The Ryzen "ThreadRipper"... 16 cores of awesome

[maddie]

That was the point I was making; the only way it would work would be if the controller was on-die with the eDRAM and communicated with the CPUs over Infinity Fabric, in which case it'd hardly be any better than going to main memory. A proper controller wouldn't add too much die space, but obviously it's something that would have to be added in Zen's successors.

(Of course, in an ideal world AMD would fill the space with a nice big HBM2 cache, but even if the necessary controller logic was there I shudder to think how much the required interposer would cost)

I'll keep posting this link as it appears too many of us are misinformed.

Shall we say roughly $3.

A 200mm^2 APU die + a 100mm^2 HBM2 stack.

http://electroiq.com/blog/2012/12/lifting-the-veil-on-silicon-interposer-pricing/
"Sesh Ramaswami, managing director at Applied Materials, showed a cost analysis which resulted in 300mm interposer wafer costs of $500-$650 / wafer. ----------- Since one can produce ~286 200mm2 die on a 300mm wafer, at $575 (his midpoint cost) per wafer, this results in a $2 200mm2 silicon interposer."

 

[Phynaz]

Really, where is yield reported. I searched. This does seem to be quite closely guarded secret.

Anybody that claims to know manufacturing yield is pulling a number out of their butt.

 

[Malogeek]

Anybody that claims to know manufacturing yield is pulling a number out of their butt.

The Guru3D article begins with "Good news for and from AMD". So we should probably take it as gospel?

 

[PeterScott]

The Guru3D article begins with "Good news for and from AMD". So we should probably take it as gospel?

No. Click the source link for that "Good News" on the Guru3D it, it's a clearly marked rumor from another site.

 

[Malogeek]

No. Click the source link for that "Good News" on the Guru3D it, it's a clearly marked rumor from another site.

Thanks. I assume one of the sites that love to make "articles" out of reddit rumor posts and forum musings. We know who they are.

 

[maddie]

Anybody that claims to know manufacturing yield is pulling a number out of their butt.

The point that started this was an estimate of $20-$30 for a Ryzen die. Raw cost to AMD.

A VERY broad range. $25 (+ or - 20%)

It's very reasonable to assume that unless yields are in the toilet, an intelligent thinking person can make useful assumptions and calculate a range. Your attempt to equating that not knowing exactly what the yield is right now and inability to make any assumptions is quite ridiculous. I hope you surely know better.

 

[PeterScott]

The point that started this was an estimate of $20-$30 for a Ryzen die. Raw cost to AMD.

A VERY broad range. $25 (+ or - 20%)

It's very reasonable to assume that unless yields are in the toilet, an intelligent thinking person can make useful assumptions and calculate a range. Your attempt to equating that not knowing exactly what the yield is right now and inability to make any assumptions is quite ridiculous. I hope you surely know better.

And the original claim was going on about the super high markup on chips. You know if it is as rosy as some suggest, why have both AMD and GloFo consistently lost money for years on end?

This is based on guesses, stacked on guesses, stacked on more guesses, with a few numbers pulled from the air.

If we believe this, Intel must be building Skylakes (122mm2) on their super mature 14nm process, for about $10, and Intel gets to keep all the margin too, since it all one company. Their gross margins should be 90+%.

 

[Topweasel]

It's not schoedengers wafer. We can get a good feel for the cost per chip by knowing rough wafer costs. Defect rate redundancy and a rough yield rate, since we know the chips die size. It's not like it as an equal chance to be $30 and be $300 until we get the exact numbers.

This really detracts from the conversation to use small inconsistencies and variables to distract people away from the more important topic.

 

[maddie]

And the original claim was going on about the super high markup on chips. You know if it is as rosy as some suggest, why have both AMD and GloFo consistently lost money for years on end?

This is based on guesses, stacked on guesses, stacked on more guesses, with a few numbers pulled from the air.

If we believe this, Intel must be building Skylakes (122mm2) on their super mature 14nm process, for about $10, and Intel gets to keep all the margin too, since it all one company. Their gross margins should be 90+%.

What are you going on about?

I think I addressed your attempts to ridicule the estimated $20-$30 Ryzen die cost. How is anything you posted here relevant?

It's not schoedengers wafer. We can get a good feel for the cost per chip by knowing rough wafer costs. Defect rate redundancy and a rough yield rate, since we know the chips die size. It's not like it as an equal chance to be $30 and be $300 until we get the exact numbers.

This really detracts from the conversation to use small inconsistencies and variables to distract people away from the more important topic.

Exactly.

How is the cat?

 

[jpiniero]

I def think it's a decent amount more than 20-30 if only because AMD is likely getting a raw deal from GloFo. It'd certainly be a lot of money going to waste if they really are putting two otherwise good chips (even assuming partially defective) on TR.

 

[Topweasel]

I def think it's a decent amount more than 20-30 if only because AMD is likely getting a raw deal from GloFo. It'd certainly be a lot of money going to waste if they really are putting two otherwise good chips (even assuming partially defective) on TR.

There was a link pointing to AMD referring to them as spacers. So its possible they might not even be real CPU dies.

 

[Topweasel]

Exactly.

How is the cat?

If it knows whats good for it, dead.

 

[GoNavy1776]

Lol the cat paradox lol... I remember tha from College Gen Chem.

 

[csbin]

HOW-TO Install AMD X399 CPU (MSI)

 

[GoNavy1776]

HOW-TO Install AMD X399 CPU (MSI)

Was this video made by a tard? I understand if you had parkinsons disease you would apply that much goop but wow they must have hired a complete imbecile to install that cpu thermal grease.

 

[Ajay]

HOW-TO Install AMD X399 CPU (MSI)

Worst application of thermal paste evah! Anyway, just twiddling my thumbs wondering when TR will actually be release (or at least reviewed).
August 10th now?

 

[GoNavy1776]

Worst application of thermal paste evah! Anyway, just twiddling my thumbs wondering when TR will actually be release (or at least reviewed).
August 10th now?

August 10th is just the date AMD anounces we have to wait until Oct for anothr announcement to wait until January.

 

[ddogg]

Not sure what was wrong with that thermal paste application? I mean it's a monstrous CPU. Either way can't wait to see what this is capable off

 

[Ajay]

Not sure what was wrong with that thermal paste application? I mean it's a monstrous CPU. Either way can't wait to see what this is capable off

Too much paste and sloppy as all heck. An yes, really interesting an seeing how these perform and what sort of systems ppl build around TR.

 

[GoNavy1776]

Too much paste and sloppy as all heck. An yes, really interesting an seeing how these perform and what sort of systems ppl build around TR.

Yes except that ws a socket AM1 i the video.

It was holy inaccurate. Socket tr4 id absolutely huge. 5 or 6 times the size of that am1.

 

[ddogg]

Oh that video is wrong, I saw the real video on reddit. The real video is here https://youtu.be/Z1j3S6MKB_E
I figured you would need that much thermal paste as the CPU is massive.

 

[Yakk]

Ah yes, much better, maybe even a little anemic, but looks pretty good. Except I'll probably do a visual test on my first TR and see how the paste spread. Just to make sure it spread properly everywhere as it is really big heat spreader with multiple hot spots.

 

[Topweasel]

Big fan of them using almost a cartridge system for the CPU. Have never been a fan of setting the CPU directly on the pins personally.

 

[ddogg]

Ah yes, much better, maybe even a little anemic, but looks pretty good. Except I'll probably do a visual test on my first TR and see how the paste spread. Just to make sure it spread properly everywhere as it is really big heat spreader with multiple hot spots.

Same. I will definitely be doing a trial and error method to make sure the paste has spread across the heat spreader. Apply too little and the CPU will likely be too hot.

 

[GoNavy1776]

Same. I will definitely be doing a trial and error method to make sure the paste has spread across the heat spreader. Apply too little and the CPU will likely be too hot.

According ro Dr8Baeur or however you spell his gaming name... the dies are not that large. Only the heatspreader and PCB has to be so damn large to support all the external connections to the dies. He was stating that essentially a cooler would only have to cover the center portion of the chip to effectively cool it.

Yes this is a big chip. No it is not any different that cooling any other past Intel monstrosity like socket 2011. My 3930K would pull over 350-400 watts easy when overclocked. High wattage CPU is nothing new. This is just a physically large chip. Plus this thing burns less overall power at stock settings than an I9 Intel 10 core does at stock speeds. Don't get bent out of shape about having to spread the thermal goop across every ^2 mm of the chip until we know for sure how the thing needs to be slathered anyways.