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
PC LOAD LETTER
May 23, 2005
WTF?!

Slippery Tilde

SwissArmyDruid posted:

Hm. The heatspreader has that weird rib in the middle...
The whole 4 dies on 1 package is pretty weird right now so odd solutions are probably necessary to make it all work. Maybe they were worried about the package flexing too much on install? Pressure should be pretty constant once the CPU is in the socket and the HSF is on right?

Cygni posted:

either way, the whole thing is weird.
If they're just salvaging Epyc's with dies that are a lil' too bad that can't be removed and reused for some reason and/or failed packages it makes perfect sense. Just weird to see.

Adbot
ADBOT LOVES YOU

nerox
May 20, 2001


Oh man, I know we have made jokes in the thread, but if there turns out to be a way to turn TR into Epyc with a graphite pencil, I will be so happy.

SwissArmyDruid
Feb 14, 2014



Looking over the R3 reviews... that R3 1300X looks like it is _the_ chip to get in that segment.

edit: Wow, I take it back, I just looked at the pricing. $130 and no graphics silicon compared to the i3s? :hmmm:

SwissArmyDruid fucked around with this message at 17:39 on Jul 27, 2017

Risky Bisquick
Jan 18, 2008

PLEASE LET ME WRITE YOUR VICTIM IMPACT STATEMENT SO I CAN FURTHER DEMONSTRATE THE CALAMITY THAT IS OUR JUSTICE SYSTEM.



Buglord

There is no way AMD are plopping 2 extra dies onto a TR package. Barring some sort of strange technical limitation where they need to use 4 interconnects this is pretty ridiculous of a rumour.

Twerk from Home
Jan 17, 2009

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


SwissArmyDruid posted:

Looking over the R3 reviews... that R3 1300X looks like it is _the_ chip to get in that segment.

edit: Wow, I take it back, I just looked at the pricing. $130 and no graphics silicon compared to the i3s? :hmmm:

Street price on the R5 1400 is $135 right now, is hyper-threading worth $5?

AVeryLargeRadish
Aug 19, 2011

WolfDad is Best Dad.



Twerk from Home posted:

Street price on the R5 1400 is $135 right now, is hyper-threading worth $5?

Yeah, definitely.

Combat Pretzel
Jun 23, 2004

No, seriously... what kurds?!

Krailor posted:

The quad vs octo channel memory between Threadripper and Epyc is a direct result there being half as many active dies on the chip. Each Ryzen die has 2 memory controllers resulting in dual channel memory access. If you have 2 active dies on a chip (Threadripper) each die can access two banks of memory simultaneously giving it 'Quad-Channel' memory access. Doubled again for Epyc to get 'Octo-Channel' memory access. A single core can't actually access more than two banks of ram at a time.
I guess I was unclear.

Four dies. Die 1 and 2 are connected to memory. Say die 1 is rendered unfunctional, due to binning or what, so for instance die 3 does computing, but die 1 still does memory duty, since die 3 isn't hooked up to the right pins. Worst case being die 1 and 2 doing memory, and die 3 and 4 doing computing. The alternative variant being each die doing a single channel of memory and only two dies of the four doing computing.

--tl,dr: The dies on the quad memory channel pinouts do memory IO, and computing is done by the best bins on the package and get memory access via IF, if not directly hooked up.

Combat Pretzel fucked around with this message at 17:50 on Jul 27, 2017

Cygni
Nov 12, 2005

raring to post



Combat Pretzel posted:



Four dies. Die 1 and 2 are connected to memory. Say die 1 is rendered unfunctional, due to binning or what, so for instance die 3 does computing, but die 1 still does memory duty, since die 3 isn't hooked up to the right pins. Worst case being die 1 and 2 doing memory, and die 3 and 4 doing computing. The alternative variant being each die doing a single channel of memory and only two dies of the four doing computing.

Based on ServeTheHome's Epyc tests, having to hit IF for the memory calls would have a ~40% hit on latency and ~50% hit on bandwidth vs each cores integrated controller. X399 only supports 4 channel memory, so only 2 of the dies can have active memory controllers. The 2 extra cores on the engineering sample really are completely vestigial.

Im putting my chips in the "they just reused Epycs as engineering samples and production models will only have 2 dies" pile, because its the only thing that makes sense to me. At least weve got somethin weird to talk about!

GRINDCORE MEGGIDO
Feb 28, 1985




Risky Bisquick posted:

There is no way AMD are plopping 2 extra dies onto a TR package. Barring some sort of strange technical limitation where they need to use 4 interconnects this is pretty ridiculous of a rumour.

It doesn't make sense not to test the chips before they are packaged, too. They aren't going to be just sticking them on a package and testing it afterwards and having enough post - packaging failures to make it worthwhile. IMHO.

3peat
May 6, 2010



AMD is now selling the Wraith Max xXxRGBxXx cooler.... for 60 dollars http://www.tomshardware.com/news/am...oler,35098.html

Watermelon Daiquiri
Jul 10, 2010




Considering they test chips before they leave the wafer...

HalloKitty
Sep 30, 2005

Adjust the bass and let the Alpine blast


Watermelon Daiquiri posted:

Considering they test chips before they leave the wafer...

See, that's what I assumed, so none of this makes sense, unless they're failed dies put in place just to keep the heatspreader level or something weird.

Methylethylaldehyde
Oct 23, 2004

BAKA BAKA


HalloKitty posted:

See, that's what I assumed, so none of this makes sense, unless they're failed dies put in place just to keep the heatspreader level or something weird.

Single socket style, single interposer/package style, single soldering setup to work the kinks out of, all for the cost of whatever wasted/dead dies. From a process simplification standpoint, it could work.

Mr Shiny Pants
Nov 12, 2012


Maybe they just wanted to get something out and an Epyc chip was at hand? Flashed the microcode to a threadripper and there you are.

ArgumentatumE.C.T.
Nov 5, 2016

by Jeffrey of YOSPOS


Maybe they'll start a 32 core Threadripper line if all the 16s sell.

SwissArmyDruid
Feb 14, 2014



Has AMD said if TR cores are disabled symmetrically across all cores a la R5, or are we looking at potentially goofy-rear end core counts like 1/7/6/2?

eames
May 9, 2009



ArgumentatumE.C.T. posted:

Maybe they'll start a 32 core Threadripper line if all the 16s sell.

maybe not 32 core but I'm sure we'll see 20/24 core Threadrippers with 7nm

MaxxBot
Oct 6, 2003

you could have clapped

you should have clapped!!


Ryzen Threadripper II, 64C/128T

PerrineClostermann
Dec 15, 2012

by FactsAreUseless


MaxxBot posted:

Ryzen Threadripper II, 64C/128T



repiv
Aug 13, 2009



HalloKitty posted:

See, that's what I assumed, so none of this makes sense, unless they're failed dies put in place just to keep the heatspreader level or something weird.

It could be like the RX480, where they shipped all early units with 8GB of VRAM and just artificially capped some units at 4GB. Later on they started making actual 4GB cards.

Similarly they might phase out the quad-die Threadripper and introduce a native two-die design later.

inkwell
Dec 9, 2005


https://www.youtube.com/watch?v=Ni9MdFNQCCE

lmfao. brb going to see threadripper at lollapalooza

Combat Pretzel
Jun 23, 2004

No, seriously... what kurds?!

"PCIe" is a registered trademark? wat?

Subjunctive
Sep 12, 2006

careful now


Combat Pretzel posted:

"PCIe" is a registered trademark? wat?

http://pcisig.com/sites/default/fil...ated_112206.pdf

Pretty much all standards are trademarked. That's how they control compatibility claims.

fishmech
Jul 16, 2006

by VideoGames


Salad Prong

Combat Pretzel posted:

"PCIe" is a registered trademark? wat?

Probably the last major one that wasn't is ISA, and then only because that was a retroactive name applied after it'd already been in use for several years with no real name besides "the expansion bus IBM PCs use"

Macksy
Oct 20, 2008


MaxxBot posted:

Ryzen Threadripper II, 64C/128T



Nice photoshop of the original xbox cpu!

Anime Schoolgirl
Nov 28, 2002






NewFatMike posted:

Bristol Ridge has been out with OEMs for like a year now though, hasn't it?
if the UK preorder pricing is anything to go by, the APUs will be AMD's 50-100 dollar range for the meanwhile

https://www.scan.co.uk/shop/compute...-apu-processors

NewFatMike
Jun 11, 2015



inkwell posted:

https://www.youtube.com/watch?v=Ni9MdFNQCCE

lmfao. brb going to see threadripper at lollapalooza



Hell yeah CPU division.

Anime Schoolgirl posted:

if the UK preorder pricing is anything to go by, the APUs will be AMD's 50-100 dollar range for the meanwhile

https://www.scan.co.uk/shop/compute...-apu-processors

Oh nice. That A12 isn't bad at 720p IIRC.

Give me Raven Ridge already! I want to see the laptops! I want to make my SFF gaming boxes!

wargames
Mar 16, 2008

official yospos cat censor


NewFatMike posted:



Hell yeah CPU division.


Oh nice. That A12 isn't bad at 720p IIRC.

Give me Raven Ridge already! I want to see the laptops! I want to make my SFF gaming boxes!

my laptop uses an A8 and works pretty drat well at 720p.

NewFatMike
Jun 11, 2015



Freesync 2 monitors on laptops is going to dramatically improve laptop gaming too, I think.

Good golly Ms. Molly, please give me a beefy APU, Freesync 2, IPS, 3:4, and an active stylus. I need it*

Inasmuch as anyone needs a laptop when they have a beefy desktop.

ufarn
May 30, 2009


.

repiv
Aug 13, 2009



http://www.pcworld.com/article/3211...ious-chips.html

PC World posted:

So did AMD really waste two perfectly good "Zeppelin" dies? Nope.

Those other two "chips" are nothing more than spacers to help maintain the structural integrity of the gigantic heat spreader, PCWorld has learned from a source who declined to be identified.

GRINDCORE MEGGIDO
Feb 28, 1985





That's good. Intel would have just used more glue.

inkwell
Dec 9, 2005



Wierd, but that kinda makes sense, if the rest of the threadripper package is based off of Epyc. Remember that shot of the IF traces between the die (and how two of the dies were mirrored)? Makes sense if they replaced the mirrored dies with blanks; thered just be the one diagonal IF bridge between top left and bottom right instead of whatever multi layer thing they were doing to get single hops between all four dies on epyc. Im totally talking out of my rear end here; but assuming it was using the same PCB as epyc would that mean that tr4 has a ton of unused lands/pins?

Also i gotta wonder what that means for cooler design if you've got hotspots on opposite corners of the ihs rather than near the center...

repiv
Aug 13, 2009



what if the spacers are completely defective zeppelin dies

100% yields

Anime Schoolgirl
Nov 28, 2002






I imagine using only one socket makes it simpler for the board designers and the unfortunate people who have to design trace paths for this poo poo.

Plus it's a great way to use completely wasted dies that aren't even good for Ryzen 3s.

isndl
May 2, 2012
I WON A CONTEST IN TG AND ALL I GOT WAS THIS CUSTOM TITLE

Anime Schoolgirl posted:

I imagine using only one socket makes it simpler for the board designers and the unfortunate people who have to design trace paths for this poo poo.

Unfortunate? You know Asrock engineers live for this poo poo, they would have gone insane years ago otherwise.

Anime Schoolgirl
Nov 28, 2002






isndl posted:

Unfortunate? You know Asrock engineers live for this poo poo, they would have gone insane years ago otherwise.
They're in an entire different league, what with the micro-stx project that isn't going to work with about 86% of MXM cards because there was no standard for that

EmpyreanFlux
Mar 1, 2013

The AUDACITY! The IMPUDENCE! The unabated NERVE!


If you gave Asrock engineers an infinite amount of money you'd have an endless cascade of weird poo poo like 5-6 people would buy. Like Imagine ASRock partnering with Nvidia to bring out a socketed Tegra line of chips.

Anime Schoolgirl posted:

They're in an entire different league, what with the micro-stx project that isn't going to work with about 86% of MXM cards because there was no standard for that

Knowing Asrock, they'll standardize MXM, and then proceed to become an AIB for MXM GPUs. They'll then buy up as much old stock of MXM cards as possible and then transfer all the chips to the new standardized PCB.

But seriously I'll buy their STX form factor if it comes in AM4 though because that shits perfect for AMD APUs. I have no interest when it's just Intel GPUs though because while it's a lot of CPU horse power the GPU is laughably bad.

Seamonster
Apr 30, 2007

IMMER SIEGREICH


NewFatMike posted:

Raven Ridge allegedly spotted today, as well.

https://videocardz.com/71232/amd-ry...raphics-spotted

1080p30 medium (with Freesync) I'll take for sure, especially if the U indicates a low voltage SKU. Looks like almost double the Iris Pro 640 performance?

http://www.ashesofthesingularity.co...52-70bcb8d3bca6



Whoa, back up a sec here. Is this with just straight system RAM? All Iris Pro parts have eRAM built in-die, right?

Adbot
ADBOT LOVES YOU

NewFatMike
Jun 11, 2015



I can't discern anything besides 4C/8T and Vega GPU architecture, but I'm glad somebody else got a little hype over it. I'm too thirsty for Raven Ridge news for that little story to be ignored.

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