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.

 

[Shivansps]

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.

We do not know if these CU are faster until someone does a 1:1 comparison. As far as we know is the same gen. AMD was talking about being faster in those 15W TDP, due to higher boost and DDR4-3200 instead of 2400.

 

[Topweasel]

We do not know if these CU are faster until someone does a 1:1 comparison. As far as we know is the same gen. AMD was talking about being faster in those 15W TDP, due to higher boost and DDR4-3200 instead of 2400.

Point is neither do you. I am not worried, nor do I think AMD is terribly worried about the nitty gritty of the performance of the video part. This CPU exists to finally have a Laptop CPU that hands down is a great laptop CPU. Continuing to pretend that iGPU graphics is all you really need hasn't gotten AMD anywhere in the market.

That said, if someone is going to come on to the forum and moan "oh noes me absolute bottom end desktop doesn't game at 1080p with a iGPU anymore", it should actually be backed up with useful information like it being actually slower. A GPU is a different beast than a CPU, having more cores allows for extra levels of usability under load of the system. A GPU is all about the amount of Polygons and what not it can push out. Performance can be pretty linear. So major clock advantages can have a great affect.

This like when you were moaning about AMD not officially supporting A320 anymore, mostly because they couldn't get the motherboard manufacturers to a.) not make crap and B.) support them. Yet you still see A320 boards with 3k support.

So rather then fill up the thread about how your crappy configurations took a hit, so AMD could offer a better CPU. Why don't we actually see if that's the case first.

 

[Shivansps]

Point is neither do you. I am not worried, nor do I think AMD is terribly worried about the nitty gritty of the performance of the video part.

The CU downgrade is a fact, and AMD still have to prove that these CU are actually faster and not due to faster ram, we are only going to see that once the desktops ones are launched.

Continuing to pretend that iGPU graphics is all you really need hasn't gotten AMD anywhere in the market.

Whiout the APUs and the GPU division AMD probably wouldt exist today.
And they would never have gotten far with the DDR4-2400 limit on mobile Picasso, they entire iGPU is wasted due to that strange decision.

This like when you were moaning about AMD not officially supporting A320 anymore, mostly because they couldn't get the motherboard manufacturers to a.) not make crap and B.) support them. Yet you still see A320 boards with 3k support.

It turns out that cutting A320 support was never a thing due to not having a replacement chipset at the ready and being a huge market outside EEUU. If i was wrong A320 would be long gone by now, who takes the extra mile to support a small platform that not even AMD wanted to support? Looks like i was right, the world is not only the EEUU belive it or not.

So rather then fill up the thread about how your crappy configurations took a hit, so AMD could offer a better CPU. Why don't we actually see if that's the case first.

Fine by me, just dont come here talking about the "big gpu performance improvement that AMD tryied to hide under a rock that im not seeing".

 

[Thunder 57]

The CU downgrade is a fact, and AMD still have to prove that these CU are actually faster and not due to faster ram, we are only going to see that once the desktops ones are launched.


Whiout the APUs and the GPU division AMD probably wouldt exist today.
And they would never have gotten far with the DDR4-2400 limit on mobile Picasso, they entire iGPU is wasted due to that strange decision.


It turns out that cutting A320 support was never a thing due to not having a replacement chipset at the ready and being a huge market outside EEUU. If i was wrong A320 would be long gone by now, who takes the extra mile to support a small platform that not even AMD wanted to support? Looks like i was right, the world is not only the EEUU belive it or not.


Fine by me, just dont come here talking about the "big gpu performance improvement that AMD tryied to hide under a rock that im not seeing".

There are no numbers or benchmarks yet! Why are you so angry? And who is claiming a huge GPU increase? It looks more like AMD is saying this is a mild GPU performance increase with a huge CPU performance increase.

 

[NostaSeronx]

I am so mad! Ohmygosh!

That guy on the utubez said 20 CUs and 15 CUs!!!! How dare AMD betray us like this! I'm selling all the stock I don't have and preparing a SUIT!! My revenge will be SWIFT and without MERCY, believe me!!!

/s

~150 mm2 on 7nm TSMC should be better than ~150 mm2 on 14nm GlobalFoundries. So, $50 for 3000G and $100 for R1606G leaves the price being $150 through $300 for the full SKU.

Less cache = cheaper 8-core/16-thread??

 

[DrMrLordX]

When Renoir was revealed to use Vega rather than Navi, it became rather clear that AMD wasn't focusing on iGPU performance.

 

[lobz]

The CU downgrade is a fact, and AMD still have to prove that these CU are actually faster and not due to faster ram

Tell me please, if you think that the Picasso iGPU was already so BW strained in the first place, that a supposedly weaker or equal solution (what you think Renoir's iGPU is) appears to be faster just because it has faster RAM, then honestly, how incredibly dumb would it be from AMD to even try to make the iGPU stronger? Ibviously there won't be any dramatic BW increase till DDR5, and LPDDR4X is not cheap either. Accordingly, how much sense does it make from you to call them out on that?

 

[amd6502]

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

Definitely. Zen3 is near and it should come with Navi.

 

[Shivansps]

Tell me please, if you think that the Picasso iGPU was already so BW strained in the first place, that a supposedly weaker or equal solution (what you think Renoir's iGPU is) appears to be faster just because it has faster RAM, then honestly, how incredibly dumb would it be from AMD to even try to make the iGPU stronger? Ibviously there won't be any dramatic BW increase till DDR5, and LPDDR4X is not cheap either. Accordingly, how much sense does it make from you to call them out on that?

MOBILE PICASSO was. No idea why they limited mobile Picasso to DDR4-2400, even 2666 gives a nice boost, they really killed their own product there, you can overclock a 3400G igp on desktop running on DDR4-3200 to 1700mhz and see sustancial gains(and most of 3400G cant reach 1700mhz, 1700mhz is a rather extreme OC for Picasso). Is clear that at around 1600-1650 gpu clock you do need more than DDR4-3200. You can be sure that if Mobile Picassos were running DDR4-2933 Renoirs would not be able to get anywhere near that "30% more".

When Renoir was revealed to use Vega rather than Navi, it became rather clear that AMD wasn't focusing on iGPU performance.

Agreed, you cant focus on IGP perf with DDR4, but i expected them to keep Vega as it was on Picasso rather than making it smaller. The only reason they were able to affort that is due to Intel trash IGPs, there is no other reason.

Definitely. Zen3 is near and it should come with Navi.

What a waste, you cant really get much more than what the Vega 11@1600mhz do on mainstream DDR4. No sure about the reasults with DDR4 speeds over 4000, but the price makes that pointless.

 

[RetroZombie]

When Renoir was revealed to use Vega rather than Navi, it became rather clear that AMD wasn't focusing on iGPU performance.

Time to market might be another.
But the real benefit is the software part and driver stability, it just works

 

[uzzi38]

MOBILE PICASSO was. No idea why they limited mobile Picasso to DDR4-2400, even 2666 gives a nice boost, they really killed their own product there, you can overclock a 3400G igp on desktop running on DDR4-3200 to 1700mhz and see sustancial gains(and most of 3400G cant reach 1700mhz, 1700mhz is a rather extreme OC for Picasso). Is clear that at around 1600-1650 gpu clock you do need more than DDR4-3200. You can be sure that if Mobile Picassos were running DDR4-2933 Renoirs would not be able to get anywhere near that "30% more".



Agreed, you cant focus on IGP perf with DDR4, but i expected them to keep Vega as it was on Picasso rather than making it smaller. The only reason they were able to affort that is due to Intel trash IGPs, there is no other reason.



What a waste, you cant really get much more than what the Vega 11@1600mhz do on mainstream DDR4. No sure about the reasults with DDR4 speeds over 4000, but the price makes that pointless.

RDNA1 is approximately 25-30% more bandwidth efficient than Vega. RDNA2 probably improves on it (it kind of needs to).

You'll get your better iGPU next year.

 

[DrMrLordX]

Agreed, you cant focus on IGP perf with DDR4

See @uzzi38 's post. DDR4 isn't going to be as big a bottleneck with Navi. Had AMD made "big iGPU" their priority, they easily could have married DDR4-3200 to Navi and had a significant boost in performance, assuming Navi was ready at the time they were taping out Renoir (it might not have been).

 

[guachi]

I don't think Navi was available. I read a statement by one AMD executive that timing was the reason Navi wasn't incorporated into the new mobile chips. And if you can boost Vega, free up die space, and still beat your competition it seems like a win for AMD.

The only negatives I see are system integrators giving us second-rate laptops.

 

[Shivansps]

RDNA1 is approximately 25-30% more bandwidth efficient than Vega. RDNA2 probably improves on it (it kind of needs to).

You'll get your better iGPU next year.

oh, i dindt know that, well that changes things, RDNA2 APU on DDR4 will be great, posible GTX1050 perf on DDR4, if they make the iGPU large enoght (remember that they need to sell dGPUs). And the DDR5 version has the potential to be equal or better than a RX570, too bad thats 2022+.

I would guess then that Renoir was meant to have a similar sized Navi GPU as a Vega 8. That makes more sence.

 

[uzzi38]

oh, i dindt know that, well that changes things, RDNA2 APU on DDR4 will be great, posible GTX1050 perf on DDR4, if they make the iGPU large enoght (remember that they need to sell dGPUs). And the DDR5 version has the potential to be equal or better than a RX570, too bad thats 2022+.

I would guess then that Renoir was meant to have a similar sized Navi GPU as a Vega 8. That makes more sence.

Renoir was always supposed to be Vega, because the RDNA IP wasn't ready at the time of development.

It has less CUs because AMD wanted something they could create in large volumes, so smaller chips were necessary. That's why the cache was also cut down.

But when Navi APUs release, they absolutely will be lower CU counts as well, though this is because Navi uses more bandwidth per CU. It just spits out more performance per CU than it does use extra bandwidth.

 

[Hulk]

Since these Ryzen 4000 series mobile parts are Zen 2 based are they expected to have better IPC than the Ryzen 3000 series mobile? If so how much and why?

Otherwise it seems as though Ice Lake will still be comparable core for core so long as Intel gets 8 core Ice Lake's out there.

Of course the game will change if Zen 4 can beat Ice Lake IPC-wise.

 

[moinmoin]

Since these Ryzen 4000 series mobile parts are Zen 2 based are they expected to have better IPC than the Ryzen 3000 series mobile? If so how much and why?

Ryzen Mobile 3000 is using Zen+ cores, so you can expect to see the IPC increase from Zen+ to Zen 2 in Ryzen Mobile 4000. That IPC increase is commonly put between 13-18%.

 

[rainy]

Since these Ryzen 4000 series mobile parts are Zen 2 based are they expected to have better IPC than the Ryzen 3000 series mobile? If so how much and why?

Similarly to the desktop parts (Mattise - Ryzen 3xxx) you should expect about 15 percent higher IPC over Zen+ (Pinnacle Ridge/Picasso - Ryzen 2xxx/3xxx).

Of course the game will change if Zen 4 can beat Ice Lake IPC-wise.

I know it's a bit confusing but you mean Zen 3 (Ryzen 4xxx)?

 

[Hulk]

Similarly to the desktop parts (Mattise - Ryzen 3xxx) you should expect about 15 percent higher IPC over Zen+ (Pinnacle Ridge/Picasso - Ryzen 2xxx/3xxx).



I know it's a bit confusing but you mean Zen 3 (Ryzen 4xxx)?

Yes that's what I meant. I didn't know that ryzen 3000 series where Zen plus! Ryzen 4000 series mobile are going to be wicked.

 

[rainy]

I didn't know that ryzen 3000 series where Zen plus!

Only for mobile/desktop APUs.

 

[tamz_msc]

Looks like Geekbench 4 results are out:

LENOVO INVALID - Geekbench

Benchmark results for a LENOVO INVALID with an AMD Ryzen 7 4700U processor.

browser.geekbench.com

It boosts to ~4.2GHz in this run. Here is a 3700X at ~4.4GHz for comparison:

Micro-Star International Co., Ltd MS-7C02 - Geekbench

Benchmark results for a Micro-Star International Co., Ltd MS-7C02 with an AMD Ryzen 7 3700X processor.

browser.geekbench.com

As you can see integer score is up roughly 10% from a 5% increase in clock speed. To me it seems like Renoir is losing IPC compared to Matisse.

 

[uzzi38]

Looks like Geekbench 4 results are out:

LENOVO INVALID - Geekbench

Benchmark results for a LENOVO INVALID with an AMD Ryzen 7 4700U processor.

browser.geekbench.com

It boosts to ~4.2GHz in this run. Here is a 3700X at ~4.4GHz for comparison:

Micro-Star International Co., Ltd MS-7C02 - Geekbench

Benchmark results for a Micro-Star International Co., Ltd MS-7C02 with an AMD Ryzen 7 3700X processor.

browser.geekbench.com

As you can see integer score is up roughly 10% from a 5% increase in clock speed. To me it seems like Renoir is losing IPC compared to Matisse.

Hold your horses a little bit on the doom and gloom mate. 2667mhz memory was used for that run, and looking at memory bandwidth, it's also single channel.

 

[DrMrLordX]

To me it seems like Renoir is losing IPC compared to Matisse.

It has less L3 cache, so what else do you expect?

 

[tamz_msc]

Hold your horses a little bit on the doom and gloom mate. 2667mhz memory was used for that run, and looking at memory bandwidth, it's also single channel.

Memory doesn't affect the integer subscore that much, which is what I'm comparing.

 

[HurleyBird]

It has less L3 cache, so what else do you expect?

But Renoir should have significantly better memory latency, and that reduced L3 probably has tighter timings, and there may be other small improvements here and there. I suspect it will be somewhat of a mixed bag, but if I had to guess I'd say Matisse should be a bit faster overall given the L3 advantage.