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.

 

[Roland00Address]

Power Efficiency numbers mean squat for that is just putting 8 cores in the same TDP as 4 cores, doing more work at max load is not the relevant information (a form of power efficency at max load.)

I want to know their idle power draw. I crave this knowledge!

[The surface AMD vs surface Intel comparison was very disappointing.]

 

[krumme]

Power Efficiency numbers mean squat for that is just putting 8 cores in the same TDP as 4 cores, doing similar power efficency at max load.

I want to know their idle power draw. I crave this knowledge!

[The surface AMD vs surface Intel comparison was very disappointing.]

Wrong. I have 8250u and 3500u with same battery capacity in Lenovo thinkpad and for office and web surfing performs about the same for batterylife.
My experience mirrors Notebookcheck testing:

"Battery Life

A major weakness of the old ThinkPad E485 was the battery life, which was significantly shorter than the Intel models. However, that changes with the new E495, because with the integrated 45 Wh battery, the AMD ThinkPad achieves realistic runtimes (WLAN, video) of around 8 hours. In the Wi-Fi test, it is thus on the level of the ThinkPad E490. Only in the video test was the Intel ThinkPad somewhat more endurable.

"

Lenovo ThinkPad E495 Laptop Review: Inexpensive office device with a lot of power, but without keyboard illumination

Notebookcheck is testing the Lenovo ThinkPad E495 with AMD Ryzen 5 3500U processor. The performance is very good, but unfortunately Lenovo dispenses with a backlit keyboard.

www.notebookcheck.net

There is only minor differences imo. What matters more is screen efficiency, software setup, idle programs running so on.

AT is simply wrong in their testing methology. Yes there is 1w or so difference at idle situations but in the grander sceme it means little.

Picasso is not Raven ridge. It needs to be put to rest. Yes Icelake is better. Yes Renoir is better but its also different higher priced products.

 

[tamz_msc]

AT is simply wrong in their testing methology.

Just because NotebookCheck's result is different from AT doesn't mean that AT's methodology is wrong. You can't compare reviews for laptops across different websites.

 

[krumme]

What i notice:

AMD is starting to segment at bit more as if they have monopoly. Eg like Intel 3 years ago. Thats what lack of compettition looks like. People need to opose that crap before they start behaving like Intel and nv.

Lots of disabling of smt. Do we not need efficiency on mobile? No thank you amd.!

Ryzen 9 not released - do they even have to? Shareholders surely glad but thanx for nothing !

4c renoir apu. Really nessesary? At such a small die? 6c cut gfx is plenty to get near 100% yield.

Amd is kicking Intel to death all over. Renoir will be far superior and the artificial hamstrung of the apu shows it. Intel needs to change top management and get other culture. Whatever. We need some compettition now or at least in 4 years. Not some paper product produced in tiny numbers. I cant endure 6 bulldozer years again.

 

[krumme]

Just because NotebookCheck's result is different from AT doesn't mean that AT's methodology is wrong. You can't compare reviews for laptops across different websites.

Agree we cant compare, but imo Notebookcheck results is relevant for most consumers:

AT testing methology is made to show differences and highlight technical design differences.
Notebook check methology is made to meassure practical batterylife surfing and video.

Notebookcheck imo mirrors practical usage more. They do it for a living.

Picasso is perfectly okey. I think part of this trashing is because first gen apu was very disappointing in battery life and it hits picasso.
Secondly we have to balance this stuff. Haswell was a huge leap for efficiency, and imo solved most of the issues. What we want in a notebook is 8 hrs usage. Including brakes in a day thats plenty for all except Bill Gates

 

[krumme]

Renoir will look stellar in reviews.
Its st perf will be on par, and its mt and efficiency at load will be far superior. Gpu will be about 20-60% better than comparable Intel apu. So mt cpu and gpu perf uplift vs Intel will be about similar.

Now amd aparently decides to make a 10cu part and not offer it yet, but I think the interesting part here is the relative downsizing of gpu area vs picasso.

I think the irony is it will look good in reviews but as a single apu solution for consumers - not business - imo picasso is far better balanced cpu-gpu.

Now for consumers this apu is a beast and made to be paired with discrete gfx - hello 5600 - and the designwins shows it. All untill now is with discrete gpu!?

And for business this apu have plenty of gpu power and hopefully does close the battery life difference to icelake even more.

It looks to me it was a darn smart move to have a small vega gpu and go for the high fmax learned from prior vega 7nm to edge out Intel. Its smaller than i anticipated, and it enables it to go firmly into midrange.

The result was a fairly small ~150mm2 that can win in all market for mid and high end. The high and midrange market was just turned upside down.

 

[moinmoin]

AMD is starting to segment at bit more as if they have monopoly. Eg like Intel 3 years ago. Thats what lack of compettition looks like. People need to opose that crap before they start behaving like Intel and nv.

You didn't mention the low end 2c2t Athlon Silver SKU which may be also a salvaged/fused off Renoir die (or Raven2, dunno).
I don't think this is down to monopolistic behavior, rather platform building, allowing ODM to create one single board and reuse that for a whole range of devices, low end to high end (like they are already used to do for Intel devices).

 

[coercitiv]

Wrong. I have 8250u and 3500u with same battery capacity in Lenovo thinkpad and for office and web surfing performs about the same for batterylife. [...]

AT is simply wrong in their testing methology. Yes there is 1w or so difference at idle situations but in the grander sceme it means little.

AT compared a 10nm Intel CPU /w LPDDR4X vs. a 12nm AMD CPU /w DDR4. You compared a 14nm Intel CPU /w DDR4 vs. 12nm AMD CPU /w DDR4, so it's only natural that results look different, in the AT review both the CPU and memory subsystem had a clear power/process advantage on the Intel platform.

This time around AMD has a 7nm CPU running LPDDR4X, so we're finally heading for a true apples to apples comparison.

[Edit] wrong process name for Intel

 

[Gideon]

You didn't mention the low end 2c2t Athlon Silver SKU which may be also a salvaged/fused off Renoir die (or Raven2, dunno).
I don't think this is down to monopolistic behavior, rather platform building, allowing ODM to create one single board and reuse that for a whole range of devices, low end to high end (like they are already used to do for Intel devices).

It would make no sense to gut a 8 core chip that much. Athlon Silver and Gold are the 2-core Raven Ridge die based almost certainly. They 100% aren't Renoir as they have the FP5 package (that 3500u has) rather than FP6 (that renoir has).

Straigth from AMD website:
Athlon Gold 3150 (FP5)
Athlon Silver 3050 (FP5)
(god I hate that they stole the stupid Intel naming convention for pentiums)

 

[Gideon]

Overall Renoir looks like an excellent CPU. The biggest elephant in the room being Idle-Power. A few things were unexpected and slightly disappointing.

Having only 8CUs is a very unexpected approach (I thought they'd have at least 12). I can see why they did this. According to Ian Cutress the die is about 150mm2 (thats abotu same as Tiger Lake!) So the GPU choice indeed seems to be a compromise in order to get the CPU size down. It's a huge tradeoff for some, but overall a quite beautiful one as the H series were getting discreet GPUs anyway as are 99% of all of the non-business Destop CPUs (and the business ones don't really need more GPU power)

What I'm less pleased about is the segmentation, particularly how SMT is handled.

I get the need of making the lowest-spec 4C/4T 4300U (as to not gut the top line too much), 6C/6T 4500U already seems a bit iffy but ok, it's still a budget CPU. But removing SMT from a Ryzen 7 4700U really is uncalled for. That really smells too much like Intel's segmentation to my taste. Having a slower GPU and maybe lower all-core boost by an extra 100 Mhz would have been good enough.

As it stands now, 4700U really sits in an odd place. IMO 4600U with it's 12 threads is a much better CPU (considering that SMT scales better on AMD than Intel, I wouldn't be surprised if it beats 4700U in quite a few MT benchmarks).

 

[itsmydamnation]

Perf per watt of 6/12 could very well be lower then 8/8 and thus allowing the 4700 to hold higher clocks in high thread situations.

 

[beginner99]

Looks interesting but as all others have said, we need to see battery life especially for idle/low usage scenarios like sitting in meetings or a conference. Also will be interesting to see the effect of the halved L3 on performance.

 

[krumme]

AT compared a 10nm Intel CPU /w LPDDR4X vs. a 12nm AMD CPU /w DDR4. You compared a 14nm Intel CPU /w DDR4 vs. 12nm AMD CPU /w DDR4, so it's only natural that results look different, in the AT review both the CPU and memory subsystem had a clear power/process advantage on the Intel platform.

This time around AMD has a 7nm CPU running LPDDR4X, so we're finally heading for a true apples to apples comparison.

[Edit] wrong process name for Intel

Agree. What i have been trying to say in the surface thread this comparison AT did was meaningless. Like comparing Renoir lpddr4x to picasso. Yet people keep harping about how bad Picasso battery life is. I just tried to put it into perspective.
Picasso should be compared to similar priced products. It means quad kbl/whatever they call it.
Sure for business usage why not take the 1w idle benefit for kbl as you dont need the gpu - at similar price or so.
But same for consumers. If you just ocationally game a little bit there is zero reason to go kbl vs picasso. All at similar price.

Now all this is low end. Renoir just took mid and high end. Surely for consumers but probably also business. We will see. I think the biggest hindering will be huge demand and high prices.

 

[krumme]

People went crazy and bought all 3900x and 3950x. The demand for highest end products is always surpricing.
In a gaming or highload productivity laptop efficiency is king to consistent high performance and we are more tdp constrained.
7nm will rule even more in mobile.
Together with traditional highend frenzy I think the 4800h and 5600m will make huge strides in gaming laptops ~100-140w tdp, so actually portable. And be the benchmark to beat. The designwins indicates it.
And for business. Man, a 8c/16t in 15w tdp. Wow.
I am sure people will line up for this. Get ready to grab one before they are gone! Lol

 

[coercitiv]

Now all this is low end. Renoir just took mid and high end. Surely for consumers but probably also business. We will see. I think the biggest hindering will be huge demand and high prices.

I think we'll be surprised by the time 2020 truly unfolds.

All the bullet-points on the 4000 series complain list help yields. They have a winning product as is, all they need to do now is churn them out like crazy, especially the 6c/12t product which will probably be the real hero of this story.

 

[Topweasel]

Agree. What i have been trying to say in the surface thread this comparison AT did was meaningless. Like comparing Renoir lpddr4x to picasso. Yet people keep harping about how bad Picasso battery life is. I just tried to put it into perspective.
Picasso should be compared to similar priced products. It means quad kbl/whatever they call it.
Sure for business usage why not take the 1w idle benefit for kbl as you dont need the gpu - at similar price or so.
But same for consumers. If you just ocationally game a little bit there is zero reason to go kbl vs picasso. All at similar price.

Now all this is low end. Renoir just took mid and high end. Surely for consumers but probably also business. We will see. I think the biggest hindering will be huge demand and high prices.

I would agree with you if we were talking about 35 or 45w sku's. But we are talking about 15w sku's (even if a lot got set to 25w), sku's. Now what you can get done in 15w has changed a lot in last 2 years. But the end point is you are using a 15w CPU for either thermal or battery life reasons. So when after everything is said and done Picasso uses like twice as much power in low load situations, that is going to be 80% of the use case in with these systems, its not going to be a very good option. If this was a question of which CPU goes in my HDET, the Ryzen would win. But if the question is what CPU would go in lets say a Surface Pro I was getting and in light usuage I had one that could get me 7 hours and the other 5, I would go with the 7, even if that means choosing an Intel based system. So I get the desire to see Renoir fix that. Problem is that that IF has a power penalty compared the ring bus and even if the cores their selves is more efficient the mesh makes getting idle power down harder. I wonder and am guessing that AMD worked on that more here (vs 3 gen of basically the same Raven Ridge design).

 

[coercitiv]

Problem is that that IF has a power penalty compared the ring bus and even if the cores their selves is more efficient the mesh makes getting idle power down harder. I wonder and am guessing that AMD worked on that more here (vs 3 gen of basically the same Raven Ridge design).

They are working on it.

These CPUs all support LPDDR4X memory, up to 64 GB, and AMD says that the infinity fabric is not tied to this memory clock. This helps the chip reach even lower power in its idle states, and the company said that they have rearchitected a good portion of the power delivery in the APU in order to be able to power down and power gate more elements of the SoC than was previously possible. AMD said that this decoupling of the infinity fabric and memory support, especially with both CPU and GPU accessing it, was made substantially easier due to the APU being a monolithic solution (with that in mind, it’s likely that AMD might not be going down the chiplet APU route any time soon). Also worthy to note is that AMD is saying that they have reduced the latency for parts of the chip to enter/exit idle states by 80%, and it’s this that helps enable the power gating in such a way to remain responsive. In previous products, certain elements of the design had to remain powered in order to be as responsive as the user required.

Renoir gets more power gating, more aggressive power gating, lower overall power due to better process and lower memory power usage due to LPDDR. All of these will bring down idle power. They may even get close enough for a practical tie with Intel.

If Lenovo isn't mistaken/lying about their Yoga Slim battery life estimates, then Renoir can match ICL in similarly built systems. We need to see actual test data though.

 

[krumme]

I would agree with you if we were talking about 35 or 45w sku's. But we are talking about 15w sku's (even if a lot got set to 25w), sku's. Now what you can get done in 15w has changed a lot in last 2 years. But the end point is you are using a 15w CPU for either thermal or battery life reasons. So when after everything is said and done Picasso uses like twice as much power in low load situations, that is going to be 80% of the use case in with these systems, its not going to be a very good option. If this was a question of which CPU goes in my HDET, the Ryzen would win. But if the question is what CPU would go in lets say a Surface Pro I was getting and in light usuage I had one that could get me 7 hours and the other 5, I would go with the 7, even if that means choosing an Intel based system. So I get the desire to see Renoir fix that. Problem is that that IF has a power penalty compared the ring bus and even if the cores their selves is more efficient the mesh makes getting idle power down harder. I wonder and am guessing that AMD worked on that more here (vs 3 gen of basically the same Raven Ridge design).

Agree with your fundamentals. But the numbers is wrong. Look at the notebook test i linked. You get close to 8 hrs on picasso in the wifi test for a 45whr battery. The difference vs kbl on idle is close to 1w. With a minor load for wifi test their results is similar. I have theese machines and unless you are gaming its simply impossible to tell them apart. I dont watch video though and kbl have a solid advantage here.

As you I think its impossible for renoir to beat icelake for idle numbers. Ringbus. But if it closes some of the distance, have a more efficient videodecoder, and is superior under load imo i will think baterylife is a non issue for consumers as well as business.

What i think matters then is eg noise under load and here the 7nm zen2 will probably be better. Not to mention the size of the chassis needed for similar tdp.

 

[majord]

What a killer looking chip - Such a shame Navi couldn't make it on board, but Pre Tigerlake, that's not going to be a hindrance competition wise clearly.. I can see why they took this path if they're getting these clocks and subsequent performance with only 8 CU's

and with that said, I wonder if there's actually more CU's on die - pushing for maximum volume with 8CU's for now (given it's more than enough to compete for now) , Release fully enabled on 7+ refresh to counter TGL?

 

[Gideon]

Problem is that that IF has a power penalty compared the ring bus and even if the cores their selves is more efficient

At least now Infinity fabric is decoupled from memory speed, meaning AMD can power down some parts of the chip 5x faster (AMDs words). This with 20% less SoC power draw and LPDDR4X will definitely improve things. If quite up to ice-lake levels ... I'm a bit sceptical (otherwise they sure would have shown it) but it should certainly be more than 20% improvement

 

[PingSpike]

It'll be interesting to see what happens with idle power. As others have said, its been a weak point of the IF design. If there is a way to solve it, I wonder how the solution could exist here but not on the existing desktop parts?

 

[beginner99]

and with that said, I wonder if there's actually more CU's on die - pushing for maximum volume with 8CU's for now

I doubt it because they also lower CU count with core count. So a defective cpu core or defective CU both can be used in a lower binned sku. 3 defect CUs? make it a 4300U even if all CPU cores work. The chip itself and the skus tells us they already are optimized for maximum yield.

Also I like that they limited the CU count. No need to use too much die space and power for iGPU. If you really need it, you will need a dgpu anyway.

 

[beginner99]

I wonder how the solution could exist here but not on the existing desktop parts?

The article actually mentions it: decoupling IF from memory clock and appliyng more power gating is easier in a monolithic design than with chiplets.

 

[coercitiv]

A small observation about platform power potential that isn't even directly tied to SoC arch/process:

Memory I/O voltage:
DDR4 ......... 1.2V
LPDDR4X ... 0.6V