Info Ryzen 4000 Mobile Chips Unveiled at CES

[guachi]

Acer Swift 3 with Ryzen 7 4700U
AMD Ryzen 4000 Mobile APUs

AMD debuted (finally!) their Ryzen 4000 mobile chips at CES that are confusingly Ryzen 2 CPUs that are analogous with their Ryzen 3000 desktop parts.

I like what I see, especially on the power front. I know many people said that 7nm Ryzen chips had the potential to be very power efficient and if AMD's slide deck is to be believed they have succeeded. Most of the power efficiency has come from the 7nm process. As well, the 7nm process looks to be allowing AMD to cram up to 8 cores onto a laptop chip.

Do you guys think AMD has a product that will be as competitive on the laptop as the 3000 series is on the desktop? I'm thinking that the 4600 will be the best buy like the 3600 is in the desktop space. The problem in the laptop space is AMD needs design wins. At least on the desktop I don't need some company to choose for me, I can just by the chip myself.

 

[majord]

HP EliteBook 845 G7 Notebook PC Performance Results - UserBenchmark

4800U looks like yet another ehh IGP, but attached to an awesome CPU.

We still trust UserBenchmark?

Is only running ddr4 in that link?

 

[Shivansps]

Did you stop buying Intel when they used cache, AVX, and HT for segmentation purposes?

I had one intel CPU in my entire life, a 2500K due to BD being a dissaster. When you see me defending such things quote me. Thanks.

And you are right about being unannounced, but we know what the Renoir die is, im not sure what else you expect them to do. They either segment Vega or they dont. The reason behind not being a 8CU 45W Renoir is due to igp being useless there, as those Renoirs are meant to use dgpus and die harvesting. They did it right there, the entire mobile line is fine.

What in the world is happening here.

Stock overclocked? seriously? Because a 7nm die can clock faster than a 12nm one, that means it's overclocked?

AMD has a history of clocking stuff beyond its ideal point to make it look better at launch. Vega and Polaris 10 are well know examples of dgpus that should have come out with a little less of stock vcore and clocks. The entire FX cpu line is another huge example.

Using more power overclocked and stressing components? It's a different chip in the same power range. Why would you even make such a case?

Did you use a APU as your main PC for a while? Do you know that there are B boards with chipset VRMs that reach 90°C on stock? So yeah when i hear 1700mhz at stock may be a thing and 7mm dosent do miracles i cant help but wonder how much you can actually push it.

Lastly, when have we never had segmentation?

Never, dosent means i have to like it or defend it. Specially when they are taking away stuff.

AFAIK, AMD isn't selling CUs but performance, but what do I know. If you don't think it's worth the price, then don't buy it, simple.

What other option are there? This is what monopoly causes.


As there is no point in keep arguing this im going to wait until they announce the desktop skus, at least there we can know for sure what they are going to do..

 

[Thunder 57]

I had one intel CPU in my entire life, a 2500K due to BD being a dissaster. When you see me defending such things quote me. Thanks.

And you are right about being unannounced, but we know what the Renoir die is, im not sure what else you expect them to do. They either segment Vega or they dont. The reason behind not being a 8CU 45W Renoir is due to igp being useless there, as those Renoirs are meant to use dgpus and die harvesting. They did it right there, the entire mobile line is fine.

And I went with a 3570k for the same reason. I didn't like how they cut cache/threads. Back then it didn't matter as much though, at least for what I was using it for. Segmentation sucks, but like you said, "This is what a monopoly causes".

AMD has a history of clocking stuff beyond its ideal point to make it look better at launch. Vega and Polaris 10 are well know examples of dgpus that should have come out with a little less of stock vcore and clocks. The entire FX cpu line is another huge example.

I think the problem here was too much voltage rather than clocking too high. At least with the GPU's. It's well known that most AMD GPU's can be undervolted and oftentimes perform better. They even offer an "undervolt" option in the driver now. My guess is that was because they didn't have as much resources as they wanted for validation and were in a rush to get the parts out. Hopefully the GPU side is back on equal footing now that AMD has money again. The FX series though; Yea, that was pretty bad.

Did you use a APU as your main PC for a while? Do you know that there are B boards with chipset VRMs that reach 90°C on stock? So yeah when i hear 1700mhz at stock may be a thing and 7mm dosent do miracles i cant help but wonder how much you can actually push it.

We'll see when it comes out. Hopefully boards will support it and VRM's will be OK. If not, then those cheapo boards just won't be updated to handle the new APU's. No point in getting upset until it gets here.

 

[Shivansps]

it is just that im having a hard time beliving that these new APU will provide a good improvement over the old ones outside notebooks(we all said it back them, why just 2400mhz on mobile Picasso when the desktop ones supports 2933? i think it is more clear now the reason of why). AMD said so themselves, they hit the 30% improvement they wanted on notebooks, there was no reason for more. If they can produce a 4200G that performs 30% faster than the 3200G like they are doing with the notebooks APUs im not going to complain, im not blind. But im not going to ask that, im going to say it now, 3400G perf on the 4200G that i what i would consider a resonable improvement. For that they would need to match a 11CU @ 1400mhz GPU, with just 5-6CU, you need like what 2Ghz? can someone do the math here? I can wait for reviews but it just seems impossible to me.

 

[IntelUser2000]

11CU @ 1400mhz GPU, with just 5-6CU, you need like what 2Ghz? can someone do the math here? I can wait for reviews but it just seems impossible to me.

You need 2.5GHz for 6CU and 3GHz for 5CU if you assume it scales linearly, but it doesn't.

Because if you boost clocks it also boosts other parts of the CPU such as the graphics command processor, ACEs, and the HWS.

But what makes you think they'll cut it from the H parts?

If you consider all that an 8CU version might be enough should they "evaluate" that they want to make that. Sure, its a disappointment that the GPU doesn't improve, but at least it has a chance it won't regress.

Plus I don't think it'll regress, because it has a faster CPU. At playable frame rates, CPU impacts it a decent amount.

 

[jpiniero]

it is just that im having a hard time beliving that these new APU will provide a good improvement over the old ones outside notebooks(we all said it back them, why just 2400mhz on mobile Picasso when the desktop ones supports 2933? i think it is more clear now the reason of why). AMD said so themselves, they hit the 30% improvement they wanted on notebooks, there was no reason for more. If they can produce a 4200G that performs 30% faster than the 3200G like they are doing with the notebooks APUs im not going to complain, im not blind. But im not going to ask that, im going to say it now, 3400G perf on the 4200G that i what i would consider a resonable improvement. For that they would need to match a 11CU @ 1400mhz GPU, with just 5-6CU, you need like what 2Ghz? can someone do the math here? I can wait for reviews but it just seems impossible to me.

Well, yeah. They want you to buy a dGPU. Plus they didn't want to make Renoir too big because of how expensive the 7 nm wafers are, and prioritized more CPU cores over the IGP.

 

[lobz]

Well, yeah. They (...) prioritized more CPU cores over the IGP.

Thank God!

 

[Shivansps]

Thank God!

The whole idea behind an APU is that should be balanced, if you want a 8C APU just get a CPU and a dGPU. There is zero point in a >$300 APU with slower CPU perf than similar priced CPU and the IGP perf of a < $100 dgpu that needs a >$150 mb and takes away up to 2GB of your ram while in use and has reduced conectivity.
This is true on notebooks as well, there is really a market for a 4800U notebook? I would rather have a 4600H+dGPU even if that means lower battery life.

 

[Topweasel]

The whole idea behind an APU is that should be balanced, if you want a 8C APU just get a CPU and a dGPU. This is true on notebooks as well.

Not if you want to drive up performance of thin and lights. The crown of Renoir isn't the 4800H, there isn't much of a reason to get that over a 9980HK outside of maybe price. I mean in the end it might end up being a better chip, but the fact is once the shackles are off the performance will be a lot closer. What Renoir does is give them what intel can't, performance driven thin and lights. You can't do that if AMD said based on the niche and low margin requests of someone like yourself, meant they should should redesign everything about Zen and it's CCX design for a fully unique configuration, just so they could go with 6 cores and put more CU's in.

 

[eek2121]

Not if you want to drive up performance of thin and lights. The crown of Renoir isn't the 4800H, there isn't much of a reason to get that over a 9980HK outside of maybe price. I mean in the end it might end up being a better chip, but the fact is once the shackles are off the performance will be a lot closer. What Renoir does is give them what intel can't, performance driven thin and lights. You can't do that if AMD said based on the niche and low margin requests of someone like yourself, meant they should should redesign everything about Zen and it's CCX design for a fully unique configuration, just so they could go with 6 cores and put more CU's in.

You cannot automatically assume that. There is quite a bit we don't know about the 4800H. What we DO know is that the 4800H will beat the 9980HK in multi-threaded performance. We also know that the 65 watt 3700X can compete with the desktop versions of the 9900K and 9900KS (which consumes more than 200 watts at 5 GHz). The things we don't know are:

1. What role SmartShift will play in increasing performance.
2. What the LPDDR4X speeds will be, and how that will factor into the performance of Renoir.
3. How much of a performance improvement we'll see from the IO die being integrated and shrunk.

I'm going to venture an educated guess and say that Renoir is faster. The 9980KS may technically boost to 5 GHz, but if it does do so, it will likely be outside it's rated TDP. While this isn't an issue on the desktop, it's a HUGE issue in the laptop world.

 

[IntelUser2000]

You cannot automatically assume that. There is quite a bit we don't know about the 4800H. What we DO know is that the 4800H will beat the 9980HK in multi-threaded performance.

Actually they are about equal.

 

[beginner99]

Well, yeah. They want you to buy a dGPU. Plus they didn't want to make Renoir too big because of how expensive the 7 nm wafers are, and prioritized more CPU cores over the IGP.

Yeah and it was finally a sane decsion by AMD compared to their previosu APUs with anemic CPU and way to much GPU compared to it. If you want to play games you need a dGPU anyway. I don't see the reason for the beefy iGPUs really. It's a waste of die space that could be used for more cores or some fixed function stuff like in smartphones.

 

[krumme]

15w tdp is what matters.
Not desktops there is old cheap 12nm to cover that lowend.
In the real world 3200 MHz will be the practical limit as 2400 is today. Dual on a good day.
What sells today is the 3500u not the 3700u. Its not fair or realistic to compare to those high cu. So what sells is 8cu, supported by at best 2x2400.
4900u will come and i think in the same price segments as 3500u covers now, the 4700u will replace it. It will be 7cu with a 59% uplift vs 8cu in the 3500u, mated to probably 2x3200. Seems like a fine solution to me and a solid uplift in real world gaming. It will be a bit more expensive and 3500u will go a tier lower. Having more cu wouldnt make sense at all at 15w tdp. Heck It will even use 2x3600 ram fine.
Having say 20-30% more cu for 2x3600 makes sense when navi arives. I assume we are not in ddr5 then?

 

[amd6502]

OMG give it a break already. Please. The world does not revolve around desktop budget APUs that are marked down as low margin products.

Especially around folks in some odd niche who are looking to max out the iGPU from a low margin discounted APU with some very expensive non-mainstream high frequency DDR4 (.... who then go further and try to OC this low budget APU's iGPU).

Renoir is targeting high-end mobile. That is high end laptops that either don't target a gaming audience or that come with a dGPU (and only use the the iGPU for unplugged gaming).

There will always be at least some minimal segmentation (due to die salvage and binning strategy).

I think their die salvage team is already saving the bulk of desktop worthy silicon having 8 well clocking CUs for desktop (hence the 7CU silicon going to the 45w BGA dgpu-gaming laptops and OEM-bga desktops).

What you get is most likely 8c/16t AM4 parts all having 8CU.

8c/8t part probably with 8CU (clocked at a moderate freq).

6c/12t part with 6CU

4c/8t may be so rare it only makes it to OEM, and these might be 3CU and not intended as gaming APU. So budget iGPU gamers will probably find the 3400g hard to beat.

I hope AMD comes out with a budget rebin of the 3400g. Something a bit discounted from the 3400g, with the same cores and CU as the 3400g, but with milder CPU frequencies. 3390g or something

 

[Gideon]

Actually they are about equal.

In bursty ST loads i9-9980HK will surely win because of it's clock-speed.
In bursty MT benchmarks I can imagine them being equal, if cooling is adequate.
In sustained MT workload? I find it hard to believe (and 500 MHz base-clock diff seems to suggest that).

The perf/watt advantage 7nm has in the all core turbo range (3.0 - 3.5Ghz) seems to big for that to be true. Let's not forget that Matisse has a separate I/O die and 4x the L3 cache, while (i9-9980HK is essentially a mobile-binned 9900K). If anything, Renoir should perform a bit better than Matisse in this metric.

 

[lobz]

Actually they are about equal.

Which would still be great, considering how much less the 4800H will (well, probably) cost to the laptop vendors. BUT as soon as boosting is over, 2.9 vs 2.4 GHz is a lot, even when it's Zen2 with much less cache - but hopefully better mem latency.

 

[Shivansps]

OMG give it a break already. Please. The world does not revolve around desktop budget APUs that are marked down as low margin products.

Well i said that what AMD is doing is good for notebooks but bad for desktop. And i dont agree about desktop not being important for the consumers, the reason of why is not important for AMD is that Intel iGPU is trash and so they have full monopoly no matter what, is not the same as not being important.

Simple as that, you think notebook is important, and i think both are important, no point in arguing that.

 

[amd6502]

Well it's not really a monopoly, because there are alternatives, like intel APUs and the new KX series from China. The KX may not really be retailed (in much of the world outside of asia) but I expect at least used systems to make their way to ebay.

Consumers also have the choice of getting a low end or used graphics card.

Lastly there are also non-x86 choices, with pi boards and acorn boads with higher than pi specs being popularly supported with linux OSes.

I've mentioned this before, but I believe an MCM project is a much better candidate for desktop APUs, especially big ones. Here they could have a new line by giving the consumer a choice between a vanilla IO hub and one with an iGPU. Here they might go with a choice of iGPU that can just about max out even very pricy high frequency DDR4 (somewhere along 10 to 12 CU or the Navi equivalent).

I believe another reason besides reducing die area for cost is that they wanted the die to be small enough for the same mini-BGA socket that Stoney (125mm2) fits in. This way the also would fit into some very compact laptops and even 2-in-1 tablets (I think a die salvaged single CCX 8 thread Renoir would be perfect for a high end 2-in-1).

 

[IntelUser2000]

Well i said that what AMD is doing is good for notebooks but bad for desktop. And i dont agree about desktop not being important for the consumers, the reason of why is not important for AMD is that Intel iGPU is trash and so they have full monopoly no matter what, is not the same as not being important.

They are clearly aware of this, and using it to boost profits. They have no competition in the iGPU space.

BUT as soon as boosting is over, 2.9 vs 2.4 GHz is a lot, even when it's Zen2 with much less cache - but hopefully better mem latency.

The boost isn't immediately over as the thermal headroom ends. The cost of running Cinebench is different from say, loading a text document.

The 9980HK is an unlocked part and they can increase TDP if they choose to. The DTRs are running at 55W, or even 65W for PL1. Of course that's the caveat for 9980HK - thermals.

 

[Shivansps]

I've mentioned this before, but I believe an MCM project is a much better candidate for desktop APUs, especially big ones. Here they could have a new line by giving the consumer a choice between a vanilla IO hub and one with an iGPU. Here they might go with a choice of iGPU that can just about max out even very pricy high frequency DDR4 (somewhere along 10 to 12 CU or the Navi equivalent).

High cost DDR4 is also pointless for a APU, but DDR4-3200 is not high cost ram... around $10 more compared to 2400/2666 for 16GB. And the 3400G plays along just fine until the 1600 gpu clock on DDR4-3200 so there is room to improve a little.

Anyway, do you think Navi APU will be still a DDR4 product?

 

[RetroZombie]

I see much more product longevity (for the user and the manufacturer) with 8 cores + low end gpu than one with 4 cores + super duper gpu.
I don't consider the vega 3.0 in renoir low end by the way.

 

[Topweasel]

You cannot automatically assume that. There is quite a bit we don't know about the 4800H. What we DO know is that the 4800H will beat the 9980HK in multi-threaded performance. We also know that the 65 watt 3700X can compete with the desktop versions of the 9900K and 9900KS (which consumes more than 200 watts at 5 GHz). The things we don't know are:

1. What role SmartShift will play in increasing performance.
2. What the LPDDR4X speeds will be, and how that will factor into the performance of Renoir.
3. How much of a performance improvement we'll see from the IO die being integrated and shrunk.

I'm going to venture an educated guess and say that Renoir is faster. The 9980KS may technically boost to 5 GHz, but if it does do so, it will likely be outside it's rated TDP. While this isn't an issue on the desktop, it's a HUGE issue in the laptop world.

Ehh. I wasn't saying it as in a number crunching way. The 9980HK fills the role of the 4800H well enough, has been out longer, and exists on a platform that 90% of their sales in the market come from. You aren't going to get to many design wins there because the 4800H isn't going to be enough "better" to make it worth it, specially when it comes to AMD and just being the better product doesn't help it sell. What you need is a killer feature, that's the 4800U. Now all of a sudden you have something that just can't be matched by the competitor. Not on CPU or GPU. It's just plainly a better product, so much so, you would be stupid not to offer it in your lineup. With it then comes models with 4700u's 4600u's 4500u's 4300u's. See the trend. The 4800U not only is something everyone must offer at some point but its the gateway to sell more of their other chips as well.

 

[IntelUser2000]

High cost DDR4 is also pointless for a APU, but DDR4-3200 is not high cost ram... around $10 more compared to 2400/2666 for 16GB. And the 3400G plays along just fine until the 1600 gpu clock on DDR4-3200 so there is room to improve a little.

Especially nowadays, when I could buy a used RX 470 for $90 cdn(that's 70 for you american folks) and play games at much higher quality settings, its not worth it.

The real crucial feature for a desktop iGPU is having support for the latest media playback. The better 3D performance is only there for competitive reasons(Intel vs AMD) than anything else.

You and I can justify spending $30 extra on the board and the $10 extra for the RAM. It's because we see it as a form of entertainment. But most people won't, and this is why official speeds are king.

So can I buy a motherboard for the 3400G that costs $60 and also supports DDR4-3200?

 

[Shivansps]

So can I buy a motherboard for the 3400G that costs $60 and also supports DDR4-3200?

Any board will work just fine, even the cheapest A320 supports DDR4-3200 (and ram overcloking). I would not recomend some of the cheapest asrocks but thats about it.

Most of the combinations i sell over here are:
A320/3200G/2x8 3200 (Crucial Black)
B450M-DS3H/3400G 2x8 3200 (Crucial Black)

Those are the top sold combos/pcs here.
Followed with 8GB 2666 combos with A320 for both the 3200G and 3400G.

 

[Topweasel]

I see much more product longevity (for the user and the manufacturer) with 8 cores + low end gpu than one with 4 cores + super duper gpu.
I don't consider the vega 3.0 in renoir low end by the way.

Isn't that last part the key. All of this complaining about loss of CU's . But we haven't even seen anything. It could be faster, or slightly faster. But it doesn't matter here because the number of the thing no one but one person is actually watching noticed went down. CU count should be per architecture comparison. We know these CU's are faster so why count them before you know the actual performance numbers.