Question Speculation: Ryzen 3000 series pricing

Page 2 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Speculation: Ryzen 3000 series retail pricing

  • 3700X -> 8 cores at $330, 3800X -> 12 cores at $500, 16 core possibly at higher price.

    Votes: 26 40.6%
  • 3700X -> 12 cores at $330, 3800X -> 16 cores at $500

    Votes: 31 48.4%
  • They are both too expensive. AMD's high end can't cost that much because no one will buy it

    Votes: 3 4.7%
  • They are both too cheap. AMD will go after market share at all cost.

    Votes: 4 6.3%

  • Total voters
    64

Mopetar

Diamond Member
Jan 31, 2011
4,445
372
126
#26
I want to belive that AMD is no forced to populate all 8 chips on the big I/O die.
The only reason they did it with ThreadRipper (the first chips only had 2 working die) was for stability. Since they just used dead chips, there's no technical reason that they should have to fully populate the CPU.

However, it doest make some sense to assume that they might want to use at least 4 chiplets (functional or not) on each ThreadRipper part for the same reasons as before. Perhaps it's less necessary now that there's a huge IO die in the middle of the chip to balance the IHS on, but it still seems reasonable to assume they will.

Whether they do or not should probably give us a good indication of what their yields are like. Since they moved most of the IO off of the chiplet, there isn't going to be anywhere near as much die area where a defect outright kills the chip.
 

maddie

Platinum Member
Jul 18, 2010
2,436
358
136
#27
This is wrong the moment you mention 6C/8C "G" model, there will be no such thing unless the I/O die is hidding something like a Vega 11, and even if thats the case, they may just go forward and call everything G. So remove those 2 G models, add Ryzen 3 3200G 4C/8T Picasso at $99 and it is possible.
I too don't think a "G" model will happen, but your reason as to why not, is wrong. The I/O die does not have to host the GPU. There is room for another die on the package with all the IF traces already existing and waiting to be used.

Reminds me of all the certainty by some that there would never be a multi-die + I/O for desktop Ryzen 3xxx. Conveniently forgotten though.
 

Shivansps

Platinum Member
Sep 11, 2013
2,425
173
126
#28
I too don't think a "G" model will happen, but your reason as to why not, is wrong. The I/O die does not have to host the GPU. There is room for another die on the package with all the IF traces already existing and waiting to be used.

Reminds me of all the certainty by some that there would never be a multi-die + I/O for desktop Ryzen 3xxx. Conveniently forgotten though.
AMD already said NO to GPU chiplets, at least for this gen there is no Zen 2 APU chiplets.
 

Kenmitch

Diamond Member
Oct 10, 1999
7,132
55
126
#30
I have no idea on what AMD will price the new offerings at.

I'd imagine AMD's plans for Zen 2 is more geared towards brand recognition and market share. AMD is much leaner than Intel so it's not like they have to inflict crazy pricing on the end user to make what is considered a reasonable profit.

Interesting times ahead. I look forward to reading the reviews and any new info that pops up.
 

Timorous

Senior member
Oct 27, 2008
366
57
116
#32
This is wrong the moment you mention 6C/8C "G" model, there will be no such thing unless the I/O die is hidding something like a Vega 11, and even if thats the case, they may just go forward and call everything G. So remove those 2 G models, add Ryzen 3 3200G 4C/8T Picasso at $99 and it is possible.
I almost did not add APUs because I was not all that sure as the APUs tend to be behind the latest and greatest so they might just do a small bump like they did with the mobile line up.

The other option is a custom io die with a built in GPU paired with a zen2 die but since they have already designed and built the 3xxx series mobile APUs they might as well use those designs at higher tdp in the desktop.
 
Dec 31, 2016
193
21
71
#33
12 core version: 399 €, -50 € for low power version. Über halo version maybe 499 €.

8 core version: 299 €, -50 € for low power version.

6 core version: 199 €, -50 € for low power version.

Rest: cheap. No 16 core version for AM4.
 

StinkyPinky

Diamond Member
Jul 6, 2002
6,355
42
126
#35
I think they will hold back on the 16 core also. They can do one obviously, but they will wait and see what Intel are doing with their 10nm first. Intel cannot even compete with 12 cores at 14nm.

The other advantage of this approach is that going 12 core gives them more yields as 16 cores will require no defects. This also gives them something to show for the Ryzen 4000 series.
 

maddie

Platinum Member
Jul 18, 2010
2,436
358
136
#36

maddie

Platinum Member
Jul 18, 2010
2,436
358
136
#37
I think they will hold back on the 16 core also. They can do one obviously, but they will wait and see what Intel are doing with their 10nm first. Intel cannot even compete with 12 cores at 14nm.

The other advantage of this approach is that going 12 core gives them more yields as 16 cores will require no defects. This also gives them something to show for the Ryzen 4000 series.
This reasoning, "but they will wait and see what Intel are doing with their 10nm first", cedes the initiative to Intel. Same argument used with the pricing. Intel remains calling the shots.

We keep forgetting that Zen 3 is 2020.
 

IEC

Super Moderator
Super Moderator
Jun 10, 2004
13,527
276
136
#38
I think they will hold back on the 16 core also. They can do one obviously, but they will wait and see what Intel are doing with their 10nm first. Intel cannot even compete with 12 cores at 14nm.

The other advantage of this approach is that going 12 core gives them more yields as 16 cores will require no defects. This also gives them something to show for the Ryzen 4000 series.
I would be happy if they'd go full 16 cores from the get-go. Those 8-core chiplets probably have good yields due to their tiny size. If there was any reason they'd hold back, it would probably be to funnel all the 8-core chiplets into server products until they ramp up production enough. But given that their mid-range single chiplet part is 8c/16t that scenario is apparently not true.

As far as precedent: No one envisioned them launching an 8-core Ryzen 1000 series when Intel had only 4-core CPUs. It wouldn't be a big surprise to see them double-up versus Intel yet again for the PR/performance crown. Though Lisa Su sure is good at teasing us by only stating "more than 8 cores" rather than specifying a number.
 

Shivansps

Platinum Member
Sep 11, 2013
2,425
173
126
#39
I would be happy if they'd go full 16 cores from the get-go. Those 8-core chiplets probably have good yields due to their tiny size. If there was any reason they'd hold back, it would probably be to funnel all the 8-core chiplets into server products until they ramp up production enough. But given that their mid-range single chiplet part is 8c/16t that scenario is apparently not true.

As far as precedent: No one envisioned them launching an 8-core Ryzen 1000 series when Intel had only 4-core CPUs. It wouldn't be a big surprise to see them double-up versus Intel yet again for the PR/performance crown. Though Lisa Su sure is good at teasing us by only stating "more than 8 cores" rather than specifying a number.
The only reason for going 16C on AM4 may be that it is just too expensive to do it on TR4... That big I/O die alone....
 

Topweasel

Diamond Member
Oct 19, 2000
4,742
345
136
#40
This reasoning, "but they will wait and see what Intel are doing with their 10nm first", cedes the initiative to Intel. Same argument used with the pricing. Intel remains calling the shots.

We keep forgetting that Zen 3 is 2020.
Honestly you look at some off Intels moves X299 (kicking up cores last second), Coffee Lake -R, Comet Lake. All of these the public and even some people here that should know better as reactionary products where only the X299 is. Some exist because rumors regarding AMD before Zen's launch going back to 2015 and 14 even. But CPU development, testing, and release is a much slower process then most are willing to understand. AMD will know this and releasing first makes them look like the Pioneer and not some stupid game of oneupmanship.

On top of that AMD had a recent CPU release where they had a chance to come out with a CPU that had so much extra value in it that it shocked the world. What they do. Big guns at a premium first and work back. All these people wanting a 9900k killer 8c super fast CPU. Are they going to be able to ignore a 9900k killing $500 16c CPU or 9900k killing $375 12c CPU (lets pretend they can clock it high enough for now we can balance out how much slower or faster these are later), for a month betting on a $300 8c super clocked single chiplet 9900k killer? That's what they did with the Ryzen 1, I don't know how many times I read about someone caving in when they were so sure the 1600x was going to be the better performer (honestly they were equal because the power usage doesn't drop much with 2 less cores).

Releasing the 16c and 12c first at reasonable (not cheap, but not TR expensive) allows AMD's products look like amazing values (good value doesn't mean super cheap), puts them back in that position again.
 

beginner99

Diamond Member
Jun 2, 2009
4,001
122
126
#41
The only reason for going 16C on AM4 may be that it is just too expensive to do it on TR4... That big I/O die alone....
Yeah my thought as well that TR3 starts with 24 cores and up. That would be 4 chiplets with 6 active cores each.
 
Feb 2, 2009
12,882
176
126
#42
With the current problems Intel has with 10nm and right now with 14nm volumes, Im expecting AMD to sell the 7nm 8C 16T at higher than $399. But we have to wait and see how Intel 14nm volume will change the coming months until Ryzen 3 launch.
If Intel 14nm volume will increase significantly and Desktop availability and prices of Intel processors will return to the normal points then AMD will rearrange its price strategy again.
 

krumme

Diamond Member
Oct 9, 2009
5,704
68
136
#43
They are launching the 8 core first, not a 12 core. Lisa said so in her interview.
No she did not. She said the opposite. We can expect more than 8 cores. How many is unknown. So its 12 or 16 day one.
 

krumme

Diamond Member
Oct 9, 2009
5,704
68
136
#44
With the current problems Intel has with 10nm and right now with 14nm volumes, Im expecting AMD to sell the 7nm 8C 16T at higher than $399. But we have to wait and see how Intel 14nm volume will change the coming months until Ryzen 3 launch.
If Intel 14nm volume will increase significantly and Desktop availability and prices of Intel processors will return to the normal points then AMD will rearrange its price strategy again.
We have paid nearly double to go from 4c4t to 4c8t for years. And are willing to do so. Nv shows us how this can be done to the extreme.

With a product stack ranging from 6c to 16c i think to most important strategy is to cover all segments as there is plenty opportunity. Those who buy 1030 gpu and those who can afford 2080ti. In that way you maximize revenue and profit.

I think it's a strategy to get so much doe from all the Joe's as possible. All.

Regardless of supply and competition I think that tells us something about bottom and top pricing. It needs to be stretched out. Because they can wipe the entire floor.
So rough guess 139usd 6c low freq to 799 uber binned 16c.
I see no reason not to go 16c out the gate. They just need to price it like the old days. 799usd is cheap here. 999 usd wouldn't be out of order.

That is if mem latency have come under control. If not they are stuck at 499 for 16c top end. Like 1800x vs 7700k.
 

Topweasel

Diamond Member
Oct 19, 2000
4,742
345
136
#45
Yeah my thought as well that TR3 starts with 24 cores and up. That would be 4 chiplets with 6 active cores each.
Honestly I could see them do like they did with the original 1900x. Have a 16c TR at $50-75 more than the AM4 version for workstation users that need more IO. But that probably depends on how much stock is left on the 2950, I think they stayed away from the 8c TR2 because they still had a bunch of 1900's in the wild.
 

PeterScott

Platinum Member
Jul 7, 2017
2,605
226
96
#46
As far as precedent: No one envisioned them launching an 8-core Ryzen 1000 series when Intel had only 4-core CPUs. It wouldn't be a big surprise to see them double-up versus Intel yet again for the PR/performance crown. Though Lisa Su sure is good at teasing us by only stating "more than 8 cores" rather than specifying a number.
AMD was selling 8 core parts for years, before they released Ryzen. It was just that they sucked.

In fact they had a very similar product stack before and after ryzen.

8 core desktops without iGPU, and 4 core APU parts with iGPU.
 
Apr 27, 2000
11,175
690
126
#47
AMD was selling 8 core parts for years, before they released Ryzen. It was just that they sucked.
It wasn't just that they sucked . . . their dubious status as 8c products forced people to use the "module" nomenclature to end ridiculous arguments. We had numerous fights on this very forum over whether you could call Bulldozer or Piledriver an 8c CPU. It has the INT resources of an 8c/8t CPU, and the FP resources of a 4c/4t CPU (or 4c/8t CPU in fp, if you were being generous).

AMD doubled their maximum thread count to 16 in 2017 with a true 8c/16t CPU that performed just as well in fp as it did in int. That was the shocker. That along with the per-thread performance.
 

Shivansps

Platinum Member
Sep 11, 2013
2,425
173
126
#48
It wasn't just that they sucked . . . their dubious status as 8c products forced people to use the "module" nomenclature to end ridiculous arguments. We had numerous fights on this very forum over whether you could call Bulldozer or Piledriver an 8c CPU. It has the INT resources of an 8c/8t CPU, and the FP resources of a 4c/4t CPU (or 4c/8t CPU in fp, if you were being generous).

AMD doubled their maximum thread count to 16 in 2017 with a true 8c/16t CPU that performed just as well in fp as it did in int. That was the shocker. That along with the per-thread performance.
AMD does considers them as full cores, i had a talk about that here with an AMD rep when the 200GE launched locally, they consider BR as quad cores. Even trought Windows says otherwise. I dont think what WE think is important on this matter.
 
Apr 27, 2000
11,175
690
126
#49
AMD does considers them as full cores, i had a talk about that here with an AMD rep when the 200GE launched locally, they consider BR as quad cores. Even trought Windows says otherwise. I dont think what WE think is important on this matter.
As you wish.

Point being, AMD went from 8t to 16t in 2017. That was a big deal. @IEC wasn't wrong.
 
Nov 6, 2014
114
10
101
#50
12 core version: 399 €, -50 € for low power version. Über halo version maybe 499 €.

8 core version: 299 €, -50 € for low power version.

6 core version: 199 €, -50 € for low power version.

Rest: cheap. No 16 core version for AM4.
Thats more or less what I was thinking. 12-core should be more than enough for AMD in round 1 (before Intel responses) especially if clocks and IPC go up as well.

But if they're going all in then we'll certainly see a 16-core variant at ~500-600USD. The top 12c will cost more than the 2700X regardless which route they choose (expecting 350-400USD MRSP).
 

Similar threads



ASK THE COMMUNITY

TRENDING THREADS