Playing with AT case bench data

[Bootleg Betty]

So, when AT onlined their case bench, someone said "hey, there should be graph of noise versus temperature, so we can find cases that are good compromises and stuff". Since AT didn't do that, and since deploying feature like that might take a long time, I decided to make them myself.

First, we will play with the data a bit.

However, since you can't really download CSV or something from AT benchmarks, the data are limited, because I can't really writer scrapper while at work, and I'm not going to copy-paste, then repair, then merge in excel, for more then few columns.

Here are the data, I truncated name of Cooler Master Storm Scout to make the table smaller. This is the "(Stock)" version.

                          Case CPULoadTemp GPULoadTemp IdleNoise LoadNoise
1         Antec Eleven Hundred        37.7          49      32.5      34.9
2                  Antec GX700        39.6          49      32.5      34.7
3            AZZA Genesis 9000        43.7          51      30.0      33.8
4               AZZA Silentium        48.0          56      30.0      31.2
5               BitFenix Ghost        47.0          59      30.0      33.2
6              BitFenix Raider        41.9          57      30.0      36.9
7               BitFenix Ronin        45.8          63      30.0      39.7
8          BitFenix Shinobi XL        46.1          61      34.0      36.6
9    Cooler Master Storm Scout        42.7          58      30.6      37.5
10        Corsair Carbide 200R        44.6          60      30.0      33.9
11        Corsair Carbide 300R        43.1          54      32.2      34.8
12     Corsair Carbide Air 540        35.8          52      36.7      39.0
13       Corsair Obsidian 550D        50.7          58      30.0      32.7
14       Corsair Obsidian 900D        37.8          52      33.6      35.3
15       Corsair Vengeance C70        49.1          58      31.4      36.1
16           Cougar Challenger        42.1          54      33.9      37.1
17            Cubitek HPTX Ice        46.4          54      37.8      39.6
18    Fractal Design Core 1000        50.3          60      30.0      40.4
19    Fractal Design Define R4        48.0          60      30.0      33.2
20 Fractal Design Define XL R2        42.1          52      30.4      32.8
21                In-Win GRone        44.6          53      30.0      33.6
22                  In-Win GT1        38.9          55      37.9      39.9
23              Lian Li PC-A55        55.8          65      30.0      39.2
24             Lian Li PC-A76X        42.9          53      30.0      30.0
25      Nanoxia Deep Silence 1        45.0          54      30.0      30.6
26      Nanoxia Deep Silence 2        44.4          56      30.0      33.0
27                   NZXT H630        42.3          57      30.4      35.3
28     NZXT Phantom 530 (High)        36.4          53      30.4      35.5
29      NZXT Phantom 530 (Low)        39.3          56      30.0      33.9
30   NZXT Phantom 530 (Medium)        38.0          54      30.0      34.8
31            NZXT Phantom 630        36.9          51      30.0      31.0
32            NZXT Phantom 820        40.6          52      30.0      30.7
33          Rosewill Armor EVO        35.2          49      34.3      35.8
34    Rosewill Blackhawk Ultra        33.5          52      37.1      38.5
35             Rosewill Line-M        39.1          53      38.2      39.8
36      Rosewill Throne (High)        34.2          51      43.6      43.6
37       Rosewill Throne (Low)        36.0          52      30.7      35.2
38      SilverStone Raven RV04        31.2          54      33.7      36.3
39       SilverStone Sugo SG09        41.7          58      33.4      39.5

First of all, what are the relationships in the data?

                  bench.CPULoadTemp bench.GPULoadTemp bench.LoadNoise
bench.CPULoadTemp         1.0000000          0.730098      -0.1284284
bench.GPULoadTemp         0.7300980          1.000000       0.1876910
bench.LoadNoise          -0.1284284          0.187691       1.0000000

So, CPU and GPU load temperatures are correlated. That makes sense. They are not complete correlated, but the relationship is strong enough.

Load noise is inversely correlated with CPU load temp. That makes sense, hotter cases are quieter and vice versa.

Now, GPU temp is very weakly correlated with noise. They are also correlated with CPU temps, that are inversely correlated with noise. That's strange, but looking at the data, it doesn't seem wrong. It looks the GPU temps are doing their own thing.

                bench.LoadNoise bench.IdleNoise
bench.LoadNoise       1.0000000       0.6923913
bench.IdleNoise       0.6923913       1.0000000

Idle noise is correlated with load noise. Not surprising.

Plots ... in the next segment. Stay tuned.

 

[Bootleg Betty]

Damn. Anandtech has no attachment feature, and I can't get to imgur or some such site to share the plots. It'll have to wait until I get home.