Benchmarks as Fast As Possible

Techquickie ·Techquickie ·2015-05-07 · 982 words · ~4 min read
Floatplane YouTube

Transcript

JSON SRT VTT 73
0:00 life used to be so simple when we wanted to know how computer a compared to
0:04 computer B all we had to do was look at the little tag under it and if you know
0:09 number a was 20% more than number B we knew that give or take computer a was
0:13 going to be 20% faster than computer B but then as architectural designs of
0:17 different processors diverged and a CPU
0:20 that ran at 2 GHz could suddenly be as fast at performing a given real world
0:25 task is another one that ran at 3 GHz we needed a new way to measure performance
0:31 benchmarks no I'm not talking about those kinds of benchmarks I'm talking
0:35 about software programs designed to evaluate the performance of PC
0:40 components or complete systems they fall into two main categories the first main
0:45 type is synthetic benchmarks examples of these include ID to 64's cach and memory
0:51 speed testing module and super pie a small lightweight program that tests how
0:55 quickly a processor can calculate pi to
0:58 a predetermined number of decimal places now synthetic benchmarks are great for
1:03 certain things because their creators are focused on obtaining results with
1:06 maximum repeatability a key characteristic of any test that is to be
1:10 used for scientific style comparisons and minimal potential bottlenecks
1:15 elsewhere in the system to simplify comparisons across different test
1:19 benches and prevent a situation where say you're trying to I don't know test
1:23 the speed of your shiny new Lightning Fast SSD by copying data from a hard
1:29 drive a much slower device in the first place which would make all ssds appear
1:34 to perform identically to each other the same speed as a hard drive the problems
1:39 with synthetic benchmarks are several though first up is that because they
1:43 work to ignore system bottlenecks they often deliver results that don't reflect
1:48 the end user's experience because just because one component is getting twice
1:53 as fast doesn't mean that the rest of the system improved in any tangible way
1:57 and second because benchmarks Run High specialized workloads they are
2:01 notoriously misused by Hardware manufacturers who might manipulate their
2:06 Hardware or drivers to perform better in Industry standard benchmarks even if
2:11 this comes with no day-to-day benefit or even with reduced performance in real
2:17 world tasks which leads us to the other main type of benchmarks realworld
2:22 benchmarks simple versions of these like cinebench or The Benchmark drop down
2:26 built into szip take a predetermined but
2:29 otherwise totally real workload in these cases 3D rendering or file compression
2:34 and decompression and have the computer process it as fast as possible then spit
2:38 out a number to tell you how well it went while more complicated real world
2:42 benchmarks often need to be custom developed by PC Hardware Publications
2:47 like aanch with their storage Benchmark Suite that was created by recording the
2:52 actions of a user performing everyday tasks on a computer than running that IO
2:56 trace on every new drive that enters the lab but some benchmarks don't really
3:01 fall cleanly into either of these two categories future marks PC Mark and 3D
3:06 Mark series are great examples here they feature components that are purely
3:11 synthetic like measurements of in-game physics performance on a CPU components
3:16 that are real world based like webpage loading times and then they spit out a
3:20 score that in theory quantifies the performance of the machine all right
3:26 lonus then so which one should I believe
3:29 synthetic or real world or like the hybrid model and the answer is kind of
3:34 all three if a real world Benchmark is available for your exact given workload
3:39 let's say 4K video editing then go with that first but if it's not available or
3:45 you're planning for a future hypothetical workload a hybrid or
3:49 synthetic Benchmark may be your only choice so just make sure that you
3:53 educate yourself about what the numbers you're looking at mean before making a
3:57 purchasing decision and speaking of educating yourself about numbers before
4:01 making a purchasing decision dollarshaveclub.com you got to educate
4:05 yourself about the numbers in this case price tags before you make a purchasing
4:10 decision in this case razors Dollar Shave Club delivers high quality razors
4:14 straight to your door for much less and it's amazing because they're actually
4:18 like it includes shipping in the fact that they cost less than the ones that
4:21 you actually have to get off your butt go to the store get them to unlock the
4:26 stupid cabinet and buy in person I mean
4:29 this is more work for you and you're spending more money great little program
4:33 there save money and save time actually I think their slogan is shave time shave
4:38 money which is very funny they're funny guys and those funny guys also have
4:42 other products they've got their Dr Carver shave butter which goes on clear
4:45 so you can see what you're doing while you're shaving they've got their one wipe Charlie which also help you see
4:50 what you're doing when you're wiping your butt with peppermint scented wipes
4:54 and uh get over to dollarshaveclub.com Linus to join the club and start looking
4:59 good without having to go out of the house looking like a scrub in the first
5:02 place to get the supplies you need to look good when you leave the house
5:07 thanks for watching this episode as fast as possible guys like this video if you liked it dislike it if you disliked it
5:11 leave a comment letting me know if you have suggestions for future fastest possibles and as always if you aren't
5:15 already don't forget to subscribe and share this video if you thought it was
5:19 helpful in somewhere