Thursday, November 24, 2016

Happy Thanksgiving!

If you are reading this, then you probably have a computer or smartphone, electricity, shelter, food, and all kinds of things that make living in the 21st century a lot nicer than most of the other centuries.*

There'd be a lot more to be thankful for, though, if you had your computer or smartphone just twenty years ago when the whole world would have completely dumbfounded by it. In fact, the government probably would have had your computer or smartphone confiscated in the name of national security so I guess you'd be less thankful for that.

In 1996, people still had dialup internet and a power user 0.2GHz Windows machine set you back $5,000 (almost $8,000 in 2016 dollars).

At the time, this got my heart racing

Flipping through the August 1996 issue of Boot magazine shows how far we've come but also a bit of what we've lost. It might have been part of an unsustainable tech bubble brought down by the prosaic realities of the Business Cycle, but at least we got these fevered dreams of the future. Cyberpunk, virtual reality, Skynet. There's little to be thankful about our current asset bubble. I suppose we've traded away some future prosperity to prop up the deteriorating grandeur of McMansion America. 

Even then, I think about my great uncle who was a top flight lawyer back in the Philippines. He had a car and a refrigerator back in the 1930s. That was high living even by American standards so you can imagine how impressive it was in the Philippines.** No one brags about owning a refrigerator today.

Even the first generation iPhone obliterates 1996's Dream Machine once you add in speakers, monitor, and input devices. This is why buying high quality peripherals makes more sense than buying faster hardware. Now if only someone would make a good mouse.***

Apart from IT advances, life in 1996 was largely the same and I'd still be thankful for many of the same reasons so it gets old. You're never really thankful until you no longer have those things. And in an age of increasing abundance, the potential of losing it all becomes rarer. 

Less than a century ago, around 30% of Americans were farmers. Being thankful for a good harvest was near universal and that sense of privation and bounty were a natural incentive to the formation of good people. Voluntary privation, whether it's fasting or cutting back or even going camping are ways to bring that sort of character building back. There's some sense of that in Survivor or disaster themed shows which, at least personally, is part of their appeal; though, I would never wish disaster on civilization just to make people feel thankful later on – let alone for the economically dubious reason of increasing demand to help the economy. 😆😆😆****

I really, really, really, like the face with tears of joy emoji but Microsoft's version is poorly done (😂??) so here's a superior version I found online.

* Or maybe you don't have any of those things and are a homeless guy reading this at a library but if I were homeless I would be loading up on turkey at the local shelter since Thanksgiving is one of the few days when society shifts focus somewhat toward the less fortunate.

** He bought into the fast life and lost it all.

*** Low input lag, firmware configured settings, durable finish, easy to clean, custom ergonomics, no sensor issues like angle snapping or acceleration, repairability, quiet clicks. As a lefty, my choices are sharply curtailed and I'm using the Logitech G900 after having gone through two faulty Logitech mice and the infuriating Razer software. The G900 hits most marks but: isn't easy to clean, its finish is starting to wear, probably difficult to repair, and sounds like a nail clipper with every click.

**** Yeah, couldn't help taking a swipe at Krugman and his Keynesian buddies but I can't help feeling perversely thankful for the Establishment getting taken down several notches this election year. They haven't given up (and why would they? The odds favor the house at the end) but if the corporate tax rate is reduced to 15%, that will provide favorable conditions for entrepreneurship – the key reason we have all these things to be thankful for in the first place.

Tuesday, November 15, 2016

Methodology done RIGHT

"Pick your battles"
Some battles are a lost cause. Decimate is beyond saving. The historically specific meaning about the Roman practice will probably stick around in specialized discussion, but the proper general meaning of "a 10% loss" isn't coming back.

The same is probably true of methodology and problematic. I only complain about these because I see them used incorrectly all the time. I'm being a hypocrite here because I am a bad writer wont to begin sentences with "and", intentionally use sentence fragments, on top of all kinds of other writing mistakes.

But a tech article this morning got it half right, and that ain't bad. The writer still conflates method and methodology here:
I've been working behind the scenes on a radically new test methodology.
Yet what follows is a discussion of why he believes some methods are better than others, and why a particular method was chosen for his review. In other words, actual methodology. This made me very happy! The article is a review of the Samsung 960 EVO SSD which is of limited interest, but within his methodological discussion is this gem
 Analyzing trace captures of live systems revealed *very* low Queue Depth (QD) under even the most demanding power-user scenarios, which means some of these more realistic values are not going to turn in the same high queue depth ‘max’ figures seen in saturation testing. I’ve looked all over, and nothing outside of benchmarks maxes out the queue. Ever. The vast majority of applications never exceed QD=1, and most are not even capable of multi-threaded disk IO. Games typically allocate a single thread for background level loads. For the vast majority of scenarios, the only way to exceed QD=1 is to have multiple applications hitting the disk at the same time, but even then it is less likely that those multiple processes will be completely saturating a read or write thread simultaneously, meaning the SSD is *still* not exceeding QD=1 most of the time.
I will admit to confirmation bias here since I've long believed QD1 to be the most important SSD metric. And until developers start paying attention to multi-threaded disk IO, QD1 will remain important in much the same way CPU frequency and memory latency are generally more important than CPU cores or memory bandwidth.

* Ironically, people who think the incorrect use of decimate should be acceptable like to defend their illogical view by accusing their opponents of employing the rather logical sounding etymological fallacy, which, Wikipedia tells me, is a species of genetic fallacy.

But these sorts of fallacies are informal ones, i.e., undeserving of sharing even part of the certitude that "fallacy" conveys in formal logic.

Wednesday, November 9, 2016

Things to look forward to in 2017

Thankfully the election is over! For the amount of time people, myself included, devote to this thing, the actual impact the president has on most people's lives is very low. Ask yourself how Obama's eight years have affect you personally. Me? I've had to pay fines for not having health insurance. Thanks Obama.

But here are some cool things that will probably have more effect on my life (and maybe yours too if you are a filthy computer nerd) next year than the president ever will! That is, unless the Draft is called; one reason to prefer candidates who want isolation peace.

OLED displays

eBay was running a deal for a 55" 4K OLED TV for $1500 a few days ago. So I think that's a sign of things to come. Back in 2008 I was predicting that OLED prices would quickly drop from the stratospheric $2,500 Sony was asking for an eleven inch OLED TV and dominate the display market by 2010 with cheap, huge, flexible screens.

Boy was I wrong. But these things are finally getting cheap and are the very-best-like-no-one-ever-was of display technologies. The deepest blacks, super low persistence, outstanding viewing angles and colors. TN, IPS, VA, and even Plasma/CRT, have too many compromises that OLED takes care of. OLED might degrade faster than other technologies and use more power in some situations, but it's a small price to pay for picture and motion quality supremacy. As a parenthetical, even though OLED is capable of very high brightness, I'm not sure how well that brightness holds up in low persistence modes, e.g., most strobing monitors have drastically lower brightness in strobe mode although EIZO seems to have figured out a way to mitigate the issue with its Foris monitors.


Intel got a bit of its OC groove back with the Haswell Refresh, but this years Skylake and next year's Kaby Lake look to be a firm step back to good overclocks. The mainstream Haswell i7 averaged about 4.47GHz whereas Refresh and Skylake average 4.65GHz on air.

Kaby Lake is clocking in 4.2GHz stock turbo boost frequencies so I'm really hoping 5GHz chips become a regular feature on Silicon Lottery.

Speaking of which, Silicon Lottery now offers delidding for the i7-E CPUs (which already use a higher quality TIM, i.e., Thermal Interface Material) than the normal i7. The thermal advantage isn't as large as say, delidding a 4770k, but it looks to be nearly 10C. Impressive! I don't know about the price, but removing the solder-based TIM from an enthusiast i7 is a lot more involved. It's still worth it, I think, given how long we are holding on to CPUs these days.

An easier route to get cooler temps? More people are experimenting with cooling the underside of the processor with basic heatpads, heatsinks, and fans. It should be fairly simple and inexpensive while offering several degrees of additional cooling. It makes a lot more sense for motherboard and case manufacturers to accommodate this rather than put features geared towards LN cooling and external radiator setups on power user products.

Alphacool has developed a replacement for the venerable D5 pump which promises to offer the same performance with lower noise and vibration.** Pump noise drives me nuts and is one of the two things I hate about watercooling. Maintenance is the other, though now that the All-In-One (AIO) units are getting really good; if Alphacool creates a premade AIO with this new pump, it'll go into my next build. It might just be on par with the D5 so this sounds like a job for SilentPCReview.

Intel's fancy new Turbo Boost 3 (TB3) seems to be a way to realize gains from per-core overclocking. Right now it's only on their latest i7-E chips, but that's where it makes the most sense. Usually, as you add CPU cores, the top frequency drops because people usually limit their overclocks to the weakest core. This is because it's not only easier to test all the cores at the same time for stability but it's a pain to make sure the different required voltages are supplied properly and make sure that core affinities are set optimally. Even with core affinity automation software like Process Lasso, I read somewhere that a given core assignment doesn't always correspond to the same physical core.

TB3 fixes that by uniquely identifying each core according to their quality and automatically assigning the most intensive threads appropriately. In theory, that means that buying a processor with more cores should mean a higher top frequency because the chances of getting a golden core improve with more cores.

But my experience with Intel's software has been just wait for the next version – if there is a next version. Step in the right direction though.***

Caching Hardware

The Integrated Memory Controller (IMC) on Intel's chips seems to have been refined bigly with Skylake. It's still the silicon lottery with both processor and RAM, but Silicon Lottery also tests IMC strength – for a price.

Speaking of RAM, 2016 saw the widespread adoption of Samsung's high performance B-die in DDR4 sticks. Perfect for the beefier Skylake IMCs (though a quality motherboard is also required to achieve good overclocks). Synergy.

Optane should be hitting the market in 2017 along with appropriate supporting hardware. It's meant to be much faster than SSDs though early benchmarks give me the suspicion that this is might end up RAMBUS-tier. But if it can deliver the low queue depth 4K goods, then sign me up. If not, Samsung's recently released 960 Pro should be on the short list of any 2017 build. Unlike most SSDs, some kind of extra cooling is needed to keep the 960 Pro (and its OEM predecessor SM961) from thermal throttling, something I take perverse delight in.


Zen is probably not going to be as fast as Intel's best. So 2017 will probably see the prices on Intel and nVIDIA's flagships creep up. Most people don't buy the top of the line CPU or GPU, but competition at the top means fiercer price competition at all levels. If AMD can produce a $1000 10-core CPU that outperforms Intel's $1700 10 core CPU, not only does Intel's 10 core CPU price get slashed to around $1000, but it also means Intel's $1000 8 core CPU gets a price reduction and so on.

Hope springs eternal, but maybe it's a return to Athlon form. Zen, like Athlon, is an elegant architecture name. Note to AMD, names like: Hammer, Bulldozer, Piledriver, and Sledgehammer, don't really make sense if the product performs worse than the competition. I don't know what I'm going on about but I do know what I want. And that is ... for Jim Keller to Make. AMD. Great. Again.^


Pretty sure Jim Keller's pixie dust didn't land on the ATI team so I Vega 10 is probably going to be a bust. It might be a good product but I think nVIDIA has already telegraphed, via locked voltages on its newest cards, an intention to drop a slightly weaker Titan Pascal with whatever clocks are necessary to beat Vega 10. Even if Vega 10 wins in DX12 performance, that (still) won't be relevant in 2017. On the other hand, ATI vs nVIDIA seems to have had a lot more back and forth than AMD vs Intel.


All indications are that Sony's PS4 VR is a great product and the Chinese are going to have their Vive/Rift clones out soon which will be good for everyone. The physical and spatial aspects that VR adds could be a powerful adjunct to meditation and learning. It's the Wild West out here and it's wonderful.

End notes:

** Aquarium filter manufacturers are you listening?

*** Intel has this legendary reputation in hardware that makes its software shortcomings hysterical. On reflection, Intel hasn't been infallible at all but there's this kind of reverence computer people have toward the company. Like pre-2008 Greenspan.

^ I know he left but writing "Jim Keller to have 'Made AMD Great Again'" is awkward

Tuesday, November 8, 2016

Gaming the vote

A few weeks after I send in my ballot, the election commission sends me a letter asking me to update my signature. It's apparently changed from the time I registered. The only time I'd bother to update that signature is in the Nozickian situation where my vote actually did count, i.e., I am the tie-breaker.

Statistically, it's unlikely.

But is the statistical unlikelihood of being the tie-breaker a mathematically compelling argument against voting? I have my doubts. Everyone else is a potential tie-breaker and so the correct strategy is for no one to vote. But as the number of voters go down – assuming a homogeneous reduction – the likelihood of being the tie-breaker goes up.

What about a heterogeneous reduction in the numbers of voters? If it wasn't obvious, the decision of whether to vote has some Prisoner's Dilemma type aspects. Deciding to vote carries with it a burden of having to pay at least some attention to political ads, arguments, and all the other things that make the election season wearying.

Imagine two Nozickian voters on opposite sides, Red versus Blue. If Red votes but Blue does not, Red wins and vice-versa. If both vote, it's a tie but both will have wasted time and money in the process. If neither vote, neither will have wasted time and money. The stable "intelligent" choice is to not vote. But if the policy differences appear large enough, the payoff for coöperating becomes insignificant.

If this election's presidential contest were between The Apprentice Trump and First Lady Clinton, no one would vote. But when it's between Literally Hitler Trump and In Your Heart You Know She Might Clinton, the payoff is tantamount to saving the world. We saw this earlier with Brexit where I could not understand the hysteria over a largely inconsequential vote.

Although Trump is definitely the non-establishment protest figure in the election (and the candidate I would expect to win), neither offer anything radically different*. So the stable intelligent choice is still to not vote. However, if there is a meaningful difference in intelligence between Red and Blue, then the stable intelligent choice becomes unlikely because less intelligent people are less coöperative. Thus, given unequally intelligent players, the correct strategy is to vote even if the outcome is almost guaranteed to be less than optimal.

* Principled to the point of consistency. The only believable differences are a stricter immigration policy and lower corporate taxes on Trump's part.