Bandwidth Place 2

I got a Hero4 Black a couple months back.

It's been a solid little video camera that allows me to capture 60Mbit/s of video at 4K 30fps, or 1080p 120fps.
I am happy that a firmware update that came out this month also added 720p 240fps.

Why the higher fps? It allows the option to slowww-mooooo!

Its casing has already received badges of honor (from it rubbing against the zipline while ziplining):
http://www.mylilsite.net/images/gopro/t4damage1.jpg
http://www.mylilsite.net/images/gopro/t4damage2.jpg

Area of disappointment: YouTube.
The bitrate/quality level that YouTube uses on even their 4K resolution option is disgraceful.
Here's the quality of what I uploaded to them:
http://www.mylilsite.net/images/gopro/c ... upload.png
Here's the quality of what their end result is:
http://www.mylilsite.net/images/gopro/c ... outube.png
I mean.. look at these edges:
http://i.imgur.com/QcDICDq.png

Both are the 4K video playback, scaled down to my 1080p monitors.

YouTube also took an upload that looks like this:
http://i.imgur.com/hAG5cpl.jpg
And encoded it to look like this:
http://i.imgur.com/Tw2G0Mw.jpg

Very discouraging to me. I don't feel that uploading to YouTube is worthy due to the quality of footage I get from them, but it is the main/only video site people would actively go to watch.

BTW, 90% of my YouTube videos are not public and are specifically shared with (IRL) friends. I'm not posting their links online. Just if you were wondering.

Playing back my footage also, for the first time in approaching three years, let me realize that my PC is a bit slower than I'd like. Playing the footage uses 70%+ of my Core i7 at 4.5GHz. You know what happens when I have a video from my GoPro and a video on YouTube play at the same time? Stuttering. My PC is too slow to play two videos at once.

Basically, it wasn't video games or anything regular that makes me want to upgrade my PC, it was the sheer amount of processing required to keep up with the footage I get from the GoPro that led me to realize I need more speed.

Two things, wow does youtube do a crappy job at "encoding" your videos. Looks like hammered dog shit.

Secondly, how much faster are you going to get than an i7 at 4.5ghz? I'd imagine to get much more processing
power than that, you'd need a crazy server based setup with multiple high end CPU's ?

It may be OC'd well, but it's still an old generation CPU. A Core i7 4770k is 15.9% faster per clock than mine according to passmark. Sure, it's not that much faster and certainly isn't fast enough to make me want to put down several hundred dollars to upgrade... But the 4770k is a CPU from 2013. What I'm waiting for is Skylake: http://en.wikipedia.org/wiki/Skylake_%2 ... tecture%29

And in particular, I'm considering moving on to their high end socket which appears to be Socket 2017-A. For whatever reason, Intel only makes quad cores for their consumer-grade socket but if I move on to the high end socket I can choose perhaps a six or eight core CPU (and sounds like it'll go up to 14).

Goals for next desktop: Skylake, DDR4, PCI-Express-based SSD

Hmm.... Skylake probably will be significantly faster then your Sandy Bridge chip.

It's insane to see how much more efficient the Haswell architecture is compared to my nehalem chip.
Most information I'm finding claims an average of around 44-45% more efficient per clock cycle. I'd imagine
that the Skylake architecture will probably be heading for 60+ % more efficient then what I have.


I'm going to have to start playing some games on my system, just to see if I can really feel it's age. I am pretty
impressed, for what I do with it the machine really doesn't seem to have missed a beat. If anything the 5 year
old HDD seems to be a bit groggy, or this current windows 8 install is starting to get dodgy. But for the machine
being 5 years old now, it doesn't feel nearly as old as the last one that I had.

I'm sure if I did the things that you do, it would probably be painful if not impossible.

I'm just about ready to make the leap to SSD. Maybe even get a different case, or at least reduce the 850 w power supply
I have. I don't think it's necessary

Another thing I wish was more responsive is when viewing my pictures in Lightroom. My pictures from the DSLR are in raw format and Lightroom takes like 2-3 seconds to process each picture at the time that I go view the picture. The processing is a CPU-intensive task.
Sure, there is a lower resolution preview and the thumbnail available for me to know what each picture generally has, but when I have to go through the several shots I took of the same thing (I do that. You never know when everyone in the picture will do their best pose so you just keep snapping away and see what you get afterward.), I do want to compare the details in each one and need to wait for the processing to finish. If I could cut the processing time down to I dont know... half a second per picture, that would be SO nice.

If you look at the performance increase trend per generation (according to Passmark):
Core 2 Quad Q9550 to Core i7 920: 30.2% faster per-clock
Core i7 920 to Core i7 2600k: 34.6% faster per-clock
Core i7 2600k to Core i7 3770k: 9.1% faster per-clock
Core i7 3770k to Core i7 4770k: 6.2% faster per-clock

It's entirely possible that Skylake will give, for example, Haswell owners no reason to want to upgrade.

I'm not TOO excited about Skylake being significantly faster than Haswell because it probably isn't going to be that much faster. That's why I mentioned wanting to switch to the high end socket to get myself more cores.

As for SSDs, at this point I'm holding off on new SSDs until I can upgrade myself to SATA Express SSDs. The fastest regular SATA SSDs are all being limited by SATA III, which maxes out at 600MB/s theoretical throughput. That speed is slow. SATA Express can do 1969 MB/s.

BTW did you know that instead of doing further improvements to the SATA protocol, they just started doing direct PCI Express? SATA Express has no "SATA" in it, it's just a PCI Express connection in a cable.

I can completely understand what your saying with the picture processing. Your probably someone who takes tens of thousands of pictures every year (mostly because, like you've said, for every "keeper" you take there are probably anywhere from 5-10 more at least of the same thing that get discarded). Most people would probably say, wow... he can't wait the 3 extra seconds to view that picture? Well.... that means on average that to go through each group of photos can take significantly longer. Would it behoove the person that takes a few hundred pictures each year and does little if any editing on them to upgrade to save that time? Probably not, but it's moot because a person who takes a few hundred pictures each year probably does not have the gear like you do that takes very high resolution pictures.

Also, on the PCI Express SSD thing. Are these going to plug into a PCI Express slot? If so.... that would limit setups quite a bit unless boards are going to be redesigned. What bandwidth of PCI Express will they use? I can see it if they were using the smaller slots, or if multiple SSD's could run on the same 16x lane. Going to be interesting to see how that plays out.


When looking at the chart in your latest post, it makes sense that Skylake may not be such a big jump.

An interesting thing about the higher end boards that support more cores.... I've yet to really notice a big adaptation
of multi threaded programs? Are programs being better about that now?

SATA Express uses the PCI Express bus but does not take up any of the PCI Express slots on the board. Here's a picture of two SATA Express connectors:
http://en.wikipedia.org/wiki/SATA_Expre ... rboard.jpg

Sure, with such a big connector, it's going to limit the number of drives we can have on that sort of connector... but just like how that picture shows that there are only two SATA Express connectors, for the longest time Intel only bothered adding two SATA III (the rest are SATA II) ports to their chipsets. The reality is that 99% of customers only need the fastest connection for their OS drive and the rest of their drives can sit comfortably on slower interfaces.

Large RAID arrays are still going to consist of HDDs for the foreseeable future, and those things are definitely too slow to benefit from SATA Express.

Bandwidth-wise?

In September 2014, Intel X99 chipset became available, bringing support for both SATA Express and M.2 to Intel's enthusiast platform. Each of the X99's SATA Express ports requires two PCI Express 2.0 lanes, while M.2 slots can use either two 2.0 lanes from the Platform Controller Hub (PCH), or up to four 3.0 lanes taken directly from the LGA 2011-v3 CPU. Thus, X99 provides bandwidths of up to 3.94 GB/s for connected PCI Express storage devices

I didn't look into what programs perform well in multithreaded environments that DID NOT ALREADY perform well before, but that's because the things I want to perform faster certainly are multithreaded - video encoding (dude, 4k takes FOR. EVER.), video decoding, image processing, software compiling, etc.

According to Geekbench scores, I recalculated the performance per clock.
Core 2 Quad Q9550 to Core i7 920: 34.58% faster per-clock
Core i7 920 to Core i7 2600K: 38.08% faster per-clock
Core i7 2600K to Core i7 3770K: 4.49% faster per-clock
Core i7 3770K to Core i7 4770K: 4.62% faster per-clock
Core i7 4770K to Core i7 5775C: 6.13% faster per-clock (wasn't really a thing for very long)
Core i7 5775C to Core i7 6700K: 6.09% faster per-clock
...
Core i7 2600K to Core i7 6700K: 23.09% faster per-clock

So, respectable...ish... when you consider the 2600K is more than 4 years old, the thought that CPUs have only improved by 23% since then is a bit disheartening. I think I'll still wait for the enthusiast 6 or 8 core Skylake generation chip.