Middle School Science Fair Investigation input and advice

I've been off the site for a few weeks as my school year has ended and the rest of the time I have been working on the Science Fair Project system.

My students have been building PCs, but to help them learn further, I want them to delve a little more deeply into the way computers work. The State STEM fair has a category for Computer Science, and I am encouraging my students to come up with serious projects. Projects (effect of paint color on WiFi reception) done by others seem a little soft and simplistic.

However, it is important that the project have positive results (unlike real scientific investigations). Last year one of my students did a very powerful and effective project which conclusively proved that something had no effect on water quality. They were praised for the detail of their project, and won an industry award, but were judged to have been unsuccessful because their results were negative!

To avoid that possibility, I have spent the last two weeks messing with our system to see what positive things can be discovered.

The system we are using is as follows:

PCPartPicker part list / Price breakdown by merchant

CPU: Intel Pentium G3258 3.2GHz Dual-Core Processor ($64.98 @ NCIX US)
Motherboard: ASRock Z97M OC Formula Micro ATX LGA1150 Motherboard ($99.98 @ Newegg)
Memory: G.Skill Trident X Series 8GB (2 x 4GB) DDR3-2400 Memory ($65.89 @ OutletPC)
Memory: G.Skill Trident X Series 16GB (2 x 8GB) DDR3-2400 Memory ($134.99 @ Newegg)
Storage: Crucial MX200 500GB 2.5" Solid State Drive ($198.99 @ SuperBiiz)
Case: Thermaltake Core V21 MicroATX Mini Tower Case ($54.99 @ Micro Center)
Power Supply: XFX XTR 550W 80+ Gold Certified Fully-Modular ATX Power Supply ($81.98 @ Newegg)
Operating System: Microsoft Windows 8.1 OEM (64-bit) ($86.98 @ OutletPC)
Wireless Network Adapter: TRENDnet TEW-804UB 802.11a/b/g/n/ac USB 2.0 Wi-Fi Adapter ($19.99 @ Amazon)
Total: $808.77
Prices include shipping, taxes, and discounts when available
Generated by PCPartPicker 2015-06-24 09:16 EDT-0400

Note:

We are using the G3258 because it is cheap (in case it gets fried) easy to work with, and is Haswell. We are not suggesting that this CPU is the basis for a serious gaming system. Also, in many benchmarks, the system will be CPU limited (as opposed to the more usual GPU limited) and we can be sure about the effect of the performance changes.

We have other coolers, but to start with we want to see what is possible with stock cooling.

We are using the motherboard because it is good and cheap, with decent power phases and can handle later a 4690K and a 4790K for comparison.

The memory was contributed by a parent and is more than good enough for our purposes.

We don't need much storage and this was fast and cheap (on sale) at the time.

The case handles mITX an mATX board and is very easy to work with and will fit liquid cooling later.

The PSU is good and modular, and much cheaper on sale.

Systems at school need to use WiFi, unless 'controlled' by IT.

Here's our build log http://www.tomshardware.com/forum/id-2626627/build-log-middle-school-science-fair-project-system.html

To test effect of GPUs we have an ASUS GTX970 DirectCU .

To monitor the system we are using

CPU-Z
Memtest 86
HWMonitor
SiSoft Sandra

For benchmarking

Unigine Heaven
Unigine Valley
PCMark 7
PCMark 8
3DMark 11 (which has stopped running and I cannot re-start it)
3DMark
Prime95 26.6
Euler 3D

I'm sure that there are other things we could do. They need to be free. This has cost me enough as it is.

As we go, I will post the full results I have gotten. Here's what I have done so far.

Basic system with
1 x 4 Gb memory 2361 - 2.466 - 276
2 x 4 Gb memory 2491 - 2.767 - 326
1 x 8 Gb memory 2289 - 2.319 - 259
2 x 8 Gb memory 2466 - 2.630 - 319
2 x 8 Gb + 2 x 4Gb memory 2460 - 2.631 - 318

The first number is the average of five trials of PCMark 8 Conventional
The second number is the 'frequency in Hz' of Euler 3d for 20 cycles and 2 threads.
The third number is Unigine Heaven Low setting (higher levels are iGPU limited)

I make no claims as to the relationship between these results and 'real life', but the benchmarks seem to be a good simulation of the light gaming and productivity that a G3258 might be used for.

I need to go back and check that the system did not automatically mess with the RAM timings.

I have also done a basic overclock and collected data at:

3.2 Ghz - 2.767 (stock) and 4460 comparison
3.3 Ghz - 2.730 i5 4590 comparison
3.5 Ghz - 2.805 i5 4690K and 4790 comparison
4.0 Ghz - 3.087 i7 4790K comparison
4.1 Ghz - 3.044
4.2 Ghz - 3.145 (best speed at 1.3V and <80C)
4.3 Ghz - 3.163 last stable speed, at1.34V and 92C on Prime95.

The number is the same Euler 3d with 2 x 4 GB memory at stock settings.

Questions or suggestion gratefully accepted. More to follow.
 
Here's some spreadsheets showing the various benchmarks. Missing entries did not get done, and I cannot go back easily.

Various systems that we have used or built as school:

15i2gxk.png


Memory configurations:

2hxuark.png
 
Here's where I am now with my overclocking.

2bva0g.png


The best I can get stable are:
4.2 Ghz at 1.305V and 80C.
2400Mhz memory 10 12 12 31
1300Mhz iGPU overclock at 1.250v

This gives a benchmark improvement from between 16% to 40%, depending on whether mostly CPU limited or iGPU limited.
 
They are going to investigate how the amount of memory and memory speed affects benchmark performance of various CPU/GPU configurations, looking at 1 x 4, 2 x 4, 1 x 8, and 2 x 8 memory and 1333, 1600 and 2400 memory speeds, and then relating that to what the benchmarks test and real life.

For example, if you do basic home productivity work, where is the sweet spot? What's the best configuration for number crunching? This can then be compared with the cost of the components.
 

Frozen Fractal

Reputable
May 17, 2015
96
0
4,630


Alright, I really like that. This is actually a great lesson for students. Keep it up! :)