jabbadapNot quite. It's a different chip. MX250 is rebrand of mx150 aka laptop gt1030 using gp108 chip(aka mx330). MX350 in other hand uses gp107 chip, same as used in gtx 1050/ti cards. So it's actually a 1050 with half the memory bandwidth. It will probably have nvenc too which gp108 lacks(unless that is disabled by nvidia).

Lol if dealing with NVENC, NVIDIA really should've made a Turing version since that would inherently be the best for low power.

Berfs1Lol if dealing with NVENC, NVIDIA really should've made a Turing version since that would inherently be the best for low power.

Smallest Turing gpu with Turing nvenc is tu116, tu117 has Volta's nvenc so not the newest and greatest but par with the pascal. So no, Rumored MX450 with Turing gpu inside gives little to null benefit over pascal nvenc found on mx350.

One thing with these ultrabook with ulv processors and why there's a place for discrete graphics chip(albeit rather weak one) is spreading the cooling from one hot chip to two.

By Chinese tests, the 4700U scores about 1050-1100 graphics points in Time Spy and the MX350 about 1350-1400. So yes, MX350 is faster. But is that big enough of a difference to make sense of having both? Most likely not for pure (gaming) performance reasons. So CUDA, power/cooling envelope and/or marketing are the reasons. Asus has seen a market for such laptop, and sure it's possible that they're not entirely wrong.

Berfs1Ok, allow me to correct myself, how many people would buy an i9-9980HK laptop with a MX350? There is absolutely no point in having a weak dGPU when the iGPU performs its job more than perfectly. UHD 630 is really good for web browsing and small stuff.

But this is not a i9-9980HK. This is a 15W SoC for slim laptops. Adding a dGPU, among other pros I've mentioned, also helps with heat concentration and probably cooling.

The market for a high performance laptop with a dedicated weak GPU is very, very small.

No, it's not. Seriously, look around...

HD630 is fine for browsing and watching movies, but faster dGPU offer a significant bump in GPU acceleration. And then it often flats out.
Great example:
www.pugetsystems.com/labs/articles/Photoshop-CC-2018-NVIDIA-GeForce-GPU-Performance-1139/#BenchmarkAnalysis

This is precisely why MacBook Pro 16 comes with a 5300M and XPS 15 gets a 1650. Both can be had with the massive i9 -H SoCs. Both can cost over $3000.

So why purchase an 8c8t processor for content creation? We are talking about the R7-4700U not the R7-4800U, which is an 8c16t processor. The fact that the 4700U doesn't have SMT means it is automatically oriented towards gamers. If this were a 4800U laptop, then sure it would make sense.

Because 4700U is cheaper. People don't always buy the fastest CPU that exists for a task. There is usually some choice of components.

IceShroomThe same reason we dont need it in this laptop. It is not a gaming laptop.

Have you already heard about GPUs being utilized for tasks other than gaming or is this a novelty to you? :)

jabbadapSmallest Turing gpu with Turing nvenc is tu116, tu117 has Volta's nvenc so not the newest and greatest but par with the pascal. So no, Rumored MX450 with Turing gpu inside gives little to null benefit over pascal nvenc found on mx350.

One thing with these ultrabook with ulv processors and why there's a place for discrete graphics chip(albeit rather weak one) is spreading the cooling from one hot chip to two.

Why not just get a TU117 GPU and cut the core count? That should work. 1650 Super and up have the NVENC encoder, so just take one of those and cut the cores, and you will still have the NVENC lol

JPpageAccording to Tom's, the R7 4800u's iGPU is very close to the MX 250. I suppose an LPDDRX memory is enough to beat or leave the Vega iGPU safely on the same level as the MX250.
In my point of view, this MX 350 is useless, except for those who need CUDA (I win that most people don't even know what it is)

That's nice, but not really useful for this discussion. This laptop is using the 4700u, which has an iGPU that's only about 20% slower than the iGPU in the 4800u.

JPpageThe Vega 10 is on the same level as the 1030, as long as it is equipped with DDR4 3200.

It's actually Vega 11 that is on the same level as a GT 1030. Vega 10 is below a GT 1030. Oh, and the iGPU in the 4700u isn't even close to Vega 10 or 11. Vega 11 gets its name because it uses 11 Vega CUs, equalling 704 Vega shaders. Ths iGPU on the 4700u uses 7 Vega CUs, so if it was following the same naming convention would be Vega 7... Yeah, it's a weak GPU.

R0H1TIsn't it a rebrand of MX250, Pascal based, IIRC whilst the Mx450 should be Turing based. There's no way anything with 64bit wide bus would come anywhere near 1050 in performance, unless they've significantly bumped the specs & it has GDDR6 memory.

It's only a rebranded MX250 if you also think a GTX 1050 is just a rebranded GT 1030. Using the same architecture doesn't make the cards the same. According to the GPU Database here at TPU, the MX350 is twice as fast as the MX250/GT1030. It's not a totally weak GPU like some people seem to think it is. Yeah, compared to dekstop cards it is, but for a thin and light laptop it's pretty damn powerful.

I think there are a lot of people in this thread that are over-estimating how powerful the AMD iGPU is in the 4700u as well as under-estimating how powerful the MX350 is.

Berfs1Why not just get a TU117 GPU and cut the core count? That should work. 1650 Super and up have the NVENC encoder, so just take one of those and cut the cores, and you will still have the NVENC lol

Because the physical amount of space the GPU takes up also matters in laptop design. TU116 is, relatively, a massive GPU to put in a laptop. It takes up a log of board space, about double what GP107 does.

Berfs1The Ryzen processor already has an inbuilt GPU, why waste money, circuit board area, power and thermals for an extra GPU that is almost pointless?

An AMD APU that you have not installed the chipset for, is very difficult to do without a gpu firsthand. If you have a score to settle on somebody, just delete their chipset driver and they are back to stone age.

londisteAbout what exactly? Retailers trying to get rid of old stock?

This was the latest, actually. There are some reports why it is old by origin.

specopsFIBy Chinese tests, the 4700U scores about 1050-1100 graphics points in Time Spy and the MX350 about 1350-1400. So yes, MX350 is faster. But is that big enough of a difference to make sense of having both? Most likely not for pure (gaming) performance reasons. So CUDA, power/cooling envelope and/or marketing are the reasons. Asus has seen a market for such laptop, and sure it's possible that they're not entirely wrong.

Those tests were, as I mentioned, with DDR4 (3200MHz), not LPDDR4x, so it's possible we'll see better performance in laptops where the CPU is paired with faster RAM as well.

TheLostSwedeThose tests were, as I mentioned, with DDR4 (3200MHz), not LPDDR4x, so it's possible we'll see better performance in laptops where the CPU is paired with faster RAM as well.

Better, perhaps, but not significantly. In fact, 1050-1100 is already wide enough gap that I'm willing to bet 4700U will fall into it even with LPDDR4x. None of the 4800U results, even, have been more than ~1200 (GPU, Time Spy). Here's a bench with 4266 mem, GPU score is 1119. And there is a confirmed-ish Fire Strike result for 4800U with LPDDRX4266 mem, got 3900 GPU score whereas MX350 gets about 4400.

4800U with LPDDRX might be a bit faster than MX250, but 4700U won't be as fast as MX350.

specopsFIBetter, perhaps, but not significantly. In fact, 1050-1100 is already wide enough gap that I'm willing to bet 4700U will fall into it even with LPDDR4x. None of the 4800U results, even, have been more than ~1200 (GPU, Time Spy). Here's a bench with 4266 mem, GPU score is 1119. And there is a confirmed-ish Fire Strike result for 4800U with LPDDRX4266 mem, got 3900 GPU score whereas MX350 gets about 4400.

4800U with LPDDRX might be a bit faster than MX250, but 4700U won't be as fast as MX350.

Oh, sorry if I made it sounds as in I was expecting them to beat the MX350, that was not the case. I was thinking about the MX250, but clearly didn't write that out.

TheLostSwedeOh, sorry if I made it sounds as in I was expecting them to beat the MX350, that was not the case. I was thinking about the MX250, but clearly didn't write that out.

You did, originally. Just not in your reply to my post. We seem to be in mutual understanding :)

Berfs1So why purchase an 8c8t processor for content creation? We are talking about the R7-4700U not the R7-4800U, which is an 8c16t processor. The fact that the 4700U doesn't have SMT means it is automatically oriented towards gamers. If this were a 4800U laptop, then sure it would make sense.

None of the U-series of CPUs (Intel or AMD) have been aimed at gaming. These are low-powered CPUs that are good for most productive/work-related tasks. You sound like you’re getting this confused with the H-series.

CheeseballNone of the U-series of CPUs (Intel or AMD) have been aimed at gaming. These are low-powered CPUs that are good for most productive/work-related tasks. You sound like you’re getting this confused with the H-series.

To be precise:
-U SoCs are aimed at varying load, i.e. CPU stays at idle/low most of the time and often boosts.
-H SoCs are aimed at costant load.

There may be no technical differences.
For example: 4800U and 4800H are identical on the CPU side. Base freq is the only difference. So they will likely perform identically in idle and (early) boost. 4800H will boost longer because it'll be put in laptops with more robust cooling.

By principle, both Intel and AMD could be selling just a single mobile SoC lineup, but that would lead to enormous performance variance and make all benchmarks and comparisons pointless.

newtekie1It's only a rebranded MX250 if you also think a GTX 1050 is just a rebranded GT 1030. Using the same architecture doesn't make the cards the same. According to the GPU Database here at TPU, the MX350 is twice as fast as the MX250/GT1030. It's not a totally weak GPU like some people seem to think it is. Yeah, compared to dekstop cards it is, but for a thin and light laptop it's pretty damn powerful.

I think there are a lot of people in this thread that are over-estimating how powerful the AMD iGPU is in the 4700u as well as under-estimating how powerful the MX350 is.

Not sure I follow, how is the comparison valid because GT 1030 is still majorly hobbled by a 64bit wide bus besides having a much lower TDP? Unless you were pointing to something else?

No, it is not ~ the MX250 is a straight up rebrand of MX150 & the MX350, with its 64bt wide bus, is nowhere near 2x as fast as the MX250 until we're just counting theoretical FLOPS? How do I know this ~ because I still own one & for some reason Nvidia decided to sell a "max Q" kind of a version of it with even more anemic clocks later on, without of course disclosing unless you dig deep down into the specs! FYI OEMs are mostly free to set their clocks.

And I'm pretty certain you're also overestimating how powerful MX350 is, given the specs we already know of!

Welp:







From:
Amd/comments/g4xc3v
A rather nonsensical "discrete" GPU combination.

Ryzen 7 4800U is very very fast! o_O


www.notebookcheck.net/AMD-Ryzen-7-4800U-Laptop-Processor-Benchmarks-and-Specs.449937.0.html

medi01Welp:







From:
Amd/comments/g4xc3v
A rather nonsensical "discrete" GPU combination.

4800u has vega 8 gpu and 4700u has vega 7, with fewer shaders(512 vs 448) and lower clocks thus a fair bit slower. But yeah none of them are really tested in playable settings on those charts(expect 1050 on F1 2019).

CheeseballNone of the U-series of CPUs (Intel or AMD) have been aimed at gaming. These are low-powered CPUs that are good for most productive/work-related tasks. You sound like you’re getting this confused with the H-series.

Im honestly just confused why they have a non-SMT CPU for a U series spec, kinda backwards if you ask me (non-SMT should be purely for performance)

ARFRyzen 7 4800U is very very fast! o_O


www.notebookcheck.net/AMD-Ryzen-7-4800U-Laptop-Processor-Benchmarks-and-Specs.449937.0.html

not sure why a 9900KS is slower than a 9900K but ok

Berfs1Im honestly just confused why they have a non-SMT CPU for a U series spec, kinda backwards if you ask me (non-SMT should be purely for performance)

Power, probably. SMT doesn't come completely for free.

londistePower, probably. SMT doesn't come completely for free.

While it is true the maximum power consumption at 100% load will increase when SMT/HT is enabled, power consumption/thread will be greatly reduced. That is the benefit of having SMT/HT, your typical power consumption will actually be lower because the processor takes less energy per thread. For example, 9700K vs 9900K both at the exact same frequency, lets say 5 GHz at the same voltages. When both the 9700K and 9900K are at 100% load, the 9900K will take slightly more power. HOWEVER, let's say you have an application that only loads 8 threads, in that case, the 9700K will be at 100% load while the 9900K will be at 50% load (exclude the windows stuff). In THAT scenario, the 9900K would be taking a little over a half of 9700K's power consumption because the 9900K is loaded to around 50% usage and the 9700K is 100%.

Long story short, here are the advantages:
Non-SMT/HT = Lower max W, better single threaded performance, better per-thread performance
SMT/HT = Lower W/thread, better multi threaded performance, more threads

And when you say slightly, you mean 30-35%? That's the power consumption difference between 9700K and 9900K.
With 8-thread load, their power consumption will be close to identical.