PC Perspective Podcast #436 – 02/09/17
Join us for ECS Mini-STX, NVIDIA Quadro, AMD Zen Arch, Optane, GDDR6 and more!
You can subscribe to us through iTunes and you can still access it directly through the RSS page HERE.
The URL for the podcast is: https://pcper.com/podcast – Share with your friends!
- iTunes – Subscribe to the podcast directly through the iTunes Store (audio only)
- Google Play – Subscribe to our audio podcast directly through Google Play!
- RSS – Subscribe through your regular RSS reader (audio only)
- MP3 – Direct download link to the MP3 file
Hosts: Ryan Shrout, Allyn Malventano, Ken Addison, Josh Walrath, Jermey Hellstrom
Program length: 1:32:21
Podcast topics of discussion:
-
Week in Review:
-
News items of interest:
-
1:14:00 Zen Price Points Leaked
-
Hardware/Software Picks of the Week
-
Allyn: Low cost OBD II scanner
-
Closing/outro
When you inevitably test W10
When you inevitably test W10 Game Mode, you should just go buy some random $299 HP/Dell Pentium/i3/A8 machine and throw a RX460 or 1050 (non-Ti) in it. Keep all the crap-ware, load up Steam/Origin/Uplay/Xbox apps, leave Chrome open.
If you try to validate its impact on a custom-built i7/1080/NVME then you’re not going to arrive at valid conclusions about its ability to help the peasant masses. Food for thought.
Good point.
Good point.
That is the beauty of the 8
That is the beauty of the 8 core CPU, plenty of cores to manage the OS bloat under windows 10(And Oh so Much Bloat) while still having plenty of cores remaining for any games that may need them. I suggest maybe a wireshark benchmark to measure windows 10’s telemetry activity to all those websites involved in M$’s bot net of ad pushing and data gathering madness! And that includes using folk’s computers to push out undates to add more bloat at any time 24/7 365 + 1 extra day during leap years.
Oh look there, Candy Crush wants to phone home right during the Zombie Apocalypse! And You are very DEAD now, your PC just part of the Redmond Zombie Bot Army(RZBA)!
They really should do more to
They really should do more to expose just how much spyware Windows 10 is. Been using windows since the 90s and 10 is the worst. Its literally just a free shitty rootkit.
Who said anything about 8
Who said anything about 8 core CPUs?
Unfortunately, Windows core management isn’t that good. Even with 20 threads, the computer might still only use 2-3 of your 8 cores because of the way things get sequentially assigned or queued.
Still hoping for a new
Still hoping for a new floodlight for Josh. 🙂
Guys, the Ashes “Benchmark”
Guys, the Ashes “Benchmark” is obviously fake. You can manipulate the site easily by replacing the strings and taking a screenshot. The one who did it just took the ID of a Zen ES and for whatever reason even wrote down AMD Ryzen. An ES would not ID itself with a marketing name.
Have a look at this for example.
http://cdn.overclock.net/b/be/bef18c79_cf8fa7f544af83f965426bb00f26d680.png
This is the most awesome
This is the most awesome image cover of your video to date. 😉
Damn it, missed the show
Damn it, missed the show again -_- even though I’m subscribed to the live stream mailing list .
Allyn, they have three versions of the g-sync (I think freesync is excluded) predator monitor.
The first version is named predator X34 which you described in the stream (you can NOT rotate the monitor).
The (updated) second version named predator X34A that does have the new stand as the new monitor (you can rotate the monitor).
There’s also a third (unreleased) version/revision named X34P version with extra features, such as using joystick for menu navigation and brushed back
Unfortunately, Acer product page for the X34A doesn’t show the new stand.
the first version X34 from Paul’s Hardware
https://www.youtube.com/watch?v=-FVtyZPkMx4
the second (updated) version X34A from The Tech Chap
https://www.youtube.com/watch?v=xv7VTpKbx9E
The Third (unreleased) X34P from HardwareCanucks
https://www.youtube.com/watch?v=1B0TuaryJV4
GK110 was the last GPU with
GK110 was the last GPU with 1/3 DP/SP that was also used as a Geforce with DP disabled.
It was also the last DP capable chip until they released GP100, and they even released a modified version called GK210 for the K80 while they simultaneously sold their failed 28nm GM200 shit as 980Tis and Quadro M6000s and called it a deep learning architecture since they had no room for DP cores.
As far as GDDR6 is concerned,
As far as GDDR6 is concerned, I kind of wonder if it will be short lived. They are probably going to come out with an HBM spec that is a bit of a mixture between the technologies. It will probably use stacked memory, but it will use a higher clock speed, narrower interface. That will allow the die stacks to be mounted on the GPU package in an MCM without using a silicon interposer. It will not be as good as a silicon interposer, but it could be cheaper. They could use much shorter traces compared to going off package and a lower number of solder interfaces should allow for better signal integrity. They may need to use more stacks to achieve high enough bandwidth though. It looks like Vega may only use 2 stacks for 512 GB/s with HBM.
It also might be that GDDR6 doesn’t take off because the GPUs just don’t need that much bandwidth. They already have a lot of bandwidth saving features and it sounds like they may be moving toward something similar to tile based rendering also. Tile based rendering can use a lot less bandwidth since it would be designed to run from on-die cache as much as possible.
Hmm there was that… cheaper
Hmm there was that… cheaper High Bandwidth Memory they were talking about last year at Hot Chips:
https://www.pcper.com/news/Memory/Samsung-and-SK-Hynix-Discuss-Future-High-Bandwidth-Memory-HBM-Hot-Chips-28
Stacked memory has such huge
Stacked memory has such huge advantages, it seems unlikely that non-stacked memory would remain cost effective in a year or two. The interface is a problem though. You can’t go that wide on a standard packaging material. Some of the wafer level packaging technologies are interesting. That is building the packaging onto the device while it is still on the wafer or making a reconstituted wafer if the packaging needs to be larger than the device. Those technologies allow for very small packaging. It could be close to HBM on a silicon interposer, but it would be limited by the number of output pads required. If HBM remains very expensive, the GDDR6 may have a market, but it is still going to be higher power consumption and lower bandwidth than what can be done with HBM. The fact that Vega seems to be using only two stacks seems to say that it is still very expensive or Vega bandwidth saving features are working really well, or maybe a bit of both.
Why is there no link for
Why is there no link for Ryan’s pick?