AMD “Next Horizon Event" Thread

[coercitiv]

I followed that you were making a ridiculous assumption that R3 would be 8cores, when it most cerainly will not be.

Let's see what you followed.

First there was a post from a fellow forum memeber:

I was just thinking about this in combination with the 1x vs 2x CCX. If AMD keeps the same overall strategy for consumer and enterprise we would have 8 core 1x CCX as the base chip. They can bin or fuse to get less cores sure. Depending on how cheap the chiplet strategy is 8 core could possibly be the lowest core count offered. I think it gives AMD an advantage to continue ramping up core counts because basically starves Intel. Intel is stuck on process. The bigger the chips they have to pump out the less chips they can make and the less profits they can reap. The other cool thing is if 8 core is the new norm that means something bigger like 16 core has to take the place on the high end desktop. Can you imagine what devs could do with an extra 8 core chip? A new best AI mode. Run actual simulations. It would be completely new territory. Probably dreaming. Hahaha

To which I replied with my concerns that an 8 core minimum, which would imply an 8 core R3, would not be effective for consumers (cost wise):

Let's not forget this mainstream consumer product needs to keep cost down and also work efficiently with dual channel memory. The more chiplets they use, the bigger the size of the IO chip, and the higher the cost of the entire package.

After which another forum member joined in and talked about design costs, and how these would likely dictate using chiplets in mainstream. Since I felt my concerns of using up to 16c/32t in mainstream were not properly understood, I followed up by another post trying to establish the context of my reply:

What basement? I was replying to a post talking about 8c being the minimum core count in Zen 3000 series.

And another, more comprehensive explanation in the attempt to make things as crystal clear as possible:

I feel like we're speaking past each other. I expressed concern over the idea that Ryzen 3000 would end up using multiple 7nm chiplets togheter with an arguably sizeable IO die in the mainstream consumer line. That concern was directly related to manufacturing cost. You keep talking about design cost.

Whether Ryzen 3000 series should or would use a custom design is another discussion altogether, my issue is with the idea of Ryzen 3000 starting at 8c/8t minimum and building towards 16c/32t in order for segmentation to allow multiple price points, including that sweet sub $200 spot. It would be expensive and it would also have problems scaling with dual channel memory.

At this point you came in and asked what made me think there would be 16 cores on mainstream, and that my assumption of R3 with 8 cores is ridiculous. To this I say again, please read the thread. You are essentially disagreeing with someone who was expressing a similar opinion to yours, that R3 8 core is unlikely.

 

[krumme]

I suspect AMD are going to have 2 IO Controllers (IOC).

The one we already know about - Rome - with 8 Infinity Fabric (IF) ports so they can run 8x8C chiplet = 64 cores. All EPYC and TR variants will used fused down versions of this IMO - both in terms of fusing off IF ports and in dropping memory channels.


Then the other IOC is a much smaller 2x IF port version with dual channel memory controller, for use in:
(i) APU - 1x8C* chiplet + 7nm Vega (AMD already have announced 7Vega can communicate via IF).
(ii) Mainstream with 1 of the ports fused off, so 1x8C* chiplet.
(iii) High end of mainstream for 2x8C* chiplets.

*CPU cores fused off as required by market/salvage.


So 1x 7nm chiplet design and 2x 14nm IOC designs to cover the market from 4C APU right through to 64C Rome. Pretty efficient use of resources IMO.

Agree. 8x and 2x io dies.
It doesn't matter the slightest for cost if the 2x io was 1x. But hey they can make 1 and 4 too. But why?
Now with 14nm apu in the backhand I would start at 6c and go up. No need for less as yield must be very very high at 73mm2.
6 8 12 16
A second reason and that is a tactics used especially by Intel is to drain you competitor for cash. Intel have plenty of that but it's not simple when you have to feed enormous fixed cost and want to keep share price and margins. I would strike.
This chiplet idea combined with sharing cost with Apple so to speak is simply brilliant.
I would really like to to know who made this. From a business perspective it just looks like a gift from heaven but let's see how those latencies plays out. But man. Intel looks outplayed now.
Good days. More competition.

 

[Mopetar]

I also think that it's unlikely to see Ryzen use more than 8C (1 chiplet) as there's little reason for AMD to do so at this time. If they're getting ~10-15% bumps for both clock speed and IPC, there's sufficient reason to upgrade right there. For a big part of the market, 8C is already more than they need, so AMD is better off selling a 4C part as R3 that's much cheaper to produce and allows them to offer better prices to consumers.

Also, having to use 2 chiplets for Ryzen means that they can only effectively sell half the chips. Even though it might only cost an extra $20 in parts, that's potentially hundreds of dollars of revenue that they don't get by having more product. And they're already going to have limited supply due to being on a new process and wanting to use most of their chiplets for Epyc to get the most revenue. Even if they are just under parity with Intel, they'll still sell a lot of chips. If they're within 5% of Intel's 9900k, but sell that for what the top 8C/16T Ryzen goes for (~$350), it's pretty much a slam dunk in terms of which to buy.

 

[krumme]

Why should they use 2 chiplets just because the io can do it?
With an 2x io you can just use the chiplets the market demands. That's how I understand the concept?

 

[HurleyBird]

I also think that it's unlikely to see Ryzen use more than 8C (1 chiplet) as there's little reason for AMD to do so at this time. If they're getting ~10-15% bumps for both clock speed and IPC, there's sufficient reason to upgrade right there.

That reasoning sucks, and is the exact kind of thinking that single-handedly lost AMD the GTX 280-HD 4870 generation despite enjoying a massive architectural advantage. If you have the capability to go for your opponent's throat versus leaving them enough breathing room to compete against you, it's almost always the right decision to do the former.

Which isn't to say that I think 16-core is more likely than 8-core, since that depends on how aggressive AMD decided to pursue the consumer space some time ago based on then future projections, but there's no doubt of what would be the better decision in hindsight, and that's upping the core count -- at least assuming no pressing supply side constraints.

 

[PeterScott]

I also think that it's unlikely to see Ryzen use more than 8C (1 chiplet) as there's little reason for AMD to do so at this time. If they're getting ~10-15% bumps for both clock speed and IPC, there's sufficient reason to upgrade right there. For a big part of the market, 8C is already more than they need, so AMD is better off selling a 4C part as R3 that's much cheaper to produce and allows them to offer better prices to consumers.

Also, having to use 2 chiplets for Ryzen means that they can only effectively sell half the chips. Even though it might only cost an extra $20 in parts, that's potentially hundreds of dollars of revenue that they don't get by having more product. And they're already going to have limited supply due to being on a new process and wanting to use most of their chiplets for Epyc to get the most revenue. Even if they are just under parity with Intel, they'll still sell a lot of chips. If they're within 5% of Intel's 9900k, but sell that for what the top 8C/16T Ryzen goes for (~$350), it's pretty much a slam dunk in terms of which to buy.

More succinctly: An 8C minimum for R3, would probably have 99% of buyers choose the minimum. That drives down ASP/profits.

 

[HurleyBird]

More succinctly: An 8C minimum for R3, would probably have 99% of buyers choose the minimum. That drives down ASP/profits.

If the lowest of AMD's consumer processors can trade blows with Intel's highest end while consuming less power, then that's R300/G80/Conroe all over again. That situation never, ever, not in a million years, lowers profits. It's only ever the opposite. Demand for the low end products may go up faster than the high end, but the absolute demand for high end products is still significantly greater than it would be otherwise.

Or to put it another way, market psychology dictates that presumed equality between your low and a competitor's high end products serves to make your own high end parts significantly more appealing.

 

[Doom2pro]

What is this photo on AMD Reddit? Is it a photoshop or real deal? Can't seem to find any source or any details at all.

 

[DownTheSky]

Photoshop

 

[Hitman928]

Looks like a photoshop of the full Zen 2 Epyc die to me. Compare the reflections in the die to the following:

 

[Doom2pro]

Looks like a photoshop of the full Zen 2 Epyc die to me. Compare the reflections in the die to the following:

Not seeing any resemblance. Your photo is blurry, smaller and totally different reflections. Not sure why you linked to it.

 

[HurleyBird]

It's a shop based off of a different photo with different reflections.

 

[Hitman928]

Not seeing any resemblance. Your photo is blurry, smaller and totally different reflections. Not sure why you linked to it.

I'm not saying the reflections are exactly the same, but the reflections on the reddit photo don't look right compared to the legit Epyc picture. Looks like the dies have been digitally manipulated.

 

[Abwx]

What is this photo on AMD Reddit? Is it a photoshop or real deal? Can't seem to find any source or any details at all.

A reflected finger appear on the bottom chiplet but no hand that hold the chip, lol...

Exemple :

 

[Hitman928]

I think this is the photo they used for the packaging and background. They just widened the packaging, flipped the columns on the right, and mirrored the left side as well as copying the black surface in the background to make it bigger. This is a 2600.

 

[Doom2pro]

Thanks guy, had a feeling... I can't stand these reddit fan fiction no source bait posts.

 

[Abwx]

Thanks guy, had a feeling... I can't stand these reddit fan fiction no source bait posts.

Pic is a fake but not the concept, though, we may well see 8C/16C + IO chiplet sooner rather than later...

 

[PeterScott]

If the lowest of AMD's consumer processors can trade blows with Intel's highest end while consuming less power, then that's R300/G80/Conroe all over again. That situation never, ever, not a million years, lowers profits. It's only ever the opposite. Demand for the low end products may go up faster than the high end, but the absolute demand for high end products is still significantly greater than it would be otherwise.

Dream on.

Then if it is such a great idea, you must think R3 will have 8 cores?

I am certain it won't because it's a poor business case. It absolutely will lower profits, because it will kill demand for anything but R3...

 

[HurleyBird]

I am certain it won't because it's a poor business case. It absolutely will lower profits, because it will kill demand for anything but R3...

You're completely wrong. You assume that the market is hyper-rational, but it isn't. "Good enough" is never enough when it's marketed as a budget option. This same phenomenon explains why the 8700K is suddenly less enticing now that the 9900K exists, despite the fact that 99.99% of users won't be able to tell the two apart in a blind test, and why consumers buy massively expensive phones over cheaper alternatives. A 16-core high end part is more inciting than an 8-core budget part by the very reason of being a 16-core high end part. Unlike, say, a 400MHz Celeron, you can't overclock an 8-core part into a 16-core one.

What really drives consumers, time after time after time, is the idea of complete superiority. That's always been the case, and obviously so. AMD can possibly attain that with a 16-core part (or perhaps even a 12-core part with high enough clocks and IPC) in a way that's probably impossible if they stick with 8 cores max.

Of course, I'm not arguing whether or not AMD made the decision to up the core count on mainstream parts, but that knowing everything that we do currently upping the core count is the better decision.

 

[maddie]

Dream on.

Then if it is such a great idea, you must think R3 will have 8 cores?

I am certain it won't because it's a poor business case. It absolutely will lower profits, because it will kill demand for anything but R3...

Now why didn't I think of this argument when Ryzen competed with the i7 7700 and the i5 7600? Lowest Ryzen same as the highest Intel. I would imagine that none of the higher core count Ryzens would sell. Who would want 4+ cores especially with HT?

I wonder what happened? Anyone knows?

Disclaimer: This is not an endorsement of 8C R3 models.

 

[amd6502]

TR4 is there as a proper platform for those needing over 16 threads.

Having a complex IO hub to coordinate multiple chiplets is going to raise the cost of all Zen2 consumer products.

Further, it could also mean higher latencies and not be especially useful for the vast majority of consumers. My hope and guess is they will use a simple one chiplet design for AM4.

Exotic 16c/32t can go to TR4 or wait for AM5.

 

[Panino Manino]

Before the even Ian joked about having with him a SDD full of benchmarks.
So, where are them? Where's the benchs?
Just that C-Ray?

 

[epsilon84]

I actually would love an 8C R3 for $100. That would mean we are back to the days when all you needed was a $100 CPU like the low end C2Ds and you'll have more than enough CPU grunt. I'll probably still be able to sell my 8700K for $250 or more and be in profit for an upgrade

It wouldnt be so great for AMDs ASP though if everyone upgraded like me, so I don't think we will see an 8C R3. R5 maybe?

 

[majord]

Let's see what you followed.

First there was a post from a fellow forum memeber:


To which I replied with my concerns that an 8 core minimum, which would imply an 8 core R3, would not be effective for consumers (cost wise):


After which another forum member joined in and talked about design costs, and how these would likely dictate using chiplets in mainstream. Since I felt my concerns of using up to 16c/32t in mainstream were not properly understood, I followed up by another post trying to establish the context of my reply:


And another, more comprehensive explanation in the attempt to make things as crystal clear as possible:


At this point you came in and asked what made me think there would be 16 cores on mainstream, and that my assumption of R3 with 8 cores is ridiculous. To this I say again, please read the thread. You are essentially disagreeing with someone who was expressing a similar opinion to yours, that R3 8 core is unlikely.

Current r3 are 8 core under the hood already.

 

[PotatoWithEarsOnSide]

An 8C R3 with improved clocks and IPC does make good business sense theoretically, since it would be akin to starving Intel of huge swathes of revenue; whilst its true that AMD would end up with much tighter margins, their increase in sales volume should still result in large profit increases.
However, such a theory relies upon AMD actually being able to meet the hugely increased demand, which we know that they couldn't. Who then fills this unmet demand? And with which products? My best guess is that current gen Ryzen would need to be sold at cost. In all honesty, I think that Ryzen 3xxxx will be pretty expensive if its as good as we all hope it to be, especially so if its 8C R3 minimum. I just don't see how such a product could exist anywhere near its current price point unless yields were phenomenal, and even then its almost certain that supply would be heavily constrained relative to demand.