All down to yields. AMD still desire to convert as many Intel users (especially streamers) as possible.
-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
AMD Zen 2 (Ryzen 3000) - *** NO COMPETITOR HINTING ***
- Thread starter Jamin280672
- Start date
All down to yields. AMD still desire to convert as many Intel users (especially streamers) as possible.
Associate
- Joined
- 27 Apr 2007
- Posts
- 966
They will take the best chips for the 64C parts be they EPYC or TR.
I did specifically say that they will prioritise EPYC over TR hence why 64C for TR might not come for a while.
The jump from 16C AM4 Zen 2 to TR Zen 2 I think will be less appealing than from 8C Zen 1.
Soldato
- Joined
- 26 Oct 2013
- Posts
- 4,132
- Location
- Scotland
The chiplets are so small with likely an extremely high success and salvage rate. I can imagine there will be no shortage of chiplets at the high, mid or low end.
Caporegime
- Joined
- 8 Jan 2004
- Posts
- 32,584
- Location
- Rutland
"Best" isn't just one attribute. The chiplet design means yields are much better than monolithic for a high core count chip but none of us will have a clue what AMD's yields are like.
All we can say is the chiplet design will make 64 core chips far more likely. Epyc will definitely get it but 64 core TR is entirely possible and AMD will want to one up Intel in HEDT.
Just on that, its really interesting to see the language usually used to describe why someone should get Intel over AMD... reversed.
Associate
- Joined
- 2 Jan 2019
- Posts
- 617
This is the crux of the whole debate.
Chiplet design, especially at such small process nodes, is revolutionary. Yields should be phenomenal.
Its exciting because AMD struggle with any big upfront costs, and here we're expecting unprecedented demand; those costs are spread over such a large number of units that their cost per SKU is likely as low as it has ever been, and should continue to drop. These 3rd gen SKUs will likely have relevance for sales revenue even after their 4th gen comes out.
OK, I take it back, he's not trolling. He is being intentionally obtuse and hypocritical. End result is the same, he offers no valid contribution to the discussion.
Associate
- Joined
- 27 Apr 2007
- Posts
- 966
I think 32C TR Zen 2 will be far enough ahead of Intel that they don't need to introduce 64C in a hurry.
Also they could introduce it at a higher price point later
That's all I am saying because it seems a low priority halo product.
Didn't AMD say it took them a long time to build up a supply of chips for TR 32C?
To bin chips which can hit the higher frequency that TR demands plus keeping power consumption in hand demands a lot I think. Was it the top 1%?
I hope they release it of course, not that I am remotely in the market for one.
no im just asking for proof of what people are stating as almost fact with no proof. i want as much as you for these chips to be as good as some tout. i will be one of the first to buy one if they are. the problem with the internet is many make up or percieve people and judge. dont do that. assumption just makes you look silly also its wrong. at the end of the day we want great cpus. i just havent seen the proof to show me these are just another amd marketing trick that dont compete with what i can already own buy. you have to remember amd havent really been competitive for 15 years or more with intel. so hopefully these chips are the ones to go infront ( which i doubt) thats just my opinion.
as to offering no contribution to the discussion why is that ? because i dont agree with people saying nonsence ? or because my opinion isnt the same as the amd crowd ? im not on either side intel or amd. im just looking at what we know . people here are just saying theorectical things as if they know. twice as fast , no proof. faster mid range chips than a 9900k with no proof. just a cinebench shot. so what i do contribute to the discussion is common sense. logic. not make believe.
how can guessing or giving out false details be adding to a topic either ? how is that valid ?
what should be the case in these threads is just lock them down until actual details are given. so then you can debate with actual facts. nothing wrong then. people can debate with real idea. not hope and dreams. worse thing is some of this guess work or bs can lead people to buy expensive items which that are actually worse than what they could own.
Associate
- Joined
- 2 Jan 2019
- Posts
- 617
Chiplet design will offset that increased production time.
That being said, I don't see a 64c Threadripper.
At these performance levels, I'm not sure that there's a need to replace the entire product stack every 12 months.
That being said, I don't see a 64c Threadripper.
At these performance levels, I'm not sure that there's a need to replace the entire product stack every 12 months.
yeah, assuming these turn up as good as we hope, I think there will be a little bit of a break before the next lot turn up. And my guess is that the 4000 series will be on a new socket with support for ddr5 and very little actual change to the cores. Basically a new IO die and socket but very similar core chiplets
as discussed proof that these amd chips are as some state on here. 2 times as fast . faster mid range or low end than 9900ks. people are getting swept away based on a cinebench score. not seen one bit of proof out of 181 pages which shows any of this is true.
Permabanned
- Joined
- 23 Apr 2014
- Posts
- 23,552
- Location
- Hertfordshire
Nobody is stating fact, they are just speculating. Unfortunately it seems this is some kind of crime. 
@~>Dg<~
On the Cinebench score, its self explanatory, unless you want to accuse them of making the whole thing up, if so them that is your argument, so make it that, failing that... Dg if you do not believe what is there to see with your own eyes then no one can help you, how do you prove there is no milk in the bowl when you turn the dam thing upside down and still ask for proof?????
On the 2X as fast, if you accept the Cinebench score was genuine then if we double the core count from 8 to 16 then we double the score, there is not hard fact knowledge that there will be a 16 core version of it but Lisa did categorically say it will be more than 8, given that one chiplet is 8 cores it would require another chiplet to make up any number higher than that and its likely to be another 8 core chiples making 16 cores, yes its speculation but from very solid grounds.
On the Cinebench score, its self explanatory, unless you want to accuse them of making the whole thing up, if so them that is your argument, so make it that, failing that... Dg if you do not believe what is there to see with your own eyes then no one can help you, how do you prove there is no milk in the bowl when you turn the dam thing upside down and still ask for proof?????
On the 2X as fast, if you accept the Cinebench score was genuine then if we double the core count from 8 to 16 then we double the score, there is not hard fact knowledge that there will be a 16 core version of it but Lisa did categorically say it will be more than 8, given that one chiplet is 8 cores it would require another chiplet to make up any number higher than that and its likely to be another 8 core chiples making 16 cores, yes its speculation but from very solid grounds.
Last edited:
I think everybody was surprised when a 32 core 'Ripper showed up, so I'm not fully ruling out a 64 core monster showing up. However, I think it's more likely we'll have 16, 24, 32 and 48 cores for this generation, simply because those complete 8 core chiplets will be binned for the rest of the products:
the most power efficienct will go to EPYC
the highest clockers will go to Ryzen 9
the working but unremarkable will probably go to the 8 core APUs end of the year
the reasonable clockers with reasonable power draw would more likely end up as Ryzen 5s
I don't really see a bin that could be allocated for a 64 core Threadripper, unless yields are so good that AMD really do have enough chiplets to be even more granular with their binning.
Then I suggest you just go away for 6 months until tangible numbers from released products are available. Otherwise you're literally just posting irrelevant crap and stomping your feet like a child. There is no proof to give you, so why continue posting and posting demanding something you know you can't get?
Nobody is stating fact, they are just speculating. Unfortunately it seems this is some kind of crime.
And yet we're the ones who the mods will crack down on for calling out a troll's troll
These chiplets are tiny so I wouldn't rule out a crazy good yield. The 14nm IO die is going to be pretty cheap to produce and will likely have redundant logic to improve yields there too (4 memory controllers etc). Maybe a return to unlocking cores?
