Register a SA Forums Account here!
JOINING THE SA FORUMS WILL REMOVE THIS BIG AD, THE ANNOYING UNDERLINED ADS, AND STUPID INTERSTITIAL ADS!!!

You can: log in, read the tech support FAQ, or request your lost password. This dumb message (and those ads) will appear on every screen until you register! Get rid of this crap by registering your own SA Forums Account and joining roughly 150,000 Goons, for the one-time price of $9.95! We charge money because it costs us money per month for bills, and since we don't believe in showing ads to our users, we try to make the money back through forum registrations.
 
  • Post
  • Reply
Paul MaudDib
May 3, 2006

TEAM NVIDIA:
FORUM POLICE

Llamadeus posted:

I'm fairly sure that a 2500K would struggle even with decoding blu-ray quality HEVC.

Oh. Yeah, if those are H265 (like, you copied the BDMV folder off the disk) then you're going to have problems. I can tell you for sure that my 4690K can't handle decoding 4K H265. I don't think my 5820K would handle them either.

I thought these were :filez: that had been transcoded to 4K H264, which is what you're going to need to do to have a shot at transcoding them in realtime right now.

Not sure about Raven Ridge but I do know that Apollo Lake/Gemini Lake can decode HEVC Main10 in QuickSync, and it also has an encoder that could do 1080p, so ironically an Atom may be your best option. You can pick up a NUC for like $150 I think.

Paul MaudDib fucked around with this message at 05:41 on Aug 13, 2018

Adbot
ADBOT LOVES YOU

priznat
Jul 7, 2009

Let's get drunk and kiss each other all night.
Yeah it is a HEVC rip, about 50Mbps bitrate. It was more just a try ripping a disc to see how it turned out, I don't plan on doing it that much anyway..

LRADIKAL
Jun 10, 2001

Fun Shoe
I think you would want to upgrade the graphics card to one that supports the hardware decode and encodes you need.

https://developer.nvidia.com/video-encode-decode-gpu-support-matrix

Paul MaudDib
May 3, 2006

TEAM NVIDIA:
FORUM POLICE
If you're doing hardware encode/decode tasks, you want the latest gen, since NVENC encoder quality has noticeably bumped with every release. Kepler's NVENC is hot garbage compared to what Pascal can do, and I'm sure it'll bump again with Ampere/Turing/whatever. (edit: Previous gens also have HEVC decode but no Main10 support, so you'd definitely want Pascal) Do note that consumer cards only support 2 NVENC encodes at a time.

Fun fact, the GT 1030 does not have a NVENC module, so the cheapest option would be a 1050. If you want to be able to game on it at a future date, 1060 3 GBs are about $150 used.

AMD's video encoder is terribad, so don't bother. There is a newer revision of the media core in Raven Ridge, and it may have some improvements, but all discrete cards and previous-gen APUs are really bad. Oddly, Tonga is the only card that supports B-frames in its H264 encoding... support was regressed in Fiji, Polaris, and Vega.

Quicksync isn't great, but it's a little better. Oddly, Haswell and Broadwell have a kinda funky generation that regresses quality somewhat. I've heard you can tune settings to help a bit, but I don't know specifics. Nice thing, there's no hardcoded limit to stream count.

priznat posted:

Yeah it is a HEVC rip, about 50Mbps bitrate. It was more just a try ripping a disc to see how it turned out, I don't plan on doing it that much anyway..

Yeah unless you get hardware decode support, you're probably going to have to transcode that one ahead of time.

If you want a quick sanity check, it's not too hard to set up ffmpeg, or there is Staxrip if you need a GUI. Use like a veryfast preset.

Paul MaudDib fucked around with this message at 06:53 on Aug 13, 2018

ufarn
May 30, 2009
2,500 bitrate is a weird test case for encoding quality, but there aren't really any good encoding benchmarks out there, unfortunately.

Only GN's for performance benchmarks - not quality.

repiv
Aug 13, 2009

Leaked slides confirm that at least some of the CFL refresh parts will be soldered.

mewse
May 2, 2006


I thought we already knew the i9 would be soldered.

100% Dundee
Oct 11, 2004

mewse posted:

I thought we already knew the i9 would be soldered.

I think it's just been rumors up until now. But much like all other processor rumors regarding clock speeds and core counts and retail pricing etc they are usually 100% right.

BIG HEADLINE
Jun 13, 2006

"Stand back, Ottawan ruffian, or face my lumens!"
Yeah, but I won't believe it until Der8auer confirms it's soldered on a retail model - not an ES.

Cygni
Nov 12, 2005

raring to post

There is no physical way to get 8 Lake cores at 5ghz on a 95w budget, so we either lookin at a situation where it will just blow by the TDP, or something like the 8700 non-K where the chip will boost to that number for 30 seconds or so and then cut itself back to stay under the 95w budget, unless you go into the BIOS and let it ignore the TDP.

For comparison, the 8 core 7820X will pull over 180 watts by itself at max load at 4ghz, and something like 300 watts when overclocked to 5ghz. So this might actually be a situation will your VRM may matter on a Z390 board.

Khorne
May 1, 2002

Cygni posted:

There is no physical way to get 8 Lake cores at 5ghz on a 95w budget, so we either lookin at a situation where it will just blow by the TDP, or something like the 8700 non-K where the chip will boost to that number for 30 seconds or so and then cut itself back to stay under the 95w budget, unless you go into the BIOS and let it ignore the TDP.

For comparison, the 8 core 7820X will pull over 180 watts by itself at max load at 4ghz, and something like 300 watts when overclocked to 5ghz. So this might actually be a situation will your VRM may matter on a Z390 board.

edit: Misinterpreted what was being posted.

Khorne fucked around with this message at 22:48 on Aug 16, 2018

orcane
Jun 13, 2012

Fun Shoe

Cygni posted:

There is no physical way to get 8 Lake cores at 5ghz on a 95w budget, so we either lookin at a situation where it will just blow by the TDP, or something like the 8700 non-K where the chip will boost to that number for 30 seconds or so and then cut itself back to stay under the 95w budget, unless you go into the BIOS and let it ignore the TDP.

For comparison, the 8 core 7820X will pull over 180 watts by itself at max load at 4ghz, and something like 300 watts when overclocked to 5ghz. So this might actually be a situation will your VRM may matter on a Z390 board.

Both, really. The TDP is theoretical already (for current Intel CPUs it's "all cores, base clocks, specific complex load") and even with current CPUs, turbo clocks or stuff like AVX pushed CPUs way past the listed TDP. At 3.6 GHz it should be a 95W CPU at least :v:

Also, Asrock had a warning up on their leaked micro site regarding i9 on H310 boards, so I'd expect the lower end chipsets (and probably even weaker Z370 boards) to struggle running turbo clocks for long, let alone constantly.

The Illusive Man
Mar 27, 2008

~savior of yoomanity~

Welp, sold. Hitting 5 GHz 24/7 on 8 cores without having to buy special tools/thermal compounds or void my warranty was the final nudge I needed to upgrade my 6700K.

redeyes
Sep 14, 2002

by Fluffdaddy
Gonna be $500 bux for that 8 core I bet.

On a similar topic. I got a 8700K with AsRock Z370 Extreme 4. So far, I cannot ever get my Corsair LPX DDR4 3200 to run at 3200 via XMP settings. 3000 works fine but 3200 just does a boot loop 3 times followed by the mobo reporting a failure to boot and reverting boot settings.

If anyone has any clues what I might try that would be swell. I've tried upping memory voltage to 1.4v (1.35v is the normal 3200 XMP voltage) which did nothing.

GRINDCORE MEGGIDO
Feb 28, 1985


Check the vccio and vccsa voltages.

redeyes
Sep 14, 2002

by Fluffdaddy

GRINDCORE MEGGIDO posted:

Check the vccio and vccsa voltages.

What am I checking for? What would be reasonable voltages for those settings?

[edit] @ 5.0Ghz with AVX load, my 8700K Intel Extreme tuning reports near 180w TDP (or power usage). Jeeeezus

redeyes fucked around with this message at 16:16 on Aug 18, 2018

GRINDCORE MEGGIDO
Feb 28, 1985


redeyes posted:

What am I checking for? What would be reasonable voltages for those settings?

[edit] @ 5.0Ghz with AVX load, my 8700K Intel Extreme tuning reports near 180w TDP (or power usage). Jeeeezus

Try about 1.1v or thereabouts. It might be giving it that much at "stock" (xmp) settings though. 

Maybe it needs a bit more at 5ghz clocks. 

GRINDCORE MEGGIDO fucked around with this message at 16:24 on Aug 18, 2018

redeyes
Sep 14, 2002

by Fluffdaddy

GRINDCORE MEGGIDO posted:

Try about 1.1v or thereabouts. It might be giving it that much at "stock" (xmp) settings though.

Maybe it needs a bit more at 5ghz clocks.

Ah will give that a try. I got a Cryorig H5 which didn't really mount right, it's all loosy. My temperatures are great %90 of the time, and then BAM 100c suddenly and throttle. Gotta be something wrong with my mounting technique. This system is touchy.. which makes sense since Intel poo poo it out thanks to Ryzen.

GRINDCORE MEGGIDO
Feb 28, 1985


About 1.1 is safe and it'll be somewhere roughly around there. This is from Asus's oc guide for kaby

DDR4-2800 ~ DDR4-3600

Vccio 1.10V~1.25V
Vccsa 1.10V~1.30V

My 6600k ran 3600 ram with 1.15 / 1.2 afaik. At "default" it was too high.

Ouch 100C. Get a 280 aio on that fucker.

redeyes posted:

This system is touchy.. which makes sense since Intel poo poo it out thanks to Ryzen.
Heh, my 2700x needs asrock to make a better bios asap.

GRINDCORE MEGGIDO fucked around with this message at 16:41 on Aug 18, 2018

redeyes
Sep 14, 2002

by Fluffdaddy
Tried those voltages, nothing different. On a whim I decided to put the memory sticks in slots B in stead of A and what do you know.. 3200 on the money. Odd! It infers slots B are routed more directly to the CPU.

redeyes fucked around with this message at 03:05 on Aug 19, 2018

GRINDCORE MEGGIDO
Feb 28, 1985


redeyes posted:

Tried those voltages, nothing different. On a whim I decided to put the memory sticks in slots B in stead of A and what do you know.. 3200 on the money. Odd!

:discourse:

redeyes
Sep 14, 2002

by Fluffdaddy
I remounted the Cryorig too, I think maybe a dog hair got into the heatsink goop because It maxes out 5Ghz at about 80c AVX but generally in the 60s with normal max load. Gotta love dogos.

Anime Schoolgirl
Nov 28, 2002

redeyes posted:

Tried those voltages, nothing different. On a whim I decided to put the memory sticks in slots B in stead of A and what do you know.. 3200 on the money. Odd! It infers slots B are routed more directly to the CPU.
this is actually really common on most boards on both platforms in general, things just work better magically on the second set of RAM slots

BobHoward
Feb 13, 2012

The only thing white people deserve is a bullet to their empty skull

redeyes posted:

Tried those voltages, nothing different. On a whim I decided to put the memory sticks in slots B in stead of A and what do you know.. 3200 on the money. Odd! It infers slots B are routed more directly to the CPU.

The opposite, actually.

Anime Schoolgirl posted:

this is actually really common on most boards on both platforms in general, things just work better magically on the second set of RAM slots

It's not magical at all. If you install just one DIMM per channel, and you use the slots closer to the CPU, electrically speaking there is an unterminated stub hanging off the end of the transmission line (aka PCB trace) connecting that DIMM to the CPU. Long, unterminated stubs aren't good for signal integrity.

Motherboard manuals usually have a guide showing the order you're supposed to use when populating DIMM slots. The recommended order within a single DDRx channel (2 or more of slots that are bussed together) is very predictable: they're always going to tell you to start with the one furthest from the CPU and work your way inwards, and it's for exactly this reason.

(Technically, in this diagram of one channel:

CPU---DimmA--DimmB

If you populate just DimmB, DimmA's slot is still an unterminated stub. However, the stub length is really short, basically just the contacts in A's connector. That doesn't hurt signal integrity as much as the case where you populate just A, in which case the stub consists of the traces connecting slot A to slot B, and B's connector too.)

redeyes
Sep 14, 2002

by Fluffdaddy
Ah! Very interesting information. I don't suppose there is such a thing as a RAM slot terminator? (getting flashbacks of RAMBUS)
I did look at the manual again and it does say populate slots 2 and 4 first. How did I not notice this before?! My bad.

redeyes fucked around with this message at 19:08 on Aug 19, 2018

redeyes
Sep 14, 2002

by Fluffdaddy
I was looking for a fan like a Gentle Typhoon except 140mm. Something that do at least 2000rpm. Any ideas?

aaand I found this but I just want one

https://www.amazon.com/EKWB-EK-Furi...us+Vardar&psc=1

redeyes fucked around with this message at 17:53 on Aug 19, 2018

GRINDCORE MEGGIDO
Feb 28, 1985


ML140?

LRADIKAL
Jun 10, 2001

Fun Shoe
I like the be quiet! Pure Wings, but I can't honestly say I've compared them side by side to other fans.

GRINDCORE MEGGIDO
Feb 28, 1985


NFA14 industrial is another good one.

redeyes
Sep 14, 2002

by Fluffdaddy

GRINDCORE MEGGIDO posted:

NFA14 industrial is another good one.

Ah yeah I missed that one. Spendy but I bet it will last forever.

Potato Salad
Oct 23, 2014

nobody cares


redeyes posted:

Ah! Very interesting information. I don't suppose there is such a thing as a RAM slot terminator? (getting flashbacks of RAMBUS)

RAM?

redeyes
Sep 14, 2002

by Fluffdaddy

Random access memory. DDR4. You may have heard of it.

Potato Salad
Oct 23, 2014

nobody cares


redeyes posted:

I don't suppose there is such a thing as a RAM slot terminator?


as in terminate with ram ( I was failing at being funny)

BobHoward
Feb 13, 2012

The only thing white people deserve is a bullet to their empty skull

redeyes posted:

Ah! Very interesting information. I don't suppose there is such a thing as a RAM slot terminator? (getting flashbacks of RAMBUS)

I don't think there is, other than plugging in a real DIMM. DDRx is just designed to be OK with the small T-stubs from unpopulated slots in the middle of the channel. (But... have you noticed how the maximum number of DIMMs per channel drops as DDRx clock increases? Yeah.)

redeyes
Sep 14, 2002

by Fluffdaddy

My bad. That was actually funny. It's hard to read internets sometime.

So my next issue. *sigh*

Got a Cryorig H5 and while it does cool my 8700k well, it doesn't feel like its working right. The main issue right now is, I run a burn test and physically touch the bottom of the heat heatpipes, they dont really feel hot to the touch. Surely that isn't right?! Obviously those pipes should be pipin' hot *ahem*. Has anyone noticed this type of thing? Should I pull the whole mobo out of my case and remount? Maybe lap the base?

mewse
May 2, 2006

redeyes posted:

Got a Cryorig H5 and while it does cool my 8700k well, it doesn't feel like its working right. The main issue right now is, I run a burn test and physically touch the bottom of the heat heatpipes, they dont really feel hot to the touch. Surely that isn't right?! Obviously those pipes should be pipin' hot *ahem*. Has anyone noticed this type of thing? Should I pull the whole mobo out of my case and remount? Maybe lap the base?

The heat pipes use evaporative cooling so they aren't going to behave like solid metal, and if the cooler is working well then screwing around with it by lapping/etc will most likely only cause you problems

https://en.wikipedia.org/wiki/Heat_pipe

k-uno
Jun 20, 2004

Khorne posted:

Provided they use any modern scheduler, you can specify the resources you want with the job submission. Right down to socket for NUMA grouping. Googling "schedulernamehere affinity socket" will probably bring you to documentation.

You can also usually get jobs to dispatch faster by requesting the smallest reasonable duration and less RAM. In many university cluster setups, the queues with a shorter cap on runtime are also favored for quick job dispatch.

There's all kinds of neat stuff on the scheduler side of HPC that researchers never use because they don't know it exists.

Anyhow, it sounds like just building beefy lab machines is probably good enough for your use case.

Hey, quick follow-up question: if I set up a workstation to run virtual machines, is it possible to assign one entire VM to each socket? E.g. if I have a two-socket system and run two instances of Windows 10 simultaneously using VMs, can I have one copy use CPU A and one use CPU B? Doing things this way would get around a maximum core count limitation in the software I'm running, as well. If so, which would be the right VM software to use? Hyper-V? VMware workstation?

redeyes
Sep 14, 2002

by Fluffdaddy

mewse posted:

The heat pipes use evaporative cooling so they aren't going to behave like solid metal, and if the cooler is working well then screwing around with it by lapping/etc will most likely only cause you problems

https://en.wikipedia.org/wiki/Heat_pipe

While that is totally true, other systems with heatpipes get really hot. I'm going to do more testing

GRINDCORE MEGGIDO
Feb 28, 1985


Delid it

Adbot
ADBOT LOVES YOU

redeyes
Sep 14, 2002

by Fluffdaddy

I'm scared. Hold my hand.

  • 1
  • 2
  • 3
  • 4
  • 5
  • Post
  • Reply