Threadripper: The inside story

[mr moose]

1 minute ago, Drak3 said:

Given interactions with some other members in the past, generic circles aren't that bad.

could be worse I suppose.  I just found this:

 

http://wccftech.com/intel-skylake-x-kaby-lake-x-q2-2017-roadmap-leak/

 

They reckon they knew about x299 having 18c June last year.   Anyway that's WCCF so I wouldn't be surprised if they edited that graph in more recently.

[leadeater]

11 hours ago, Dylanc1500 said:

18 cores was done because of the different silicon dies (10,18, 28). I don't know what you mean by 12 core, if you could explain that a little better, I apologize.

 

like I had said it was released earlier than planned, they had plans to release but not when they  ended up. I can't really go to in depth with that.

The jump in core count is totally unprecedented in the HEDT line, something like that has never been done before. All other times it's been an addition of 2 cores. Along with the mentioned previous road maps never showing these parts before.

 

Intel may have added them a short while before their announcement internally but that still would have been in response to Threadripper. This doesn't really mean it's good or bad, there is nothing wrong with responding to competition.

 

These HEDT CPUs now, and have for I think the last 2 generations, used their own dies and are not cut down Xeon dies. Being based on the design and literally being from the same wafer is very different. Intel having large core count Xeons doesn't mean Intel had or wanted to have the capability to produce high core count HEDT CPUs.

 

Without Threadripper there isn't really any indication the 18 core HEDT would have existed and I find it very unlikely based on past history with this product line. It also puts a great deal of pressure on their Xeon lineup which is why I also think Xeon-W was unplanned also, story for a different thread, but was added so partners like Dell/HPE could build systems to customer demands but not be forced in to using Skylake-X to meet that. Xeon-W also is not a replacement for Xeon E3, E5-1600 maybe but has 1/3 the memory capacity support.

[Jito463]

3 hours ago, mr moose said:

Do some of you actually think Intel are capable of puling an extra 4 -8 cores out of their **** in such a short time with no planning or R+D?  Of course they were working on it prior to TR, given the x299 release debacle it's fairly obvious they rushed it out the door, but to assume it didn't exist (in planning or otherwise) prior to TR is a bit of stretch.

 

TL:DR: is anyone really surprised that professionals managed to do their job and design a processor worth selling?  It's really kinda insulting to call engineers "enthusiasts" or to insinuate AMD got TR out of employee luck. 

As @Drak3 pointed out, nobody is saying that Intel wasn't already working on those 18-core processors, we just don't accept that they were planning to release them for X299.  There's no evidence to support that supposition, but plenty of circumstantial evidence to support the conclusion that X299 was never intended to go higher than 12-core.

[leadeater]

3 hours ago, mr moose said:

Do some of you actually think Intel are capable of puling an extra 4 -8 cores out of their arse in such a short time with no planning or R+D?  Of course they were working on it prior to TR, given the x299 release debacle it's fairly obvious they rushed it out the door, but to assume it didn't exist (in planning or otherwise) prior to TR is a bit of stretch.

Actually I do think Intel has both the highly skilled engineers and equipment to do exactly this. I'm not saying they could add that many cores to their existing highest core count die but certainly can to a lesser die. Intel has the means and ability to execute something like this based on pre-existing architecture with a 6 month turnaround if they so wish, this is even more so the case with the move to the scalable mesh architecture.

 

3 hours ago, mr moose said:

You can only enable, disable and validate something that already exists. I think this forum just gets too hung up on basic ideals.  If TR can be designed from existing R+D and made into a usable product then so can x299.

 

What AMD did with Threadripper and EYPC is a fair amount more simple than what Intel has to do as well but still fits in the with MCM Zen design purpose that they laid out from the start. It's not unbelievable that AMD could have a very short turnaround on the Threadripper design since it is very much EYPC with less active dies, that is not a hard thing to do architecture permitting.

 

Also I'm not implying either AMD or Intel knee jerk'd anything, you can have basic ideas or concepts without any intent to execute them unless certain trigger points are meet i.e. Threaripper's existence.

[mr moose]

46 minutes ago, leadeater said:

Actually I do think Intel has both the highly skilled engineers and equipment to do exactly this. I'm not saying they could add that many cores to their existing highest core count die but certainly can to a lesser die. Intel has the means and ability to execute something like this based on pre-existing architecture with a 6 month turnaround if they so wish, this is even more so the case with the move to the scalable mesh architecture.

 

 

What AMD did with Threadripper and EYPC is a fair amount more simple than what Intel has to do as well but still fits in the with MCM Zen design purpose that they laid out from the start. It's not unbelievable that AMD could have a very short turnaround on the Threadripper design since it is very much EYPC with less active dies, that is not a hard thing to do architecture permitting.

 

Also I'm not implying either AMD or Intel knee jerk'd anything, you can have basic ideas or concepts without any intent to execute them unless certain trigger points are meet i.e. Threaripper's existence.

I'm actually not saying it's unbelievable what AMD did with threadripper,  but I am saying the concept that it was just enthusiasts on AMD's part and knee jerk on Intel's that I find offensive (to intel and AMD engineers).  They are very experienced engineers on both sides and I don't think some of the insinuations are justified.

 

1 hour ago, Jito463 said:

As @Drak3 pointed out, nobody is saying that Intel wasn't already working on those 18-core processors, we just don't accept that they were planning to release them for X299.  There's no evidence to support that supposition, but plenty of circumstantial evidence to support the conclusion that X299 was never intended to go higher than 12-core.

 

What do you make of the article I posted earlier claiming 18cores on x299 in June last year?

[TheOriginalHero]

13 hours ago, mynameisjuan said:

Depends on what? Intel is still king in the performance category, but yeah its going to cost you. 

Actually considering the server level is the highest level of performance for both companies and amd's top 32c is on par with intel's 28c for $4000 less I wouldnt say intel holds the perfomance crown.

[ATFink]

6 hours ago, RGProductions said:

No that's wrong. That's like saying a G4400 has better price/perf than the $2000 CPUs. This is a lot more along the lines of buying a miata instead of an ecoboost mustang. You arent losing much, and its a lot cheaper.

The EcoBoost mustang starts at about $26,000 with a zero to 60 time of 5.5 seconds.

https://www.ford.com/cars/mustang/2017/models/ecoboost-fastback/

http://www.cjponyparts.com/resources/mustang-v6-vs-ecoboost

 

The latest Mazda Miata starts at about $25,000 with a zero to 60 time of 7.0 seconds (only one entry for the sport variant which is the cheaper option... second link below this paragraph).

https://www.mazdausa.com/vehicles/mx-5-miata

https://www.0-60specs.com/mazda-mx-5-miata-0-60-times/

 

The Miata isn't much cheaper at all and it's losing by a lot (in a straight line).

 

Sorry for being a stickler, I'm just a fan of the EcoBoost Mustang . I agree with the sentiment of your response.

[leadeater]

1 hour ago, mr moose said:

I'm actually not saying it's unbelievable what AMD did with threadripper,  but I am saying the concept that it was just enthusiasts on AMD's part and knee jerk on Intel's that I find offensive (to intel and AMD engineers).  They are very experienced engineers on both sides and I don't think some of the insinuations are justified.

Well I have to say I haven't read much of this thread and I tend to skip straight to certain posters to see what they are saying since they typically have something worth while to read e.g you .

[mr moose]

1 minute ago, leadeater said:

Well I have to say I haven't read much of this thread and I tend to skip straight to certain posters to see what they are saying since they typically have something worth while to read e.g you .

cheers.

[Pixel5]

15 hours ago, emosun said:

wow. too bad they sat around and did nothing for several years prior to that. while intels "computer enthusiasts" got bored from waiting for someone to actually compete.

i dont really think intel is very concerned , amd has a history of not really staying on top of things , and a history of never really being at the cutting edge of anything. short of making cheap versions of cpus that already exist , amd doesn't do much better than that.

sounds very much like you have no idea why AMD is so far behind at this time.

 

If you would have any knowledge about intels shady business practice they used to drive AMD out of the market you would be cheering for them that they managed to pull that one of under the worst possible conditions possible.

Meanwhile Intel got away with paying some fines which are nothing compared to the damage they have done and the money they have raked in.

[laminutederire]

46 minutes ago, Pixel5 said:

sounds very much like you have no idea why AMD is so far behind at this time.

 

If you would have any knowledge about intels shady business practice they used to drive AMD out of the market you would be cheering for them that they managed to pull that one of under the worst possible conditions possible.

Meanwhile Intel got away with paying some fines which are nothing compared to the damage they have done and the money they have raked in.

Also sounds like he doesn't know the whole story overall. Like when AMD was stomping on Intel somewhere around the early 2000 if I remember right. (Sorry if I'm mistaken on the date, I was too young to follow those stuffs then (as was he I suppose) (my first computer was a laptop with a 3rd gen i5 ))

[mr moose]

1 hour ago, laminutederire said:

Also sounds like he doesn't know the whole story overall. Like when AMD was stomping on Intel somewhere around the early 2000 if I remember right. (Sorry if I'm mistaken on the date, I was too young to follow those stuffs then (as was he I suppose) (my first computer was a laptop with a 3rd gen i5 ))

2004-2005-2006.  The 3200 and 3500 Athlon's were released just after the Pentium 4's (northwood/prescott).  And gave them a hiding in gaming.  it was shortly after that run that Intel started doing dodgy stuff with the x86 compiler and anti-trust case with OEM's.

 

EDIT:for those who were still being potty trained at the time, it is interesting to note that Intel was generally better at video encoding and workstation/business applications.   But then the Pentium 4EE was $1000 and the A643500 was only $400.

 

 

 

 

 

 

[Bananasplit_00]

this is pretty funny but exactly how i thought it happaned

[samcool55]

9 hours ago, Jito463 said:

A bit of a backhanded compliment.

Just to be clear, i'm not saying the guys who came up with Threadripper are below average.

 

What i mean is that companies should also hire people that are very passionate about something that's relevant for the job they might get even if they are below the average employee.

Usually companies look at numbers and simple things and don't even try getting to know the person, even if that person has a lot of potential and possibly some great ideas which could lead to awesome things like threadripper.

 

The fact the guys and girls at the top of AMD heard of the idea, talked about it and figured out of it could be a worthy product is probably something that rarely happens in similar big companies and imo shows AMD takes better care of its employees than the average company.

[samcool55]

3 hours ago, ATFink said:

The EcoBoost mustang starts at about $26,000 with a zero to 60 time of 5.5 seconds.

https://www.ford.com/cars/mustang/2017/models/ecoboost-fastback/

http://www.cjponyparts.com/resources/mustang-v6-vs-ecoboost

 

The latest Mazda Miata starts at about $25,000 with a zero to 60 time of 7.0 seconds (only one entry for the sport variant which is the cheaper option... second link below this paragraph).

https://www.mazdausa.com/vehicles/mx-5-miata

https://www.0-60specs.com/mazda-mx-5-miata-0-60-times/

 

The Miata isn't much cheaper at all and it's losing by a lot (in a straight line).

 

Sorry for being a stickler, I'm just a fan of the EcoBoost Mustang . I agree with the sentiment of your response.

Well, that's because you look at US prices and the Mustang is built in the USA while the MX5 has to come from Japan. Import taxes, shipping costs, all those things add up to the MX5's high price but not to the Mustang.

 

If i would compare prices (i live in belgium) the cheapest Mustang would be 37 000 euro's and the cheapest MX5 would be 22 590 euro's.

Exactly the same thing, different location, completely different prices. You can't apply that kind of logic to hardware...

 

Of course if you would look at japanese prices the MX5 would be even cheaper and the Mustang would be expensive AF.

[Jito463]

7 hours ago, mr moose said:

What do you make of the article I posted earlier claiming 18cores on x299 in June last year?

I think it's WCCF, and I'd question them if they said the sky was blue and water was wet.  They have a known tendency to over exaggerate, if not just make stuff up completely out of thin air.  If you had posted from a more reputable source I might have conceded, but I take everything they post with a

 

[DocSwag]

20 hours ago, Dylanc1500 said:

If you want top tier enthusiast performance you have to pay top tier enthusiast price. As far as being under water that could be a lot of different things and isn't currently known. Being under water doesn't automatically mean better than air.

Even at that level it's rare for someone to just say, "I have absolutely no budget at all." There are people that have insane budgets, like $3000-$4000, but unlimited ones I've only seen one or two. Within a $3000-$4000 budget or even a $5000 one, I think it's fair to say a 1950x is usually gonna be a better option.

[RGProductions]

6 hours ago, ATFink said:

The EcoBoost mustang starts at about $26,000 with a zero to 60 time of 5.5 seconds.

https://www.ford.com/cars/mustang/2017/models/ecoboost-fastback/

http://www.cjponyparts.com/resources/mustang-v6-vs-ecoboost

 

The latest Mazda Miata starts at about $25,000 with a zero to 60 time of 7.0 seconds (only one entry for the sport variant which is the cheaper option... second link below this paragraph).

https://www.mazdausa.com/vehicles/mx-5-miata

https://www.0-60specs.com/mazda-mx-5-miata-0-60-times/

 

The Miata isn't much cheaper at all and it's losing by a lot (in a straight line).

 

Sorry for being a stickler, I'm just a fan of the EcoBoost Mustang . I agree with the sentiment of your response.

Should read the f150 can raptor/lightning comment. For some reason I remember the Ecoboost as $33,000

[Doobeedoo]

Pretty good story, passion and enthusiasm shows. 

[The Benjamins]

10 hours ago, mr moose said:

could be worse I suppose.  I just found this:

 

http://wccftech.com/intel-skylake-x-kaby-lake-x-q2-2017-roadmap-leak/

 

They reckon they knew about x299 having 18c June last year.   Anyway that's WCCF so I wouldn't be surprised if they edited that graph in more recently.

I think it is edited, I can't find a single comment on that page talking about anything more then 10c.

 

looking at the time frame of Videocardz skylake-x articles

https://videocardz.com/news/cpus/intel-skylake-x

April 9th - 10c (leak)

May 12th - 12c (leak)

May 29th - 18c

 

for AMD TR was announced May 16th at 16c

March rumor of 16c AMD cpu

http://wccftech.com/amd-working-16-core-ryzen-cpu/

 

 

WCCF

 

Apr 9th - 10c (Leak)

http://wccftech.com/intel-basin-falls-x299-hedt-platform/ (chart updated)

Apr 21 - 12c (leak)

http://wccftech.com/intel-moster-12-core-hedt-skylake-x-confirmed-launch-30-may-x299/

May 12 - 12c (leak)

http://wccftech.com/intel-skylake-x-core-i9-7920x-7900x-7820x-7800x-x299-leaked/

May 29 - 18c This is the first time I can find any meantion of a leak, rumor, announcement of a 18c x299 CPU

http://wccftech.com/intel-core-x-skylake-x-kaby-lake-x-cpu-family-leak-core-i9-7980x-18-core-36-thread/

 

I can't anywhere of info of any kind of a 18c intel HEDT cpu before may 29

[GDRRiley]

12 hours ago, mr moose said:

Do some of you actually think Intel are capable of puling an extra 4 -8 cores out of their arse in such a short time with no planning or R+D?  Of course they were working on it prior to TR, given the x299 release debacle it's fairly obvious they rushed it out the door, but to assume it didn't exist (in planning or otherwise) prior to TR is a bit of stretch.

 

TL:DR: is anyone really surprised that professionals managed to do their job and design a processor worth selling?  It's really kinda insulting to call engineers "enthusiasts" or to insinuate AMD got TR out of employee luck. 

the had the die already made for xeon they just had to cut it down.

[sydh]

Neat read and also that time line for being a pet project of a few.

[mr moose]

6 hours ago, The Benjamins said:

snip

 

6 hours ago, GDRRiley said:

snip

 

7 hours ago, Jito463 said:

snip snip

(extra snip for big picture)

 

So we can agree they had been most likely working on 18C well prior to may this year.  But just as how far they intended to go on x299 is just supposition?

 

Which leads me back to what I was saying earlier,  The idea that Intel is knee jerking adding 8 cores (almost doubling) to an existing chip without R+D is highly unlikely.  Dropping price, bringing release earlier and a new marketing campaign are knee jerk reactions. But drastic changes to CPU engineering is not.

 

I only commented originally becasue  the insinuation that this has Intel falling apart or freaking out is just wrong and the insinuation that AMD has "enthusiasts" working on TR instead of it being simply the result of actual engineers doing their job is quite frankly a little insulting to AMD.  

 

 

[The Benjamins]

54 minutes ago, mr moose said:

 

 

 

So we can agree they had been most likely working on 18C well prior to may this year.  But just as how far they intended to go on x299 is just supposition?

 

Which leads me back to what I was saying earlier,  The idea that Intel is knee jerking adding 8 cores (almost doubling) to an existing chip without R+D is highly unlikely.  Dropping price, bringing release earlier and a new marketing campaign are knee jerk reactions. But drastic changes to CPU engineering is not.

 

I only commented originally becasue  the insinuation that this has Intel falling apart or freaking out is just wrong and the insinuation that AMD has "enthusiasts" working on TR instead of it being simply the result of actual engineers doing their job is quite frankly a little insulting to AMD.  

 

 

Can you explain why motherboard manufacturers didn't know about the 14c+ parts before it became public knowledge?

[mr moose]

5 minutes ago, The Benjamins said:

Can you explain why motherboard manufacturers didn't know about the 14c+ parts before it became public knowledge?

Never bothered looking into motherboards.  what evidence is there to suggest hey had no idea what was happening?