***OFFICIAL*** Ryzen 5000 / Zen 3 Launch Thread REVIEWS BEGIN PAGE 39

[moinmoin]

If they really cared about having the best profit per wafer, then why produce SoCs for the consoles?

Because in semi custom business AMD is not the one making the decisions, AMD's customer is. With semi custom business AMD is essentially doing project work for Sony and Microsoft, being paid for the necessary R&D as well as the production of chips. It's essentially risk free money for AMD, Sony and Microsoft pay for all the necessary work and expenses and will have to worry about allocation and sales, whereas AMD essentially just does as being told and gets money for doing so.

AMD's own APU product line is completely irrelevant to all that. Both Raven Ridge and Picasso have been low end chips within AMD's Zen product portfolio, and Renoir is a successor within that line. For its own products AMD has been very conservative in many areas, be it ordering wafer capacity in advance, be it allocation for high end chips (leading to the prolonged scarcity of the popular 3900X), be it being completely unprepared for the popularity that Renoir turned out to have. Cezanne, fully in line with how AMD handled its APUs so far, seems to be little more than an upgrade of Renoir to Zen 3.

I fully expect AMD to eventually sell high end premium APUs as part of its product portfolio. But that will take some more years as the current output (result of several years of planning and execution) is still that of the more conservative cash strapped financial risks averse AMD of the recent past.

 

[Shivansps]

I fully expect AMD to eventually sell high end premium APUs as part of its product portfolio. But that will take some more years as the current output (result of several years of planning and execution) is still that of the more conservative cash strapped financial risks averse AMD of the recent past.

I dont understand how replacing a 4 year old GPU would suddenly turn a low end APU into a premium APU, thats normal progression. In fact whats making a APU premium is the number of cores, the 8 cores alone means it is a +$300 SKU in desktop what is useless, in notebooks, you probably want a dGPU with that many cores. This is why APUs arent balanced, starting with the 8C Renoir.

 

[teejee]

I dont understand how replacing a 4 year old GPU would suddenly turn a low end APU into a premium APU, thats normal progression. In fact whats making a APU premium is the number of cores, the 8 cores alone means it is a +$300 SKU in desktop what is useless, in notebooks, you probably want a dGPU with that many cores. This is why APUs arent balanced, starting with the 8C Renoir.

You are fighting for a market that too few cares about. Gaming on APU/iGPU is low prio for both OEM and most buyers.

And Cezanne will be succesful thanks to powerful CPU, either with discret GPU for gaming interested buyers, or standalone for buyers with little interest in gaming (but it can still handle many games though).

This is the right balance to continue taking market share from Intel in laptops. Wasting silcon on a niche market is not the right balance.

Your opinion/needs doesn't change this.

 

[krumme]

you guys think if zen4 have io die at 12nm too?

 

[moinmoin]

you guys think if zen4 have io die at 12nm too?

Zen 4 will be on a completely new platform, so the IOD will be redesigned as well, possibly even split up into smaller parts, some of which may stay at GloFo and others that profit of smaller nodes.

 

[lobz]

so It's ok using an 3 years old gpu arch in Cezanne when they need to defeat Tiger Lake IGP while having a better arch named RDNA?
RDNA1,2 has much higher IPC and is more power efficient and also bandwidth efficient compared to Vega, but It uses more die space.
If they want to use Vega, then at least use more CUs, ROPs even If they are at lower clocks, It's still more power efficient that way than clocking It too high, but the last leak mentioned only 8CUs.

At least I can somewhat understand Van Gogh. They probably won't release more than 4 core models so CCX from Zen3 couldn't be used, only Zen2 was possible.

100% honestly, I couldn't care less even if they used VLIW4, granted it performs competitively.

 

[Olikan]

And Cezanne will be succesful thanks to powerful CPU, either with discret GPU for gaming interested buyers, or standalone for buyers with little interest in gaming (but it can still handle many games though).

Good point, didn't thought about it... Cezanne is likely intended to ship with a Dgpu... while Van Gogh is intended for ultra low power...

Mmm... maybe zen3 is not as efficient as zen2 at lower clocks... hence the lower base clocks of zen3, while Van Gogh and consoles use zen2

 

[lightmanek]

Good point, didn't thought about it... Cezanne is likely intended to ship with a Dgpu... while Van Gogh is intended for ultra low power...

Mmm... maybe zen3 is not as efficient as zen2 at lower clocks... hence the lower base clocks of zen3, while Van Gogh and consoles use zen2

I think Zen 3 is not as frugal at the same clock as Zen 2, but certainly it is more efficient when normalising for work done in a given time.

 

[TESKATLIPOKA]

Good point, didn't thought about it... Cezanne is likely intended to ship with a Dgpu... while Van Gogh is intended for ultra low power...

Mmm... maybe zen3 is not as efficient as zen2 at lower clocks... hence the lower base clocks of zen3, while Van Gogh and consoles use zen2

Even If Zen3 couldn't keep the same clocks as Zen2 within let's say 15W TDP, It would still perform better thanks to higher IPC unless the clocks were lowered more than what IPC could compensate. I think the reason for Zen2 is that It will have only 4 cores max and because of that Zen3 CCX couldn't be used.

 

[Shivansps]

You are fighting for a market that too few cares about. Gaming on APU/iGPU is low prio for both OEM and most buyers.

And Cezanne will be succesful thanks to powerful CPU, either with discret GPU for gaming interested buyers, or standalone for buyers with little interest in gaming (but it can still handle many games though).

This is the right balance to continue taking market share from Intel in laptops. Wasting silcon on a niche market is not the right balance.

Your opinion/needs doesn't change this.

Is not my opinion, it is a fact that Vega is too old for Cezanne, it will be a +3 year old arch by the time Cezanne launchs and +4 year old by the time it gets replaced, it Rebrandt gets delayed for any reason it may be over 5 years.

if is good enough or not is very subjective, i would rather wait until they are launched to comment on that.

BTW, im not going to keep commenting on this matter here, i was just repleying to someone that said there was nothing wrong with it.

 

[maddie]

Is not my opinion, it is a fact that Vega is too old for Cezanne, it will be a +3 year old arch by the time Cezanne launchs and +4 year old by the time it gets replaced, it Rebrandt gets delayed for any reason it may be over 5 years.

if is good enough or not is very subjective, i would rather wait until they are launched to comment on that.

BTW, im not going to keep commenting on this matter here, i was just repleying to someone that said there was nothing wrong with it.

Guess you never dated younger.

 

[Kocicak]

I have not followed the discussion, so I am not sure if anyone already mentioned it.

AMD did not increase the price by 50 dollars, but by more. 3700x, 3800x and 3900x come with a nice cooler worth cca 30 USD.
So when AMD stops supplying these with the new CPUs, the price is effectivelly 80 USD higher. Almost 100. That is a substantial price increase.

Are nonX models coming?

 

[Kenmitch]

I have not followed the discussion, so I am not sure if anyone already mentioned it.

AMD did not increase the price by 50 dollars, but by more. 3700x, 3800x and 3900x come with a nice cooler worth cca 30 USD.
So when AMD stops supplying these with the new CPUs, the price is effectivelly 80 USD higher. Almost 100. That is a substantial price increase.

Are nonX models coming?

I guess if a person intended on using the stock cooler you could look at it as another negative on pricing. I guess it depends on new build vs upgrade in the end. Upgrading you'll already have a cooler anyways. If building new the stock cooler wasn't really all that great anyways. Depending on one's work load even the better of the stock coolers was somewhat noisy. It really depends on one's tolerance for noise and temps in the end.

 

[Zucker2k]

I guess if a person intended on using the stock cooler you could look at it as another negative on pricing. I guess it depends on new build vs upgrade in the end. Upgrading you'll already have a cooler anyways. If building new the stock cooler wasn't really all that great anyways. Depending on one's work load even the better of the stock coolers was somewhat noisy. It really depends on one's tolerance for noise and temps in the end.

I suspect people are going to need ever more beefier coolers for the coming chips, anyway. Whichever cooler(s) AMD suggests for the reviews is going to be telling. In effect, the XT release would've served the purpose of not only preparing consumers for price increases, but also frees AMD from the burden of having to reinvest in better stock cooling than previous generations. Of course, AMD makes more money by taking away the coolers as well. All these should add up to a significant additional profit for AMD for every chip sold. Zen 3 is going to make AMD a ton of money, and that's vital for their march forward.

 

[therealmongo]

I have not followed the discussion, so I am not sure if anyone already mentioned it.

AMD did not increase the price by 50 dollars, but by more. 3700x, 3800x and 3900x come with a nice cooler worth cca 30 USD.
So when AMD stops supplying these with the new CPUs, the price is effectivelly 80 USD higher. Almost 100. That is a substantial price increase.

Are nonX models coming?

Not for you.

How many CPUs will you RMA this time

 

[Kenmitch]

I suspect people are going to need ever more beefier coolers for the coming chips, anyway. Whichever cooler(s) AMD suggests for the reviews is going to be telling. In effect, the XT release would've served the purpose of not only preparing consumers for price increases, but also frees AMD from the burden of having to reinvest in better stock cooling than previous generations. Of course, AMD makes more money by taking away the coolers as well. All these should add up to a significant additional profit for AMD for every chip sold. Zen 3 is going to make AMD a ton of money, and that's vital for their march forward.

AMD did state that the new offerings require no more additional power then what they're intended to replace. I'd have to assume that cooling requirements will be similar. Of course we'll have to wait for independent reviews for verification.

I'm not much of a stock cooler fan myself. Sure they're better then nothing, but the noise profiles are usually on the unbearable side. When I bought my 3600 I tried the stock cooler and it worked, but I replaced it within a week or so. I don't really give much value to the stock coolers, but they are at least good enough for component testing while a person figures out which cooling solution they wish to purchase.

 

[Kenmitch]

Not for you.

How many CPUs will you RMA this time

I'm going to borrow a generator, take my rig with me, and bin them in the parking lot at Microcenter. Won't take long to swap and do a quick test for a all core 5GHz overclock! /s

 

[inf64]

I made a performance projection for Zen3 based Ryzens and charted it on computerbase graphs (single core, multicore and gaming fps @ 1080p ). I expect ~+-2% deviation from these numbers. I used AMD's claims from their presentation and crosschecked them versus both old gen Ryzen 3000 and intel 10X00 parts, it almost perfectly matches with computer base numbers. I also managed to confirm that 3900X(T) performance from AMD's materials matches almost perfectly the 3900X(T) performance in computerbase reviews. Enjoy

edit: added another chart with arranged Multicore performance for all CPUs; thanks majord!

 

[Markfw]

I made a performance projection for Zen3 based Ryzens and charted it on computerbase graphs (single core, multicore and gaming fps @ 1080p ). I expect ~+-2% deviation from these numbers. I used AMD's claims from their presentation and crosschecked them versus both old gen Ryzen 3000 and intel 10X00 parts, it almost perfectly matches with computer base numbers. I also managed to confirm that 3900X(T) performance from AMD's materials matches almost perfectly the 3900X(T) performance in computerbase reviews. Enjoy

I personally think you have hit the nail on the head. It will be interesting to see how close you come in the next few weeks (whenever the benchmarks all come out)

 

[IntelUser2000]

Do they care that much about the lost money? If they really cared about having the best profit per wafer, then why produce SoCs for the consoles?

Come on. Earning money isn't just about production cost.

Being in consoles, especially for a company with smaller marketshare and presence as AMD gets benefits elsewhere.

Like being the common architecture for a mass market gaming system. Or that everyone gets to know its AMD that powers them.

 

[maddie]

Come on. Earning money isn't just about production cost.

Being in consoles, especially for a company with smaller marketshare and presence as AMD gets benefits elsewhere.

Like being the common architecture for a mass market gaming system. Or that everyone gets to know its AMD that powers them.

What is not always appreciated, is AMD's ability to pull business to TSMC, who is investing a lot of revenue into CapEx. Both companies are obviously in this for the long haul, so having a couple quarters not fully exploited but still great, is not a problem to savvy investors. AMD intends to win big and for this to happen, TSMC must also.

 

[A///]

Zen 4 will be on a completely new platform, so the IOD will be redesigned as well, possibly even split up into smaller parts, some of which may stay at GloFo and others that profit of smaller nodes.

Layered IOD? There was speculation by @jamescox that the IOD and the chiplets would be layered to increase core counts without a bigger overall package.

 

[Panino Manino]

I also used to be "mad" with AMD because of Vega, but seeing how they decreased the size of the GPU on Renoir with no real detriment to performance I doubt that using RDNA1 would make much difference. There are two critical points for me:
One is that like other said, the CUs are larger? So this would mean a limited number of CUs, nullifying the the better memory efficiency?
But most important, would be wasted effort. Why expend time and money to put RDNA1 on an APU instead of waiting for RDNA2 that will allow them to also brag about having hardware-RT support and with faster DDR5 with that so necessary bandwidth increase?

In the end AMD is more concerned with the long run than looking good now.

 

[therealmongo]

I'm going to borrow a generator, take my rig with me, and bin them in the parking lot at Microcenter. Won't take long to swap and do a quick test for a all core 5GHz overclock! /s

Thats actually not a bad idea, probably best to do it outside some LAN gaming venue and sell at a markup

 

[majord]

I made a performance projection for Zen3 based Ryzens and charted it on computerbase graphs (single core, multicore and gaming fps @ 1080p ). I expect ~+-2% deviation from these numbers. I used AMD's claims from their presentation and crosschecked them versus both old gen Ryzen 3000 and intel 10X00 parts, it almost perfectly matches with computer base numbers. I also managed to confirm that 3900X(T) performance from AMD's materials matches almost perfectly the 3900X(T) performance in computerbase reviews. Enjoy

Nice job!
Looks brutal When you put it amongst the Pigeons like that , also starts to feel like a bit better value proposition.

Any chance you could move the MT bars into Performance order?