Why are all the Ryzen based laptops sandbagged with the 2nd best GPUs while intel gets the best ones?

[de.das.dude]

I was recommending a laptop to a friend, and i noticed that all the AMD ryzen based laptops never get the best GPU.
Finally settled on the HP Omen 15 with the Ryzen 4600H, but its a shame that you dont get the best GPUs. Most people would spend this much money on a laptop for gaming purposes only. Very few people would buy it for work.

 

[ChiNoel]

Laptop manufacturers might have an agreement with Intel so Intel can make more money before AMD laptops get a better GPU. Like the Lenovo Legion 5.
Another theory is that 3000 series of Ryzen mobile CPU kinda sucks but 4000 series is so good that they didn't prepare for it.
I personally went for the Asus TUF Gaming A15. Despite its thermal issue I still think it's a great value.

 

[Caring1]

Perhaps AMD had the sense to limit heat output rather than go balls out and then deal with countless complaints about throttling.

 

[king of swag187]

Despite its thermal issue I still think it's a great value.

Give it a week

 

[tabascosauz]

I was recommending a laptop to a friend, and i noticed that all the AMD ryzen based laptops never get the best GPU.
Finally settled on the HP Omen 15 with the Ryzen 4600H, but its a shame that you dont get the best GPUs. Most people would spend this much money on a laptop for gaming purposes only. Very few people would buy it for work.

Well, look back at the years upon years of Intel's underhanded dealings with OEMs. Did you really have to ask?

The 4 FX generations and 3000 Mobile firmly cemented AMD's reputation as a vendor of "budget" hardware. We all know that Renoir is a beast, but you gotta give them and the laptop manufacturers more than a single generation to turn that reputation around. Mindshare also doesn't come overnight; regardless of how good Renoir is, you need time to spread the word so that it's what comes to people's minds instead of Picasso.

The more that Renoir gets more design wins like the Zephyrus G14 and Thinkpad X13, the more willing the manufacturers will be to throw in a high end GPU towards the end of Renoir's lifecycle, or in the next generation of Ryzen Mobile.

As Comet Lake goes away, Tiger Lake has some massive shoes to fill without high core count parts. This might prove to be a boon for AMD.

 

[newtekie1]

In my experience it has to do with how well nVidia's cards play with Intel CPUs/iGPUs. At least in my experience, nVidia has refined it to the point where the nVidia GPU is basically turned completely off in the Intel based laptops when the iGPU is in use. I don't think they've really worked that much on getting it to do that on the AMD platform. This is of course no fault of AMD, AFAIK, it's a fault of nVidia and their lack of development for the mobile AMD platform. I assume it works like this because on the Intel based laptops I've used, when the nVidia GPU isn't being used, you can't even read the sensors from it. It goes into some kind of very deep sleep. But on the AMD laptops I've used, the nVidia GPU just goes to idle and it's always readable.

The down side of this is that, because the nVidia GPU is always active on the AMD platform, then tend to have worse battery life(again, nVidia's fault not AMDs). So the laptop manufacturers only use the weaker GPUs and not the higher end ones so they can keep advertising decent runtimes. That's just my theory.

 

[erocker]

There's some sketchy stuff going on, just ask Asus. Saw this on Reddit yesterday.

Amazon has a Lenovo 2 in 1 with a 4500u for $599 I'm very tempted to get right now.

 

[ARF]

I was recommending a laptop to a friend, and i noticed that all the AMD ryzen based laptops never get the best GPU.
Finally settled on the HP Omen 15 with the Ryzen 4600H, but its a shame that you dont get the best GPUs. Most people would spend this much money on a laptop for gaming purposes only. Very few people would buy it for work.

Maybe because if they get the fastest graphics, it won't show its full performance because of the way Ryzen APUs are designed.
PCIe 3.0 x8 is no good:

AMD Renoir (Mobility) PCIe 3.0 Limited To 8x – Could Bottleneck GPUs Above An NVIDIA RTX 2060 Super

AMD Renoir (Mobility) PCIe 3.0 Limited To 8x - Could Bottleneck GPUs Above An NVIDIA RTX 2060 Super

AMD's Renoir SOC has just got its first legitimate platform block diagram leaked courtesy of the venerable Igor over at Igor's Lab. The source in question has a spotless track record with leaks and we have no reason to doubt the legitimacy of the slide. The AMD Renoir SoC block diagram (for the...

wccftech.com

 

[Assimilator]

The only answer to this question (which I already posted, but apparently straight answers are discouraged here in favour of uneducated guesses) is: ask the OEMs. Only they know why they are making the choices they are making.

It's most likely to be as @newtekie1 stated: OEMs simply have more experience with the Intel CPU/NVIDIA GPU combo than the AMD CPU/NVIDIA GPU combo. Coupled with AMD's historically poor drivers and integration support, it means building a laptop with an AMD CPU is somewhat of a higher risk and higher complexity than using a known quantity like Intel.

There's some sketchy stuff going on, just ask Asus. Saw this on Reddit yesterday.

View attachment 163117


Amazon has a Lenovo 2 in 1 with a 4500u for $599 I'm very tempted to get right now.

Yeah, because you should always trust things you read on Reddit...

Especially ones that are fundamentally retarded in their premise, because why would any manufacturer willingly cripple their own products? Wait, lemme guess, it's because they're being paid to by Intel. Right? Much like how jet fuel can't melt steel beams, right?

Maybe because if they get the fastest graphics, it won't show its full performance because of the way Ryzen APUs are designed.
PCIe 3.0 x8 is no good:

AMD Renoir (Mobility) PCIe 3.0 Limited To 8x – Could Bottleneck GPUs Above An NVIDIA RTX 2060 Super

AMD Renoir (Mobility) PCIe 3.0 Limited To 8x - Could Bottleneck GPUs Above An NVIDIA RTX 2060 Super

AMD's Renoir SOC has just got its first legitimate platform block diagram leaked courtesy of the venerable Igor over at Igor's Lab. The source in question has a spotless track record with leaks and we have no reason to doubt the legitimacy of the slide. The AMD Renoir SoC block diagram (for the...

wccftech.com

But we aren't talking about APUs here.

Regardless, the impact of PCIe x8 vs PCIe x16 is only noticable with RTX 2080 Ti, and you aren't going to find a 2080 Ti in a laptop. Considering laptop GPUs are also downclocked compared to desktop counterparts it's even less likely that the lower lane count will have a performance impact.

 

[erocker]

Lol, don't get all worked up about it!

 

[nguyen]

Regardless, the impact of PCIe x8 vs PCIe x16 is only noticable with RTX 2080 Ti, and you aren't going to find a 2080 Ti in a laptop. Considering laptop GPUs are also downclocked compared to desktop counterparts it's even less likely that the lower lane count will have a performance impact.

Not really, even the lowly 5500XT 8GB is affected when running at PCIe 3.0 x8 when compare to PCIe 4.0 x8

This could the perfect explanation for why OEM don't want to integrate high end GPU with Renoir Laptop...

 

[Assimilator]

Not really, even the lowly 5500XT 8GB is affected when running at PCIe 3.0 x8 when compare to PCIe 4.0 x8

This could the perfect explanation for why OEM don't want to integrate high end GPU with Renoir Laptop...

The 5500 XT isn't a high-end GPU, and it isn't available in laptops. So what was your point exactly?

 

[nguyen]

The 5500 XT isn't a high-end GPU, and it isn't available in laptops. So what was your point exactly?

My point is a GPU as slow as the desktop 5500XT (which is slower than mobile 1660 Ti) is already affected by the reduced PCIe bandwidth, higher performance mobile GPU will see bigger handicap when restricted to PCIe 3.0 x8...
Not that hard to interpolate informations though.

 

[TheLostSwede]

AMD Ryzen 4000H series capped to PCIe 3.0 x8, limiting high-end GPU options - VideoCardz.com

AMD Renoir FP6 (Mobile) is capped to PCIe 3.0 x8 Igor Wallosek has shared a new block diagram for the Ryzen 4000H and U series platform codenamed Renoir FP6. The diagram reveals the sad truth about PCI-Express interface bandwidth limitation. This could be the main reason why notebooks makers...

videocardz.com

 

[TheoneandonlyMrK]

My point is a GPU as slow as the desktop 5500XT (which is slower than mobile 1660 Ti) is already affected by the reduced PCIe bandwidth, higher performance mobile GPU will see bigger handicap when restricted to PCIe 3.0 x8...
Not that hard to interpolate informations though.

Your talking 2-4% for a x16 running at x8.
An insignificant amount.

But do go on.

 

[BoboOOZ]

AMD Ryzen 4000H series capped to PCIe 3.0 x8, limiting high-end GPU options - VideoCardz.com

AMD Renoir FP6 (Mobile) is capped to PCIe 3.0 x8 Igor Wallosek has shared a new block diagram for the Ryzen 4000H and U series platform codenamed Renoir FP6. The diagram reveals the sad truth about PCI-Express interface bandwidth limitation. This could be the main reason why notebooks makers...

videocardz.com

Interesting article, but it doesn't apply to performance on the mobile lineup:

NVIDIA GeForce GTX 1080 PCI-Express Scaling

In this article, we investigate how performance of NVIDIA's GeForce GTX 1080 is affected when running on constrained PCI-Express bus widths such as x8 or x4. We also test all PCIe speed settings, 1.1, 2.0, and 3.0. One additional test checks on how much performance is lost when using the...

www.techpowerup.com

Here are the scaling results for a 1080Ti, and the difference between 8x and 16X is 0, and that is more than a mobile 2080 performance...
So, no, I think it's more of a mindshare and stable software environment problem as @tabascosauz pointed out.

 

[Assimilator]

Let's get one thing straight:
I think that AMD limiting their laptop APUs to only 8 PCIe lanes for graphics is a poor decision, especially considering their own Navi discrete GPUs have been shown to suffer in performance with that lane count as compared to the full standard 16 lanes.
That might be why we aren't seeing many laptops with AMD APU + Navi GPU.

But I do not think it is the reason why we are not seeing AMD APUs + high-end NVIDIA GPUs in laptops. As others have posted, TPU's very own tests demonstrate that even the fastest NVIDIA GPU is not as dependent on PCIe lane count as Navi.

 

[nguyen]

Interesting article, but it doesn't apply to performance on the mobile lineup:

NVIDIA GeForce GTX 1080 PCI-Express Scaling

In this article, we investigate how performance of NVIDIA's GeForce GTX 1080 is affected when running on constrained PCI-Express bus widths such as x8 or x4. We also test all PCIe speed settings, 1.1, 2.0, and 3.0. One additional test checks on how much performance is lost when using the...

www.techpowerup.com

Here are the scaling results for a 1080Ti, and the difference between 8x and 16X is 0, and that is more than a mobile 2080 performance...
So, no, I think it's more of a mindshare and stable software environment problem as @tabascosauz pointed out.

That was the GTX 1080, not 1080 Ti. Furthermore the article was from 2016, with the test setup consist of 6700K + GTX 1080. Performance of mobile CPU + GPU have already exceeded desktop 6700K + GTX 1080 from 4 years ago.
The Hardware Unboxed clip also show that the 1% low FPS that take a heavier hit from PCIe 3.0 x8 bandwidth than avg FPS, while TPU didn't record 1% low FPS.

 

[Assimilator]

That was the GTX 1080, not 1080 Ti. Furthermore the article was from 2016, with the test setup consist of 6700K + GTX 1080. Performance of mobile CPU + GPU have already exceeded desktop 6700K + GTX 1080 from 4 years ago.
The Hardware Unboxed clip also show that the 1% low FPS that take a heavier hit from PCIe 3.0 x8 bandwidth than avg FPS, while TPU didn't record 1% low FPS.

Imagine if you weren't lazy and did your own research: https://www.techpowerup.com/review/nvidia-geforce-rtx-2080-ti-pci-express-scaling/7.html

 

[BoboOOZ]

Performance of mobile CPU + GPU have already exceeded desktop 6700K + GTX 1080 from 4 years ago.

So we go from 0% difference to 1% difference?

The Hardware Unboxed clip also show that the 1% low FPS that take a heavier hit from PCIe 3.0 x8 bandwidth than avg FPS, while TPU didn't record 1% low FPS.

Do you really think that this is why you don't see Renoir paired with mobile 2080?

 

[micropage7]

I was recommending a laptop to a friend, and i noticed that all the AMD ryzen based laptops never get the best GPU.
Finally settled on the HP Omen 15 with the Ryzen 4600H, but its a shame that you dont get the best GPUs. Most people would spend this much money on a laptop for gaming purposes only. Very few people would buy it for work.

sometimes i think that, for that screen it's not too necessary to put the high end GPU since medium GPU will be good on that screen, except if you buy pc where you can hook it up to multi monitors
beside the heat and another things that need to be compromised when they put hot high end GPU in laptop

 

[de.das.dude]

There's some sketchy stuff going on, just ask Asus. Saw this on Reddit yesterday.

View attachment 163117


Amazon has a Lenovo 2 in 1 with a 4500u for $599 I'm very tempted to get right now.

Glad to see you around.

I actually saw a video on youtube who took a dremel to it and actually saw significant improvement in performance. He contacted ass us and they gave some snake oil explanation saying it improves overall feel and reduces keyboard region temperature...

, lemme guess, it's because they're being paid to by Intel. Righ

Hey intel wont do something like this ever right?? Oh wait they did and lost a lawsuit already, if memory serves me right. As i mentioned, people opened up the vents and saw performance gains.

Not really, even the lowly 5500XT 8GB is affected when running at PCIe 3.0 x8 when compare to PCIe 4.0 x8

This could the perfect explanation for why OEM don't want to integrate high end GPU with Renoir Laptop...

This is interesting... They give 2 m.2 slots but 8x pcie slot? That seems silly yes? Is it because the chipset has to be nerfed? I know 570 chip has some thermal issues

yo

sometimes i think that, for that screen it's not too necessary to put the high end GPU since medium GPU will be good on that screen, except if you buy pc where you can hook it up to multi monitors
beside the heat and another things that need to be compromised when they put hot high end GPU in laptop

u can hook up a laptop to a 4K 144mhz as well...

 

[BoboOOZ]

By the way, newest comparison between mobile Nvidia GPU and desktop ones show the difference in performance is still several tiers and 1% lows are way worse then on desktop.

And by the way, all hardware news YouTube channels talk about Igor's article as the answer for the question of the OP without any of them looking at the actual numbers in the article...

 

[Vader]

Poor relative performance on the rx5500 when using pcie 3.0 was attributed to it's low memory buffer (4 gb sku) and only present in games that could utilize more than that amount. Frequent data exchange between the gpu and the ram to swap data would saturate the link and significantly reduce the performance. The 8 gb model was barely affected by the x8 link

As for why there are no laptops with gpus above the 2060, i can list a few reasons:

1. Amd themselves said that they want to target their laptops to the majority of buyers. 2060-class products and below are what most people buy. This might be why you also don't see amd in the most "exotic" variants of laptops
2. Manufacturers have to work with amd to engineer their products, see point 1
3 They might leave the upper end open for future rdna 2 mobile gpus exclusively, thus bringing sales to their cards as they have the better cpus right now

 

[Vya Domus]

Right? Much like how jet fuel can't melt steel beams, right?

Like how OEMs were paid to delay launches from AMD.
Like how OEMs were paid to restrict certain AMD products from being sold.
Like how compilers did not apply the same optimizations on AMD chips.
Like how software such as Matlab deliberately chose slower code paths on AMD chips.

Yeah, just like that ...