That graph is an amazing example of misleading axes. Tempted to use it as an example of dodgy stats.
Hmm, there's something fishy about this graph charting AMD's push into Intel's server turf
Semiconductor-pokers at Mercury Research have crunched the numbers to chart AMD's resurgence against Intel's virtual x86 server CPU monopoly. Aaron Rakers, senior analyst at Wells Fargo, has seen the second 2018 quarter numbers. He told The Register: "Intel's server CPU share is estimated to have declined to 98.7 per cent vs …
COMMENTS
-
-
-
-
Monday 6th August 2018 11:52 GMT Martin an gof
Re: (misleading stats)
I've got to disagree
Sort of. If you plotted those two graphs to the same scale with the same y=0 point you would find it difficult to see any useful information. It would have been far better to keep the scale while setting different y=0 points. I think if you did that you'd find the two graphs mirrored each other pretty much exactly, which I would have thought was the whole point of the comparison - every server sale lost by Intel is a gain by AMD, or are there other x86 vendors in the market whose figures should be added to the mix? Not in the market myself, so wouldn't know.
M.
-
Monday 6th August 2018 17:11 GMT katrinab
Re: (misleading stats)
"or are there other x86 vendors in the market whose figures should be added to the mix?"
No, but there are other chips that aren't AMD64 compatible that could be used in some cases. The Itanic is I believe still shipping, and probably someone is using it. ARM is another option
-
-
Monday 6th August 2018 13:02 GMT Jon 37
Re: (misleading stats)
> People will only be mislead if they don't pay attention.
That old chestnut. "If you'd carefully read the contract, on page 97 out of 233, in tiny print, in grey-on-slightly-lighter grey, in Latin, we clearly explain it. If you were foolish enough to rely on the summary given by the salesperson instead of reading the contract and getting people to translate the Greek and Latin parts, then that's your lookout".
-
Wednesday 8th August 2018 07:47 GMT the Kris
Re: (misleading stats)
Failed comparison.
You're comparing a paragraph hidden in a 233 page document, which we can call a surprise needle in a haystack, to axis that are extremely clear. The fact that the 1st comment is about the axis shows just that, it is not hidden.
If the axis both went from 0% to 100% it would virtually show no information, that would be a useless graph.
You sound like the type of person who needs a warning on a box of razor blades: "Do not swallow, external use only", to prevent you from doing just that.
I can agree with the comment to have them at the same scale.
-
-
-
-
-
-
Monday 6th August 2018 15:19 GMT Anonymous Coward
Re: Two sine waves 180 degrees out of phase
But it isn't a duopoly. It's a monopoly, since Intel has all the power and AMD is more or less in the noise. As the article says, it is not feasible that AMD could get even 5% of market share in any really foreseeable timeframe.
Monopolies maintain their power through barriers to entry, and in Intel's case there are many of these none of which can go away at all quickly.
-
Monday 6th August 2018 19:29 GMT Joe Werner
Re: Two sine waves 180 degrees out of phase
If it was a duopoly they would not have to cheat with the axes... As somebody above wrote: AMD is just the noise on Intel's sales.
In a case like this I believe Tufte and others would recommend using
- a single graph showing a stack (not the best, you absolutely need to include the zero)
- two graphs stacked (ditto)
- a single graph showing AMD as a percentage of Intel sales
In principle one line is (as per the duopoly comment) redundant anyway...
-
-
-
Tuesday 7th August 2018 03:28 GMT More Granularity
Lisa Su Comment
During the Q2 earnings last week, Lisa was asked if the prediction for ~5% AMD capture of the server market in Q4 was still accurate, and she confirmed it.
One way for that to be true is for the linear extrapolation of the graph to be incorrect. The growth has to be nonlinear. For example, by doubling each quarter:
2Q18: 1.3%
3Q18: 2.6%
4Q18: 5.2%
Doubling may be reasonable because the large server customers have a long qualification process, followed by large purchases. That may be what is happening.
-
Tuesday 7th August 2018 16:25 GMT Anonymous Coward
Re: Lisa Su Comment
An interesting point of comparison would be the last time AMD made a big entrance into the X86 server with HP back around 2004 or so, and look at the rate of market-share gain back then. I don't have the data but my recollection was that they gained share much more quickly back then which basically traumatized Intel, drove lots of exec changes at Intel etc.
-
Wednesday 8th August 2018 15:50 GMT msroadkill
Re: Lisa Su Comment
To this layman, it seems common sense it is more like a snowball than linear, more so given the nature of the whole server ecosystem.
It was announced in march 2017, but only publicly available recently. Its a slow process while everybody tests the water, but when a few take the plunge, many follow.
Of course it is exponential.
-