AMD Threadripper Was Developed by Engineers in Their Spare Time

Megalith

Megalith

24-bit/48kHz
Staff member
Joined
Aug 20, 2006
Messages
13,000
According to interviews with AMD staff working on the Ryzen project, the high-end Threadripper line was the result of a “skunkworks” project: the CPUs weren't originally part of the company's chip strategy but a passion project that the engineers worked on in their spare time. Luckily, AMD Senior Vice President and General Manager Jim Anderson got wind of the project and threw his weight behind it after being “blown away” by the specifications and how it could actually be built.

Threadripper never had a business plan: "We were building it because we knew it was awesome, because we could and to make it best product we could, even the name had to be big." In a bid to make the device in time, and at a reasonable cost, the engineers rejected the idea of making a larger die. "The efficiency in manufacturing came from the fact that we didn't need entire wafer runs just to produce Threadrippers, we could use Ryzens… It meant we didn't need millions of dollars for this one design; we found a way to use our existing great product and make it even more powerful," said Prior.
 
shad0w4life said:
Aren't they made with %5 binning CPUs...
Click to expand...

Eypc SKUs right now cap out at 2.2/3.2GHz. You pack two more dies in there and you have a lot bigger thermal concerns, so you can't clock any of the dies up as far. That would still leave plenty of room for TR to take the top 5% to clock upo to 4.0.

And we don't know what yields are. Maybe they are only getting 5% yield on a wafer... doubtful, but possible.
 
Elios said:
or you know a way to make money on partly failed Epic chips
Click to expand...
We know that there are two spacers instead of disabled dies in the TR package. Purportedly Ryzen has 98% plus yield, at least marginal chips can be sold as 4 core 8thread. TR is top binned dies packaged in twos. Doubt there is many failures in EPYC packaging.
 
I, for one, am glad they took the risk. I love my CoreJizzer and while it may not have the IPC of an Intel chip it is still a tremendous value in the creative space. I have been a staunch Intel user for a VERY long time thanks to AMDs near apathy in the CPU market, but I am always a fan of the underdog as competition is generally a win all-around for the consumer. Can't wait to see what the next version of the Zen platform holds.
 
Unter Dog said:
I've tried things like this at my company and got the opposite reaction from management. Kudos AMD
Click to expand...
Same here. I feel like my bosses are sometimes Bill Lumbergh from Office Space or the "Did you get the memo guy". They let you talk. Then "Umm yeah, did you get the memo?" They usually don't give you the time of day. My favorite is when we then have a meeting about innovating and thinking outside the box. Glad to see this worked for the AMD engineers.
 
Unter Dog said:
I've tried things like this at my company and got the opposite reaction from management. Kudos AMD
Click to expand...

I wrote up a White-Paper a year ago on efficiency and how the creation of a new dedicated department would actually save our building money short AND long term.... Ops went to senior management with it.. Ops got promoted and I'm still sitting here with management asking me retard questions.... because they have absolutely no idea what I do or how I do it, they just like my numbers *sigh*
 
Wonder if these people got a sizable bonus for doing this, fuck saying thanks i want to see digits in my bank balance.
 
Good for them. And yeah like others said, gives them a viable revenue stream from "failed" Epyc chips. Turns trash into profit. Reminds me of the 3 core Phenoms and Athlons that were destined for the trash and someone said "Hey wait, we can sell those!"
 
Burticus said:
Good for them. And yeah like others said, gives them a viable revenue stream from "failed" Epyc chips. Turns trash into profit. Reminds me of the 3 core Phenoms and Athlons that were destined for the trash and someone said "Hey wait, we can sell those!"
Click to expand...

Only the 1920 and 1900 would have 'failed' chips. The 1950 is two full zen die (just like the r7 is one full zen and 32 core epyc server chips are 4 die). They use the 'failed' ships for the lower end, but the r7, TR 1950 and epyc 32 core chips all have fully enabled cores. They just have 1, 2 and 4 of them, respectively. They already had revenue streams for the failed die in the r3, r5 and <32 core epyc chips. This just fills a (in retrospect) glaring hole between the r7 and epyc.
 
Chris_B said:
Wonder if these people got a sizable bonus for doing this, fuck saying thanks i want to see digits in my bank balance.
Click to expand...

Research shows that when you put people in charge of a project and make it "their own" they will work harder for less money, because it is "their's" and not "somebody else's" A sense of autonomy pays off. Just beware of "man in the box" developers. Their goals need to be clearly stated before hand and progress monitored. But they still get to be in control of development.

I'm in charge of two projects here. One is AI that is a pet project where machines assemble themselves from a parts catalog. The second is a hyper aggressive optimizer designed to reduce part sizes. The later project is currently saving our customers over $1,000,000 a year. I put a lot of time of effort into these things that was outside my contract 40 hour week. And I pushed it because I hated the legacy code and how it was written.

Do I get paid more? No. But successful implementation ensures I have a say on how things are done in the future which makes my life a lot easier.
 
Last edited by a moderator:
The story is a bit misleading. These guys didn't make chips and a platform on their own time.

The engineers worked on the CONCEPT in their spare time. Once the VP saw the concept and how it fit in the product line, then they got all their funding and such to develop hardware.
 
Unter Dog said:
I've tried things like this at my company and got the opposite reaction from management. Kudos AMD
Click to expand...

I think these grassroots projects are possible when they are this small and behind the eightball. Unfortunately for AMD (or maybe fortunately for us) they did this before with Mantle. A handful of key players got together and talked it over. Then one guy in the basement coded the basic layout and then they ran with it and the rest is history.
 
Bandalo said:
The story is a bit misleading. These guys didn't make chips and a platform on their own time.

The engineers worked on the CONCEPT in their spare time. Once the VP saw the concept and how it fit in the product line, then they got all their funding and such to develop hardware.
Click to expand...
It seemed pretty clear:

"The efficiency in manufacturing came from the fact that we didn't need entire wafer runs just to produce Threadrippers, we could use Ryzens… It meant we didn't need millions of dollars for this one design; we found a way to use our existing great product and make it even more powerful,"

"So we put together this skunkworks team where we had platform architects, people that deal with core design, business unit, marketing team, to work out how to use what's already here and to go to the boss - Jim Anderson and say we'd like to do this"
 
Meeho said:
It seemed pretty clear:

"The efficiency in manufacturing came from the fact that we didn't need entire wafer runs just to produce Threadrippers, we could use Ryzens… It meant we didn't need millions of dollars for this one design; we found a way to use our existing great product and make it even more powerful,"

"So we put together this skunkworks team where we had platform architects, people that deal with core design, business unit, marketing team, to work out how to use what's already here and to go to the boss - Jim Anderson and say we'd like to do this"
Click to expand...

They put together an IDEA for TR in their own time. That's very different from the headline "TR was developed by engineers in their spare time"
 
Bandalo said:
They put together an IDEA for TR in their own time. That's very different from the headline "TR was developed by engineers in their spare time"
Click to expand...
You aren't coming up with an idea for a year, you're developing the product on a high level. The title is ok.
 
I believe Sony used to have a lot of success with this in the past. Amazing what happens when you let genius minds operate unencumbered. Now we just need RTG to do the same thing!
 
Sounds to me they have some good engineers on the CPU team now and are interested in bringing the enthusiasm back to the PC market. We needed this for years and here it is. Kudos to AMD. If RAM wasn't so expensive I would be getting a 12 core TR myself.
 
Bandalo said:
They put together an IDEA for TR in their own time. That's very different from the headline "TR was developed by engineers in their spare time"
Click to expand...

I'm sure they didn't just breakout a box of crayons and draw pictures of 2 ryzen chips living in a house happy with mommy and daddy.

These are engineers... they detailed how it would work and what they would have to overcome design wise.

Did they build a full featured product in their spare time ? Of course not... it sounds to me like they basically wrote the threadripper whitepaper. That is a huge chunk of the planning work. From their they required real time and money to translate the concept into a working sample.
 
ChadD said:
I'm sure they didn't just breakout a box of crayons and draw pictures of 2 ryzen chips living in a house happy with mommy and daddy.

These are engineers... they detailed how it would work and what they would have to overcome design wise.

Did they build a full featured product in their spare time ? Of course not... it sounds to me like they basically wrote the threadripper whitepaper. That is a huge chunk of the planning work. From their they required real time and money to translate the concept into a working sample.
Click to expand...

No, I wasn't implying they did nothing or were incompetent.

But the headline implies they had the whole thing planned out. Reading the article, they put together an outline of an idea. Then when they got funding, they developed a full product. My point is that you can't "develop" something as complex as TR in your spare time. You can develop the idea to slap a couple Ryzen dies under a single IHS, but you can't actually "develop" anything that complicated without a lot of funding and time.
 
Bandalo said:
No, I wasn't implying they did nothing or were incompetent.

But the headline implies they had the whole thing planned out. Reading the article, they put together an outline of an idea. Then when they got funding, they developed a full product. My point is that you can't "develop" something as complex as TR in your spare time. You can develop the idea to slap a couple Ryzen dies under a single IHS, but you can't actually "develop" anything that complicated without a lot of funding and time.
Click to expand...

They are saying what they did. Really, who are you to counter them?
 
da599235a4683ff92a103a201556a479ef10f3c4aed75141c2b8eac5e1e17c7e.jpg
 
You must log in or register to reply here.
Back
Top