(UPDATED) Techradar using biased graphing techniques in AMD Ryzen Threadripper 1950X review

[Nytohan]

http://www.techradar.com/reviews/amd-ryzen-threadripper-1950x

 

The graphs for FPS measurements in TechRadar's review are garbage.

 

A few of the images used in their review are fine, (Tomb raider@1440, Total war@1080)

But then there's Total War at 2160, where the two CPUs were 1 fps apart:

 

 

 

This is extremely misleading, seems to imply that the i9-7900x performs ~3x better than the threadripper, when in fact this entire graph exists within what LMG would consider a margin of error.

More graphs below.

 

UPDATE: TechRadar has updated their performance graphs to all start at 0.

 

[Zando_]

Wow... that's.... that's retarded.

[Navbryce]

That is INCREDIBLY misleading. Those are terrible charts. When I initially saw them (from your post), I was like, "wow threadripper really sucks". Then, I looked at the key. . By putting the value at the end of the bar, one would think that the physical length of the bar represents all of that value.

[Nytohan]

Just now, Navbryce said:

That is INCREDIBLY misleading. Those are terrible charts. When I initially saw them (from your post), I was like, wow threadripper really sucks. Then, I looked at the key.

Yep. I know where I'm not going for reviews.

[IntMD]

lol, i scroll wheeled down and it landed on the TR 1440p and I was... that's not a bad scale.... then I looked up at the TR 1080...... Why!

[CasualGamerCC]

Looking at how varied the units are, they probably just slapped some values into Excel and let it auto format the axis. Put into a chart with a starting value of 0, a lot of them would show a performance difference of 10-15%, which is what the story should be. I didn't look at the original article to see what their conclusions were.

[Nytohan]

32 minutes ago, CasualGamerCC said:

Looking at how varied the units are, they probably just slapped some values into Excel and let it auto format the axis. Put into a chart with a starting value of 0, a lot of them would show a performance difference of 10-15%, which is what the story should be. I didn't look at the original article to see what their conclusions were.

That sounds plausible, and we should probably attribute this to stupidity over malice, but it is at a minimum a huge oversight. It gives casual observers the idea that a minor difference in performance which is within the margin of error is in fact a noteworthy gap that should seriously impact your buying decision.

Also, looks like it's been fixed!