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
Eyochigan
Dec 13, 2006

It's not rape unless I explicitly see it!

Watermelon Daiquiri posted:

does the flux through the ihs stay constant over its entire area if there are smaller, localized sources? or wouldn't the flux be higher above those areas?

Flux is a compound used to cover a surface to prevent oxidation during welding/soldering etc. The die(s) itself should be soldered to the IHS, and the reason stated for including the other two dies is to distribute the weight on the IHS. So in theory, it should sit flat.

edit: aw crap.

Eyochigan fucked around with this message at 21:53 on Aug 8, 2017

Adbot
ADBOT LOVES YOU

AVeryLargeRadish
Aug 19, 2011

WolfDad is Best Dad.



Eyochigan posted:

Flux is a compound used to cover a surface to prevent oxidation during welding/soldering etc. The die(s) itself should be soldered to the IHS, and the reason stated for including the other two dies is to distribute the weight on the IHS. So in theory, it should sit flat.

I think they mean flux as in variation, in this case variation of temperature.

EDIT:

AVeryLargeRadish fucked around with this message at 21:54 on Aug 8, 2017

GRINDCORE MEGGIDO
Feb 28, 1985


They mean heat flux, I think.

EF
Beaten like a kaby lake in a vice.

Watermelon Daiquiri
Jul 10, 2010




Yeah, flux as in amount of flow though a surface (integral) lol

Eyochigan
Dec 13, 2006

It's not rape unless I explicitly see it!

I imagine the space between the two dies will be significantly warmer than the edges of the heat spreader. But the edges will still get quite warm, by the assumption that a heatsink will heat up and lose its ability to disperse some of that energy.

Now I'm curious if you want full coverage of the IHS, or a smaller contact area to prevent transferring heat back into the IHS where it wasn't present before. I'd always assumed you want as much surface area as possible.

edit: Has anyone touched their cpu cooler at full load? How hot was it? I've never burned myself personally.

SourKraut
Nov 20, 2005

POST QUALITY UNDER CONSTRUCTION




Eyochigan posted:

I imagine the space between the two dies will be significantly warmer than the edges of the heat spreader. But the edges will still get quite warm, by the assumption that a heatsink will heat up and lose its ability to disperse some of that energy.

Now I'm curious if you want full coverage of the IHS, or a smaller contact area to prevent transferring heat back into the IHS where it wasn't present before. I'd always assumed you want as much surface area as possible.

edit: Has anyone touched their cpu cooler at full load? How hot was it? I've never burned myself personally.

Well the IHS is going to reach thermal equilibrium so the more cooler contact area that can be provided the better the thermal dispersion will be.

GRINDCORE MEGGIDO
Feb 28, 1985


Eyochigan posted:

I imagine the space between the two dies will be significantly warmer than the edges of the heat spreader. But the edges will still get quite warm, by the assumption that a heatsink will heat up and lose its ability to disperse some of that energy.

Now I'm curious if you want full coverage of the IHS, or a smaller contact area to prevent transferring heat back into the IHS where it wasn't present before. I'd always assumed you want as much surface area as possible.

edit: Has anyone touched their cpu cooler at full load? How hot was it? I've never burned myself personally.

I've loaded my CPU up to 75deg C, and the air cooler was warm, that's it. Probably a good 10 deg C could go away if I delidded it, but if I was going to bother I'd want to go direct die, and that's not happening on air (so CBA).

Pablo Bluth
Sep 7, 2007

I've made a huge mistake.


ServeTheHome have started to post their EPYC benchmarks
https://www.servethehome.com/dual-a...-review-part-1/

Eyochigan
Dec 13, 2006

It's not rape unless I explicitly see it!

GRINDCORE MEGGIDO posted:

but if I was going to bother I'd want to go direct die, and that's not happening on air (so CBA).

I was considering direct-die too but I don't think I'd risk it on a brand new cpu. But what's the risk factor on air. Too heavy? Athlon XP had those 4 foam pads to protect the die, I think people still managed to crush them, or maybe those were enthusiasts who shaved the pads off..

Or do you just mean water would be worth the effort at that point.

Eyochigan fucked around with this message at 23:09 on Aug 8, 2017

Combat Pretzel
Jun 23, 2004

No, seriously... what kurds?!

Threadripper to 4GHz and 1.36V does like 60-62C delta over ambient? Seems a lot? I suppose with a decent sample, it should work with less voltage and thus less heat generated?

GRINDCORE MEGGIDO
Feb 28, 1985


Eyochigan posted:

I was considering direct-die too but I don't think I'd risk it on a brand new cpu. But what's the risk factor on air. Too heavy? Athlon XP had those 4 foam pads to protect the die, I think people still managed to crush them, or maybe those were enthusiasts who shaved the pads off..

Or do you just mean water would be worth the effort at that point.

E - mine is currently intel Skylake, just chatting about direct die. Really I want a 1700 Ryzen, but I'll wait for a respin. Although I want it now particularly as Intel are not supporting 6c coffeelake on z170 / 270 for no good reason.

I suppose the risk is if the cooler rocks, or when the cooler is first fitted it isn't evenly pressing down.

With a shim, like the shim by aqua computers - I don't think it'd be that dangerous to do (I'd do it), but I think you get problems with the base of the air cooler being bigger then the socket, so it won't contact the die.

I'm not sure if removing the mounting hardware on the socket would let it work? I'm not sure if it would affect the tension much or not.

If it worked on air, I'd give it a go for sure. But I'm too lazy to remove the lid just to re add it back with better paste, although I probably should at some point. I'd rather leave it off if I could. Just because I like the fans off as much as possible with a mild overclock.

GRINDCORE MEGGIDO fucked around with this message at 23:50 on Aug 8, 2017

Scarecow
May 20, 2008

3200mhz RAM is literally the Devil. Literally.


Lipstick Apathy

Seems like someone got a threadripper early and is posting benchmarks

https://www.reddit.com/r/Amd/commen..._ghz_all_cores/



Sounds like Threadripper is going to be amazing

MaxxBot
Oct 6, 2003

you could have clapped

you should have clapped!!


Jesus I don't need 16 cores but I'll probably eventually get one because I'm a big dumb idiot.

priznat
Jul 7, 2009

Let's get drunk and kiss each other all night.

MaxxBot posted:

Jesus I don't need 16 cores but I'll probably eventually get one because I'm a big dumb idiot.

mayodreams
Jul 4, 2003


Hello darkness,
my old friend


MaxxBot posted:

Jesus I don't need 16 cores but I'll probably eventually get one because I'm a big dumb idiot.

This. I just built a 7700k box and I irrationally want a TR system instead. We should probably start a support group.

bobfather
Sep 20, 2001

I will analyze your nervous system for beer money

My overclocked i7-3770 non-K gets 41xx and 13xxx on Geekbench. Not that Threadripper isn't impressive - it totally is, but man the 3770 is almost 5 years old.

Malloc Voidstar
May 7, 2007

Fuck the cowboys. Unf. Fuck em hard.

MaxxBot posted:

Jesus I don't need 16 cores but I'll probably eventually get one because I'm a big dumb idiot.
wait for TR2, maybe amd will somehow get +50% IPC again

MaxxBot
Oct 6, 2003

you could have clapped

you should have clapped!!


Malloc Voidstar posted:

wait for TR2, maybe amd will somehow get +50% IPC again

Yeah I might hold out for TR2 not because I expect a huge IPC gain but out of (probably futile) hopes that Glofo's new process clocks significantly higher. If they just deliver their promised 10-15% IPC gain and the new process got the clocks higher it would be absolutely insane.

MaxxBot fucked around with this message at 01:26 on Aug 9, 2017

Malloc Voidstar
May 7, 2007

Fuck the cowboys. Unf. Fuck em hard.

Scarecow posted:

Seems like someone got a threadripper early and is posting benchmarks

https://www.reddit.com/r/Amd/commen..._ghz_all_cores/

quote:

Doing some overclocking testing, thought I would share some results. At 4.0 GHz, the system is stable at 1.25 V, with all cores running CB 15, so this part looks very good.

EmpyreanFlux
Mar 1, 2013

The AUDACITY! The IMPUDENCE! The unabated NERVE!


Woaaah, holy poo poo, 1.25v for 4.0Ghz? Have they tried to push it to 4.2 yet? gently caress, if AMD releases a new stepping which can hit 4.5Ghz in 2018 I'm loving sold.

Seamonster
Apr 30, 2007

IMMER SIEGREICH


I'm waiting for that 4.4-4.5 GHz regular ryzen stepping myself.

dont be mean to me
May 2, 2007

I'm interplanetary, bitch
Let's go to Mars




Seamonster posted:

I'm waiting for that 4.4-4.5 GHz regular ryzen stepping myself.

The next stepping is the key part of that, but I'd still start at 4. On desktop anyway.

eames
May 9, 2009



MaxxBot posted:

Jesus I don't need 16 cores but I'll probably eventually get one because I'm a big dumb idiot.

same, the temptation is strong even though I have no use for more than 6 or maybe 8 physical cores. fortunately for my wallet KVM/AMD is unable to fix that huge GPU passthrough performance issue so I'll probably end up with switching to Coffee Lake, particularly if the ECC support rumours turn out to be true.

also 4.0 Ghz stable at 1.25V is insane

Scarecow
May 20, 2008

3200mhz RAM is literally the Devil. Literally.


Lipstick Apathy

eames posted:

same, the temptation is strong even though I have no use for more than 6 or maybe 8 physical cores. fortunately for my wallet KVM/AMD is unable to fix that huge GPU passthrough performance issue so I'll probably end up with switching to Coffee Lake, particularly if the ECC support rumours turn out to be true.

also 4.0 Ghz stable at 1.25V is insane

Well there is a 8c x399 ment to be out at the end of the month so

Anime Schoolgirl
Nov 28, 2002






MaxxBot posted:

Jesus I don't need 16 cores but I'll probably eventually get one because I'm a big dumb idiot.

EmpyreanFlux
Mar 1, 2013

The AUDACITY! The IMPUDENCE! The unabated NERVE!


eames posted:

same, the temptation is strong even though I have no use for more than 6 or maybe 8 physical cores. fortunately for my wallet KVM/AMD is unable to fix that huge GPU passthrough performance issue so I'll probably end up with switching to Coffee Lake, particularly if the ECC support rumours turn out to be true.

also 4.0 Ghz stable at 1.25V is insane

When did they say they can't fix GPU passthrough using KVM?

eames
May 9, 2009



FaustianQ posted:

When did they say they can't fix GPU passthrough using KVM?

Nobody said that, I should have chosen my words more carefully by saying "seems unable".
The ball is in KVM's court but AMD doesn't seem interested in helping and there's zero developer activity on this >6 year old bug. I'm not expecting it to get fixed anytime soon.
Passing through graphics cards for high framerate applications () is a very niche use case so I can't even blame them.

https://bugzilla.kernel.org/show_bug.cgi?id=196409
https://community.amd.com/thread/215931

Kazinsal
Dec 13, 2011






KVM needs to fix their broken poo poo and AMD is too busy doing actually important things (like developing and releasing CPU architectures that don't suck for the first time in nearly a decade) to fix their broken poo poo for them.

At least the KVM people aren't being openly hostile to AMD about it like Torvalds vs. Nvidia.

Sinestro
Oct 31, 2010

The perfect day needs the perfect set of wheels.


You can use Xen and the performance is great, I've helped my best friend set it up on his Ryzen desktop. You have to patch the driver to fix the Code 43 "bug", and that's it.

Filthy scum
Jul 7, 2003



Sinestro posted:

You can use Xen and the performance is great, I've helped my best friend set it up on his Ryzen desktop. You have to patch the driver to fix the Code 43 "bug", and that's it.

Oh, that's nice to know. Thanks for the link

nerox
May 20, 2001


Take a look at this line in the PUBG patch notes:

quote:

Client Optimization
Optimized game performance for 6-core or higher CPU

Anarchist Mae
Nov 5, 2009

by Reene


Lipstick Apathy

nerox posted:

Take a look at this line in the PUBG patch notes:

https://www.youtube.com/watch?v=Pxeaamdwty4

And all I can think is "Why is every Phanteks case so loving damned ugly?"

SoftNum
Mar 31, 2011



Measly Twerp posted:


And all I can think is "Why is every Phanteks case so loving damned ugly?"

I like my Enthoo (though why they have such dumb names idk). Also my daughter likes to request colors on the components. (Make the ship one red!)

ZobarStyl
Oct 24, 2005

This isn't a war, it's a moider.

nerox posted:

Take a look at this line in the PUBG patch notes:
I've been on PUBG a lot lately and it can go from clean 45-60 framerates to multi second delays as soon as firefights begin. I can only hope the answer is "get 4x the cores."

ufarn
May 30, 2009


Any good videos or reads on how specifically game devs implement multicore support? Do they put netcode and physics on one core etc sort of like PhysX, or is it more about basic parallellization?

Twerk from Home
Jan 17, 2009

This avatar brought to you by the 'save our dead gay forums' foundation.


ufarn posted:

Any good videos or reads on how specifically game devs implement multicore support? Do they put netcode and physics on one core etc sort of like PhysX, or is it more about basic parallellization?

If you've got an hour to spare, the current state of the art is work-stealing algorithms and doing everything possible to minimize synchronization.

https://www.youtube.com/watch?v=0nTDFLMLX9k

NewFatMike
Jun 11, 2015



Some benches popped up of an R7 2700U APU up against the i5 7260U with Iris 640 graphics:

https://gfxbench.com/compare.jsp?be...us+Graphics+640

Against the A12 9800E:

https://gfxbench.com/compare.jsp?be...E+CORES+4C%2B8G

And it looks like the GPU is clocked at 300MHz with 11CUs:

https://gfxbench.com/device.jsp?ben...&testgroup=info

I'm getting thirsty Dr. Su.

SwissArmyDruid
Feb 14, 2014



I'm not. If Vega can't scale up and it doesn't scale down, what bleeping good is it over Fiji? poo poo had better be like, $99 compared to the i5's $309 or something.

Malcolm XML
Aug 8, 2009

I always knew it would end like this.


Twerk from Home posted:

If you've got an hour to spare, the current state of the art is work-stealing algorithms and doing everything possible to minimize synchronization.

https://www.youtube.com/watch?v=0nTDFLMLX9k

Game developers discover ancient multicore secrets

Adbot
ADBOT LOVES YOU

NewFatMike
Jun 11, 2015



From the Sisoft leaks earlier in the year, we know the APUs will do at least 800MHz.

That big Vega has a bunch of really weird bottlenecks (geometry and thermal limiting) gives me the inclination that it's designed for mobile or integrated solutions rather than full fat desktop ones. That's pure speculation, though and I wouldn't know how to confirm that.

Even still, I'm just hoping for laptops that'll do 1080p 60-ish at medium settings with Freesync 2 panels. Even if it takes 95W or so to do, I'm pretty cool with that, especially if it has a desktop equivalent APU so I can get my friends back on PC gaming on that cheap cheap.

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