Zen 2 APUs/"Renoir" discussion thread

[jpiniero]

I simply said that a 155mm^2 part is replacing a 210mm^2 part and we will be paying close to 2X the price. How many cores they manage to fit is part of the equation of giving more for the same price as technologies advance.

I think what it tells you is that AMD spent too much die space on Raven/Picasso's IGP.

 

[nicalandia]

What Intel does is irrelevant. The lesser of two evils is still evil.

Great, this is what Amd gets for making 8C/16T Available for the masses, bein called an Evil Corporation.

 

[maddie]

150 mm^2 are on a much more expensive process, it was rumored that 7nm wafers cost almost double of the cost of 12nm wafers, and development costs on smaller geometries are higher, too. Moreover the comparison is between launch cost of a new product (not even yet official) with the cost of another product that is EOL. We should look at the situation a couple of months after launch (of the retail parts, as the first APUs seems to be destined to the OEM market, according to the rumors).

With yield, area and wafer cost in the equation, I'm fairly certain that the end fab cost is quite close.

As a comparison, a bigger 7nm Navi 14 die (158mm^2) sold to a video card OEM then installed on a much larger PCB + 4GB GDDR6 + a comparatively massive cooler + tens (hundreds ?) of lesser components retails for around $180.

I don't think you can make the case for 7nm being this huge factor.

Look I'm not saying don't buy this product, but we should at least be honest and accept that AMD is increasing margins tremendously with the 4750.

 

[maddie]

Great, this is what Amd gets for making 8C/16T Available for the masses, bein called an Evil Corporation.

I just used a common term to get a point across. Would you prefer if I said the lesser of two bad situations is still a bad one? Is this nicer?

 

[VirtualLarry]

Dali is another Raven2 APU, they are gona keep using it for botton-line Athlons for some time. Regardless of what they are calling it, it is the same 2C/4T Raven Ridge with Vega 3 at 14nm. At $50 it is good.

You sure? I thought Picasso was the 2C die of Raven2, cheaper to make, etc., for the low-end, and used in things like the Ryzen 3200U and Athlon 300U laptops as well.

I thought Dali was a 7nm APU with smaller core-count, in the similar vein to Renoir. No?

Why would AMD invest in making an expensive mask set, again, for a 2C chip, if it's not going to be 7nm? Why two Raven2 2C APUs? Makes NO sense to me.

 

[VirtualLarry]

but we should at least be honest and accept that AMD is increasing margins tremendously with the 4750.

Is this a bad thing? More money, a stronger, more competitive AMD, and more R&D money for their pipeline(s). (*Maybe some of it will trickle into their GPU division, too. It could maybe use some help, although RDNA2 sounds like a strong arch. so far.)

 

[maddie]

Is this a bad thing? More money, a stronger, more competitive AMD, and more R&D money for their pipeline(s). (*Maybe some of it will trickle into their GPU division, too. It could maybe use some help, although RDNA2 sounds like a strong arch. so far.)

If I'm a stockholder, sure, but as a consumer, my intention is to get the most for the least expenditure. Do we really see ourselves as fighting the good fight and supporting a heroic company in it's struggle against dark forces?

If yes, then you might as well start a patron page where donations can be made to assist in R&D.

While I am and have been a supporter of AMD, I do not fool myself for a minute in thinking that they won't squeeze us as much as any other economic entity would, if conditions for it arise.

 

[NostaSeronx]

You sure? I thought Picasso was the 2C die of Raven2, cheaper to make, etc., for the low-end, and used in things like the Ryzen 3200U and Athlon 300U laptops as well.

I thought Dali was a 7nm APU with smaller core-count, in the similar vein to Renoir. No?

Picasso(210mm2) is the 12nm RTO of 14nm Raven(210mm2).

Dali and Pollock are 14nm APUs which are based on 14nm Raven2.(150mm2)
Dali is FP5 w/ dual-channel => Ryzen 3250U & 3250C // https://browser.geekbench.com/v5/cpu/1653279
Pollock is FT5 w/ single-channel => https://browser.geekbench.com/v5/cpu/2323968 (4.8W Sus, 6W sPPT, 9W fPPT)

Raven2(14nm) re-uses Picasso's(12nm) FW blob, while Dali(14nm)/Pollock(14nm) re-uses Renoir's(7nm) FW blob.

 

[moinmoin]

Look I'm not saying don't buy this product, but we should at least be honest and accept that AMD is increasing margins tremendously with the 4750.

Well yes, its their bleeding edge APU. It would be terribly sad if they didn't manage to increase margins tremendously with it considering it even appears to beat similar Matisse chips.

I'm not even sure why anybody would complain about AMD's MSRPs. Unlike with many Intel chips AMD's desktop chips so far rather reliably drop in price a couple months after launch (unless there's shortage). Due to that the MSRP is often not a constant but essentially an early adopter surcharge. So if you want it early pay the price, if you want it cheaper wait for the deals and/or street price drops.

 

[DrMrLordX]

No, you took one parameter (best of class), added another parameter (price) to claim that AMD is rising prices

No, I looked at the APU from desktop Renoir with the best iGPU compute performance (4750G) and pitted it against the similarly-positioned product from Picasso (3400G) because that is what was being discussed in the first place.

Page 5 - Zen 2 APUs/"Renoir" discussion thread

Page 5 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

forums.anandtech.com

Actually @Shivansps was also discussing iGPU gaming performance. In his opinion, apparently the not-quite-as-expesive 4600G is going to be the successor to the 3400G, and who knows, maybe he's right in terms of compute performance parity. Either way, if you want iGPU compute performance, AMD is making you pay more for it this generation.

any modern low level dGPU quickly passes any iGPU anyway.

Totally not the point.

While I am and have been a supporter of AMD, I do not fool myself for a minute in thinking that they won't squeeze us as much as any other economic entity would, if conditions for it arise.

Too right. It's not that AMD is deliberately bagging on iGPU performacne because they're incompetent or de-emphasizing iGPUs (per se). They just want more of your money to get that performance.

 

[Shivansps]

You sure? I thought Picasso was the 2C die of Raven2, cheaper to make, etc., for the low-end, and used in things like the Ryzen 3200U and Athlon 300U laptops as well.

I thought Dali was a 7nm APU with smaller core-count, in the similar vein to Renoir. No?

Why would AMD invest in making an expensive mask set, again, for a 2C chip, if it's not going to be 7nm? Why two Raven2 2C APUs? Makes NO sense to me.

Yeah Dali is again another 14nm Raven2, the 3150U for example is a Dali... at best it may have compatibility changes for B550, like 8K support. but that about it.

 

[Shivansps]

But guys like you want a R7 with a price of an R5? Yeah right. I say Amd have spoiled your kind too much, remember when 8C/16T were 1K a pop? Now Amd out of their kindess has given Peasants 8C/16T CPUs for $300 and the same Peasants now want those CPUs for $140.00 just because they feel entitled. I mean Really?

No, i want a R5 in replacement of a R5, at a similar price point, not an R3 just because it has the same core/thread count. Simple as that.

 

[KompuKare]

If I'm a stockholder, sure, but as a consumer, my intention is to get the most for the least expenditure.

Sure, but somewhere in the consideration is whether buying cheaper now means having to pay more the next time.
For instance the aim of dumping is to force the competition out and make up for period of losses by profiting later.
Intel's antics with Atom and contra-revenue could easily be argued to have been dumping, and at the time Jaguar was a pretty competitive design which sold very poorly (although ultimately shifted huge numbers in the XBO and PS4).

As for the cost to make Renoir vs the previous gen, do we even know that the assumption that a 150mm² 7nm part is at all similar in cost (ignoring the fixed costs of design, masks etc.) to a 220mm² 12nm part?

Okay, the defect rate of both processes were similar the smaller part should ultimate have a higher yield, but I find it unlikely that the defect rates are the same. Even packaging which we might take to be a constant is likely to be somewhat higher for 7nm.

 

[jpiniero]

No, i want a R5 in replacement of a R5, at a similar price point, not an R3 just because it has the same core/thread count. Simple as that.

If you look it that way, yeah they are raising prices. You could say that RR/Picasso's prices were suppressed because it would have been tough to justify going higher compared to Skylake.

 

[LightningZ71]

But, again, you are arguing a specific corner case, "raw compute performance of the iGPU is the only important factor and no other improvements to the entire product matter in any measurable way".

We have NO in game performance benchmarks, just a leak of system level benchmarks from one suite. Until we see something that shows otherwise, you're just beating a drum for a problem that doesn't yet exist. While compute performance is important for most games, it's not the only factor.

Let's wait and see the real numbers...

 

[AtenRa]

If those prices turn to be true, this is the worst APU release so far.

I really dont even want to speculate how ZEN 3 prices will be after those APUs.

 

[jpiniero]

If those prices turn to be true, this is the worst APU release so far.

I really dont even want to speculate how ZEN 3 prices will be after those APUs.

It's basically the same prices as the current non-XT Matisse. I would expect Renior to effectively replace them.

 

[DAPUNISHER]

If I'm a stockholder, sure, but as a consumer, my intention is to get the most for the least expenditure. Do we really see ourselves as fighting the good fight and supporting a heroic company in it's struggle against dark forces?

If yes, then you might as well start a patron page where donations can be made to assist in R&D.

While I am and have been a supporter of AMD, I do not fool myself for a minute in thinking that they won't squeeze us as much as any other economic entity would, if conditions for it arise.

Yeah, and if "ifs" and "buts" were candy and nuts, you'd have a concession stand.

We are a loooong way from AMD being in said position. And it is silly to trash them for things they have not even done yet. Sounds like thought crimes.

Some seem to forget this was AMD's situation less than a year ago AMD's meaningful debt burden

They are doing strong work, and these new APUs BETTER have a decent profit margin, they need every dime.

 

[AtenRa]

It's basically the same prices as the current non-XT Matisse. I would expect Renior to effectively replace them.

So then, no 3200G/3400G replacements.
And i highly doubt those APUs will be able to replace the non XT Matisse.

For example, there is no Renoir replacement in price and performance for the 3100 and especially the 3300X SKUs.

 

[leoneazzurro]

With yield, area and wafer cost in the equation, I'm fairly certain that the end fab cost is quite close.

As a comparison, a bigger 7nm Navi 14 die (158mm^2) sold to a video card OEM then installed on a much larger PCB + 4GB GDDR6 + a comparatively massive cooler + tens (hundreds ?) of lesser components retails for around $180.

I don't think you can make the case for 7nm being this huge factor.

Look I'm not saying don't buy this product, but we should at least be honest and accept that AMD is increasing margins tremendously with the 4750.

I was only considering the fact you were comparing two different processes with different base costs, and btw yields being higher on 7nm respect to 12nm are news to me. And yes, there are other factors in the costs like packaging, development, testing and so on. But, you all are not considering that pricing is something that is driven by the price/performance ratio defined by the market's EXTERNAL competition, not internal. Yes, 3400G may or may have not better graphics performance respect to 4350. For sure 4350 has better CPU performance. But the pricing of 3400G (and 4350 that will have a similar performance/price ratio) is dictated by what the competition offers and at what price. As 3400G will be phased out, if 4350 will be priced lower than that and the competition has a worse price/performance ratio, selling it for less will mean AMD will lose money for nothing. Same is true for the 4750. Having a 8 core APU, with same performance of a 3700X, and an integrated VGA, for the same costs of a 3400G? Yeah, dreaming is free, but when you look at the reality, AMD will shoot itself in the foot by doing so. Probably the scenery will change once the Zen3 CPU are launched, by giving the CPU-only units a performance advantage again. In that case, and accordint to what intel does, we wil see if prices will go down on APUs.

 

[leoneazzurro]

So then, no 3200G/3400G replacements.
And i highly doubt those APUs will be able to replace the non XT Matisse.

For example, there is no Renoir replacement in price and performance for the 3100 and especially the 3300X SKUs.

This is why there are rumors of a "small APU" with RDNA graphics coming next year. Current 4000 APU line was conceived clearly for the mobile market first and to fight with the U/H lines of mobile Intel processors, which they did well but left the lowest end uncovered except by the old 3000 APUS (and they lack something similar to the Y line).

 

[jpiniero]

For example, there is no Renoir replacement in price and performance for the 3100 and especially the 3300X SKUs.

I guess $149 would be a tad higher. I Figure there will actually be one but suspect that will be OEM only.

The 3100 and 3300X might stick around just for yield purposes.

 

[AtenRa]

This is why there are rumors of a "small APU" with RDNA graphics coming next year. Current 4000 APU line was conceived clearly for the mobile market first and to fight with the U/H lines of mobile Intel processors, which they did well but left the lowest end uncovered except by the old 3000 APUS (and they lack something similar to the Y line).

Next year it will be 3 years since the original Ryzen 2200G/2400G and 2 years since 3200G/3400G. That is too little too late for a small ZEN 2 APU when ZEN 3 will also be available.
Also, previous APUs were also made for Mobile first, so i dont see whats different this time with Renoir.

 

[moinmoin]

For example, there is no Renoir replacement in price and performance for the 3100 and especially the 3300X SKUs.

God forbid AMD not replacing SKUs announced a mere 3 months ago right away and give an iGPU for free on top of it. I mean, come on now...

 

[leoneazzurro]

Next year it will be 3 years since the original Ryzen 2200G/2400G and 2 years since 3200G/3400G. That is too little too late for a small ZEN 2 APU when ZEN 3 will also be available.
Also, previous APUs were also made for Mobile first, so i dont see whats different this time with Renoir.

The difference is that series 4000 APU is a serious contender for Intel in price/performance/power consumption where 3000 series was still lacking in comparison to Intel offerings, and 4000 series was aiming especially at the high end/higher margins laptops, being 8 core native. And you don't contend on a low margin market like the one covered by Intel with 2 /4core APUs with a 8 core native part on a more expensive process. AMD has not the same resources as Intel, and thus they have priorities, and of course their priorities are where the higher the margins are. In the meanwhile, if you want an AMD APU with low core count, you can go for the older 3000 series which still offers more than good competition with even priced Intel parts.