Gaming on a Ryzen 9 4900HS Laptop with a Discrete GPU

[Julio Franco]

Following up to our review of AMD's new Ryzen Mobile 4000 laptop CPU and having covered productivity performance and touched on gaming using the integrated GPU in our initial review, now it's time to tackle the other main use case for these processors, and that's gaming with a discrete GPU.

Read the full article here.

https://www.techspot.com/review/2006-gaming-ryzen-9-4900-hs-laptop/

 

[Lounds]

Obviously hard when you've got limited laptops to do these benchmarks on but definitely promising for a laptop CPU and GPU combination. Hopefully this boosts AMD'S market share in the laptop industry.

 

[krizby]

I would be interested in how this 4900HS would perform in an Ultrabook attached to a 2080 Ti external GPU gaming box.

 

[seeprime]

I wonder how long of a lifespan those machines will have, with a total of about 110 watts, which gets laptops warm even with great cooling, between the CPU and GPU. Our PC shop sees a lot of gaming laptops that barely last five years, due to higher heat levels and heavy usage.

 

[Audiogeek]

I wonder how long of a lifespan those machines will have, with a total of about 110 watts, which gets laptops warm even with great cooling, between the CPU and GPU. Our PC shop sees a lot of gaming laptops that barely last five years, due to higher heat levels and heavy usage.

My friend lost a i7 with gtx1080 laptop. He was using it more for productivity. So, yes durability is a weakness of gaming laptops.

 

[Gimp65]

No power usage comparison ? (perhaps it in another review?)

Power usage is essential for laptops. AMD dominates intel here.

 

[fred28056]

I would be interested in how this 4900HS would perform in an Ultrabook attached to a 2080 Ti external GPU gaming box.

Unfortunately that can't happen with the mobile ryzen as it doesn't have thunderbolt support. I know some motherboards for desktop ryzen have finally started adding thunderbolt but there probably isn't room in an ultrabook for the thunderbolt controller. Intel has the entire or portion of the controller built into their cpus which allows them to put it in ultrabooks. We probably won't get eGPU support for Ryzen until next years mobile 5000 series that will include USB4 support which is really just an opened up and rebranded thunderbolt 3.

 

[BrotherMichigan]

Can the author comment on the particular model used for the 9750H numbers?

 

[Squid Surprise]

Not sure why you bothered writing this article as you even admit it won’t actually tell you anything... guess everyone has lots of time on their hands thanks to the pandemic and I guess that means it wasn’t a complete waste of time...

So what conclusions can we draw? Nothing! We already knew that this CPU blows Intel’s away in productivity... we can guess that it will probably be on par or better than Intel in gaming if the same GPU is being used... but we knew that before this article was written!

The fact is, almost all games - either mobile or desktop - are GPU limited, not CPU limited. You can use 5+ year old CPUs and get virtually identical performance on a game provided you use a high end GPU with it. —> again, something we knew well before this article was written.

Of course, it wasn’t a complete waste - you killed about 20 minutes of my day for me.. thanks!

 

[hahahanoobs]

Thanks!
Gaming battery tests coming?

 

[yeeeeman]

How is exactly comparing 9750H (6 core) with 4900HS (8 core) performance at 720p fair?
AMD mobile parts are slower than desktop parts in gaming because they have smaller L3 caches, so the inter-CCX latencies will be worse. So basically a 9880H should be faster than 4900H. More so the 10th generation which gets even higher clocks.

 

[yeeeeman]

I would be interested in how this 4900HS would perform in an Ultrabook attached to a 2080 Ti external GPU gaming box.

You basically need Tiger Lake for that. It will have TB4 + 15-20% higher IPC than Zen 2 parts.

 

[Evernessince]

How is exactly comparing 9750H (6 core) with 4900HS (8 core) performance at 720p fair?
AMD mobile parts are slower than desktop parts in gaming because they have smaller L3 caches, so the inter-CCX latencies will be worse. So basically a 9880H should be faster than 4900H. More so the 10th generation which gets even higher clocks.

Just like how the 8700K ties the 9900K in gaming performance, 8 cores yields little to no benefit in gaming right now. I should also point out that the 4900HS is AMD's low power variant. You could equally lodge a complaint that the Intel CPU is going to consume significantly more power to achieve it's shown performance. In fact a higher wattage 6 core CPU has a distinct advantage over a lower wattage 8 core when it comes to gaming. That it is losing while having to power 2 less cores that do little to nothing for gaming "should" give it the advantage but in reality it didn't work out so well.

'AMD mobile parts are slower than desktop parts in gaming because they have smaller L3 caches"

I don't think we have the data to support such a determination right now. I've seen some games where the performance difference could be attributable to cache but then again it could also be power restrictions. There are also quite a few games where the 4900HS's performance is close to the 3700X.

"So basically a 9880H should be faster than 4900H"

The data is saying otherwise.

 

[neeyik]

How is exactly comparing 9750H (6 core) with 4900HS (8 core) performance at 720p fair?

From the article itself:

The biggest challenge for this test was getting an apples-to-apples platform to compare the 4900HS against Intel processors. Our Ryzen 4000 test bed, the Asus Zephyrus G14, comes with a GeForce RTX 2060 Max-Q with a 65W power limit, which we haven't seen in any other laptops.

We wanted this to be a completely fair battle with the same GPU, but at least we managed the next best thing. We've put the 4900HS up against the closest platform we could find, which is a Core i7-9750H laptop with an RTX 2060 at 80W. Unfortunately we were unable to source an 8-core Intel laptop with the RTX 2060, but that will prove easier with the upcoming 10th-gen Intel line-up that features the 8-core Core i7-10875H.

 

[BrotherMichigan]

they have smaller L3 caches, so the inter-CCX latencies will be worse.

Inter-CCX latency won't be worse because of this, but you will have to go out to main memory more often, which will result in a performance hit. Memory access and average inter-CCX latency is actually probably a bit better with Renoir given its monolithic design. It doesn't have that potential extra hop across the package to the IOD that Matisse has to make to get to system memory and it doesn't suffer the worst case scenario of two hops to access the other CCD on the package like the 3900X and 3950X do.

 

[Lew Zealand]

You basically need Tiger Lake for that. It will have TB4 + 15-20% higher IPC than Zen 2 parts.

And AMD is likely to have Zen 3 parts out then with similar IPC gains so the future will be interesting.

 

[Aryassen]

Not sure why you bothered writing this article as you even admit it won’t actually tell you anything... guess everyone has lots of time on their hands thanks to the pandemic and I guess that means it wasn’t a complete waste of time...

So what conclusions can we draw? Nothing! We already knew that this CPU blows Intel’s away in productivity... we can guess that it will probably be on par or better than Intel in gaming if the same GPU is being used... but we knew that before this article was written!

The fact is, almost all games - either mobile or desktop - are GPU limited, not CPU limited. You can use 5+ year old CPUs and get virtually identical performance on a game provided you use a high end GPU with it. —> again, something we knew well before this article was written.

Of course, it wasn’t a complete waste - you killed about 20 minutes of my day for me.. thanks!

Have to disagree, sorry....
"....but we knew that before this article was written!"
No, we didn't know. We thought. Now, we know. Because it has been tested. Thoroughly. And the fact that the testing didn't bring any earth-shattering surprises to the table doesn't render the effort unworthy (in my opinion).

 

[hahahanoobs]

And AMD is likely to have Zen 3 parts out then with similar IPC gains so the future will be interesting.

Maybe not. Rumour going around says IPC will be lower than the 18% we heard from AMD.

 

[kmo911]

Resident evil 2 can run vey low like a ps1 gpu grafics. so comparing a 4k game as low as 720p are that you can run it low res from beginning. 300x200 for non strong gpu cpu s.
if other games can run 640x480-800x600 it would run it even on a low end pc and no limitations on that low fps cpu gpu and atleast 2 gb gpu. if you try this on a p 2 200 (3) p4 you can run r e 2 R on even lower specs. 1 gb ddr 1 2 ram. so getting all games down to low fps means the laptops cant run 4k at full.
it just like running orginal r e 2 pc on 4k and then put it back on lower resolutions.
wolfenstein young blodd in that mode would give it a good fps even in 300x200 (if supported)
I would be like comparing a old pc with a new 10 gen cpu gpu but wit only 128mb-512mb ram and a tnt u 2 with low memory that can run thos game as a low fps machine.
so all that cores an it would run even better on desktop machines. a game thats needs a p 2 200 mhz can be run on low specs like p 100 60 90 with same pci with 128 mb gpu and edoram at 128 max. but future computers would run 10 000+ better with new rtx 1080 ti and 2080 ti gpus. compare them with 1k-4k even results getting low. thats a nørd LIKe ting. and get a userbenchmark.com from it to and detection tool like can you run IT.

 

[veLa]

I'd really like to see these Ryzen 4000 laptop processors compared to desktop processors like the 2700X or 3700X, but all the reviewers seem to think you're either choosing between an AMD laptop or an Intel laptop rather than a desktop or a desktop replacement.

 

[Squid Surprise]

Have to disagree, sorry....
"....but we knew that before this article was written!"
No, we didn't know. We thought. Now, we know. Because it has been tested. Thoroughly. And the fact that the testing didn't bring any earth-shattering surprises to the table doesn't render the effort unworthy (in my opinion).

No... we KNEW... Techspot already ran the article benchmarking this CPU in compute tasks...

As for gaming - this article tells us NOTHING new. It even states as much! We didn't get an "apples to apples" comparison in GPU or CPU! We got a 6 core vs an 8 core, and we got a 2060 vs a 2060 Max Q...

I understand that this was as close as Techspot could get - but it's a WASTE OF TIME!!

 

[meric]

These small powerhouse of APUs (without a dGPU) would make for an ideal 720p casual gaming machines for cheap. I'm thinking of buying one.

 

[lipe123]

No... we KNEW... Techspot already ran the article benchmarking this CPU in compute tasks...

As for gaming - this article tells us NOTHING new. It even states as much! We didn't get an "apples to apples" comparison in GPU or CPU! We got a 6 core vs an 8 core, and we got a 2060 vs a 2060 Max Q...

I understand that this was as close as Techspot could get - but it's a WASTE OF TIME!!

Wow really? you need to tone it down a few notches.

They compared what was available on the market now, it doesn't matter if its apples to oranges when the apples and oranges cost the same price, and that's all you can buy.

The whole point of these reviews/tests is so consumers can make informed decisions.
So after reading this your choice is:
Do I want an 8 core cpu that uses less power than a 6 core one, for about the same or slightly more performance?

 

[Squid Surprise]

Wow really? you need to tone it down a few notches.

They compared what was available on the market now, it doesn't matter if its apples to oranges when the apples and oranges cost the same price, and that's all you can buy.

The whole point of these reviews/tests is so consumers can make informed decisions.
So after reading this your choice is:
Do I want an 8 core cpu that uses less power than a 6 core one, for about the same or slightly more performance?

Really? That's not how benchmarks work... I suggest you actually READ the article - it even admits that they wanted to make an apples-to-apples comparison... they couldn't!

Now, at this point, most people would either A) not write the article or B) Change the article to something that they could do - like "is it worth buying the 4900 in a laptop" or some such...

They did neither... Techspot decided to go ahead with their benchmarks, spending tons of time telling us completely irrelevant information...

Please tell me where I've gone wrong - I'm all ears!

 

[Lew Zealand]

The Ryzen 3000 series laptop CPUs were OK for productivity but lagged behind Intel for gaming. The Ryzen 4000 series CPUs were a slam dunk for productivity but we did not know if that would translate to gaming. Maybe Zen2's latency was still too much of a problem and gaming would suffer.

At least with a 2060 MQ, that ended up not being the case. Actually I was expecting the gaming numbers for Ryzen 4000 to be lower. And even with 2060MQ vs 2060, there is variability in each chassis. Sometimes the 2060MQ beats the 2060 with the same CPU. Have a look at Jarrod's Tech for comparos on this.