Double x4 iSSE2: 49.53MPix/s

<<< Memory Latency >>>
--------------------------------------------------------------------------------
< Benchmark Results >
Memory Latency: 73.0ns
< Detailed Benchmark Results >
1kB Range: 0.9ns / 3.0clocks
4kB Range: 0.9ns / 3.0clocks
16kB Range: 0.9ns / 3.0clocks
64kB Range: 0.9ns / 3.0clocks
256kB Range: 4.9ns / 15.8clocks
1MB Range: 15.5ns / 49.7clocks
4MB Range: 21.5ns / 68.9clocks
16MB Range: 67.9ns / 217.1clocks
64MB Range: 73.0ns / 233.6clocks

<<< Cache and Memory >>>
--------------------------------------------------------------------------------
< Benchmark Results >
Cache/Memory Bandwidth: 63.3GB/s

< Cache Information >
Integrated Data Cache: 249GB/s
L2 On-board Cache: 115.27GB/s
L3 On-board Cache: 45.4GB/s

< Detailed Benchmark Results >
2kB Blocks: 263.86GB/s (111GB/s - 263.86GB/s)
4kB Blocks: 264.1GB/s (106.66GB/s - 264.1GB/s)
8kB Blocks: 265GB/s (124GB/s - 265GB/s)
16kB Blocks: 268.9GB/s (130.41GB/s - 268.9GB/s)
32kB Blocks: 242.13GB/s (136.3GB/s - 242.13GB/s)
64kB Blocks: 223.88GB/s (137.84GB/s - 223.88GB/s)
128kB Blocks: 214.77GB/s (128.89GB/s - 214.77GB/s)
256kB Blocks: 196.14GB/s (106.17GB/s - 196.14GB/s)
512kB Blocks: 130GB/s (81.4GB/s - 130GB/s)
1MB Blocks: 100.6GB/s (58.3GB/s - 100.6GB/s)
4MB Blocks: 45.4GB/s (34GB/s - 45.4GB/s)
16MB Blocks: 14.3GB/s (12.3GB/s - 14.3GB/s)
64MB Blocks: 13.8GB/s (11.86GB/s - 13.8GB/s)
256MB Blocks: 13.68GB/s (11.67GB/s - 13.68GB/s)
1GB Blocks: 13GB/s (11.15GB/s - 13GB/s)
4GB Blocks: 10.4GB/s (9.1GB/s - 10.4GB/s)

Looks to me like the X6 gives a significant boost over the X4 for the following...
X6------------X4---------Diff-----------% of X4---------Test
71.55------40.55--------31----------------76% -------Aggregate Arithmetic Performance :
124.77-----70.31--------54.46------------77%----------Aggregate Multi-Media Performance :
93.11------63.3---------29.81------------47%-----------Cache/Memory Bandwidth :
That last is to be expected two extra cores accessing memory would be 50% more bandwidth all other things being equal.

[mmmmmdonuts21]

When you ran your SiSoftware memory bandwidth benchmarks did the program utilize 100% of your 8GB of RAM? The reason I am asking this is because even though I have 16GB of RAM, the max I saw used in the test was 10.5GB. The reason I am asking is because I am curious if my system really can only use that amount of RAM bandwidth effectively at one time. (I know programs such as LinX can use all 16GB but I believe the RAM is used for storage in that case instead of bandwidth)

My thought process on this maybe completely run but its running through my head so I thought I would bring it up.

Looks to me like the X6 gives a significant boost over the X4 for the following...
X6------------X4---------Diff-----------% of X4---------Test
71.55------40.55--------31----------------76% -------Aggregate Arithmetic Performance :
124.77-----70.31--------54.46------------77%----------Aggregate Multi-Media Performance :
93.11------63.3---------29.81------------47%-----------Cache/Memory Bandwidth :
That last is to be expected two extra cores accessing memory would be 50% more bandwidth all other things being equal.

Your timings are a little tighter than mine so I think that is pretty great that it scales nearly linearly in the memory bandwidth category. Since my clock speed is at 3.75 and yours at 3.2 I would look a the results by doing it by the score/(# of cores / GHz). That would give us something like this:

Score per core per GHz

X6------------X4-----Test
3.18------3.17------Aggregate Arithmetic Performance :
5.55-----5.49-------Aggregate Multi-Media Performance :
4.14------4.96------Cache/Memory Bandwidth :

To get an idea what the two extra cores are doing I am going to just divide the original score by the GHz so its on an even playing field.

Score per Core

X6------------X4---------Diff-----------% of X4---------Test
19.08------12.67--------6.41-----------50.6% -------Aggregate Arithmetic Performance :
33.27-----21.97--------11.3------------51.4%----------Aggregate Multi-Media Performance :
24.83------19.78---------5.05------------25.5%-----------Cache/Memory Bandwidth :


To me these results show how well the X6's scale compared to the X4. Nearly a 50% increase with Arithmetic and Multi-Media Performance when all the cores are being utilized (at the same speed). This makes sense, but its interesting to see the results actually scale that way. The memory in the last one is very interesting to me because your timings are one notch different basically yet you see what would be a 25% loss from pure scaling. This is just by going 8-8-8-24 rather than 9-9-9-24.

This makes me want to tighten my timings.

[Mr Bill]

The scaling is quite spectacularly good. Looks to me like Sandra used half our respective memories. I don't see how a single clock tick can net you that 25% difference in the memory bandwidth. There must be some other difference. I think it most likely a difference between X64 versus Win7. At the very bottom is what the so-called OCZ Platinum modules were actually specified for in their spid.

<<< Memory Bandwidth >>>
--------------------------------------------------------------------------------

< Benchmark Results >
Aggregate Memory Performance: 14.75GB/s
Integer B/F iSSE2 Memory Bandw:14.76GB/s
Float B/F iSSE2 Memory Bandwid:14.74GB/s
Results Interpretation: Higher scores are better.

< Performance vs. Speed >
Aggregate Memory Performance: 9.44MB/s/MHz
Integer B/F iSSE2 Memory Bandw:9.45MB/s/MHz
Float B/F iSSE2 Memory Bandwid:9.43MB/s/MHz
Results Interpretation: Higher scores are better.

< Performance vs. Power >
Chipset(s)/Memory Power: 48.75W
Aggregate Memory Performance: 309.83MB/s/W
Integer B/F iSSE2 Memory Bandw:310.07MB/s/W
Float B/F iSSE2 Memory Bandwid:309.60MB/s/W
Results Interpretation: Higher scores are better.

< Capacity vs. Power >
Memory Capacity: 168MB/W
Results Interpretation: Higher scores are better.

< Integer B/F iSSE2 Memory Bandwidth >
Assignment: 14.52GB/s
Scaling: 14.49GB/s
Addition: 15GB/s
Triad: 15GB/s
Data Item Size: 16bytes
Buffering Used: Yes
Offset Displacement: Yes
Bandwidth Efficiency: 58.43%

< Float B/F iSSE2 Memory Bandwidth >
Assignment: 14.51GB/s
Scaling: 14.4GB/s
Addition: 15GB/s
Triad: 15GB/s
Data Item Size: 16bytes
Buffering Used: Yes
Offset Displacement: Yes
Bandwidth Efficiency: 58.10%

< Performance Test Status >
Result ID: ATI (AMD) RD790 GFX Dual Slot; 2x 4GB OCZ
OCZ3P20002G DIMM DDR3 (1.6GHz, 50% OC) PC3-
8500 (8-8-8-24 5-32-9-6)
Platform Compliance: x64
Total Memory: 8GB
Memory Used by Test: 4GB
NUMA Support: No
SMP (Multi-Processor) Benchmar:No
Total Test Threads: 4
Multi-Core Test: Yes
SMT (Multi-Threaded) Benchmark:No
Processor Affinity: P0C0T0 P0C1T0 P0C2T0 P0C3T0
System Timer: 3.58MHz
Page Size: 4kB
Use Large Memory Pages: No

< Features >
SSE Technology: Yes
SSE2 Technology: Yes
SSE3 Technology: Yes
Supplemental SSE3 Technology: No
SSE4.1 Technology: No
SSE4.2 Technology: No
AVX - Advanced Vector eXtensio:No
FMA - Fused Multiply/Add eXten:No
SSE4A Technology: Yes
FMA4 - 4 Operands Fused Multip:No
HTT - Hyper-Threading Technolo:No

< Chipset >
Model: ATI (AMD) RD790 GFX Dual Slot
OEM Device Name: ATI (AMD) RD790 GFX Dual Slot
Revision: A1
Front Side Bus Speed: 2x 2GHz (4GHz)
Width: 16-bit / 16-bit
Maximum Bus Bandwidth: 15.63GB/s

< Chipset >
Model: AMD F10h Athlon 64/Opteron/Sempron HT Hub
Revision: A1
Front Side Bus Speed: 2x 2GHz (4GHz)
Width: 16-bit / 16-bit
Maximum Bus Bandwidth: 15.63GB/s

< Logical/Chipset Memory Banks >
Bank 0: 1GB DIMM DDR3 8-8-8-24 5-32-9-6 2T
Bank 1: 1GB DIMM DDR3 8-8-8-24 5-32-9-6 2T
Bank 2: 1GB DIMM DDR3 8-8-8-24 5-32-9-6 2T
Bank 3: 1GB DIMM DDR3 8-8-8-24 5-32-9-6 2T
Bank 8: 1GB DIMM DDR3 8-8-8-24 5-32-9-6 2T
Bank 9: 1GB DIMM DDR3 8-8-8-24 5-32-9-6 2T
Bank 10: 1GB DIMM DDR3 8-8-8-24 5-32-9-6 2T
Bank 11: 1GB DIMM DDR3 8-8-8-24 5-32-9-6 2T
Channels: 2
Bank Interleave: 4-way
Memory Bus Speed: 2x 800MHz (1.6GHz)
Multiplier: 4x
Width: 64-bit
Integrated in Processor: Yes
Cores per Memory Controller: 4 Unit(s)
Maximum Memory Bus Bandwidth: 25GB/s

< Memory Module(s) >
Memory Module: OCZ OCZ3P20002G 2GB DIMM DDR3 PC3-8500U DDR3-
1066 (7-7-7-16 4-26-8-4)
Memory Module: OCZ OCZ3P20002G 2GB DIMM DDR3 PC3-8500U DDR3-
1066 (7-7-7-16 4-26-8-4)
Memory Module: OCZ OCZ3P20002G 2GB DIMM DDR3 PC3-8500U DDR3-
1066 (7-7-7-16 4-26-8-4)
Memory Module: OCZ OCZ3P20002G 2GB DIMM DDR3 PC3-8500U DDR3-
1066 (7-7-7-16 4-26-8-4)