Laptop with Server-Like Performance: Serious Power

The advancements of computing technology has been beyond what we expected around 10 years ago. Computer power has grown exponentially for all market segments from personal computing using laptops, business computing using desktops up to enterprise computing using servers installed on datacenters.

This trend, brings so many benefits to users from all segments, specially engineers, designers, architects, analysts, up to hobbyists. Rendering isn't a problem anymore as real-time rendering can be achieved using common gaming notebooks and not locked into mobile workstations or even desktop workstations. And the difference in terms of performance, is getting blurry each day. And now, we'll bring you another interesting experiment to show the potentials of knowing the technology and the guts to break the barrier, and success.

And this, may be your first article, showing how things be done properly, to achieve server-like performance, on a laptop.

High Performance Thin and Light Laptop. - Image Generated by AI

The Precursor
Before the Covid era (2019 - 2022), a manufacturer started to release a laptop, tailored to gamers who like play games anywhere. It was 2011, and the first "true gaming laptop" was born. It was Alienware Razer Blade, released to public, marketed it as "the world's first true gaming laptop" due to its emphasis on a thin, portable design while still delivering strong gaming performance.

For professionals market, manufacturers like HP, Dell and Lenovo, have been providing pro-grade laptops using pro-level and enterprise-level components like Quadro GPU, Xeon CPU, ECC memory, etc., and marketed as mobile workstation. It offers high performance computing, stability and reliability for professionals like engineers, architects and designers, to finish their job, quicker and more efficient at some additional cost, compares to business-grade laptops available in the market.

Seeing that mobile gaming market started to develop showed huge potentials, more manufacturers were starting to produce and manufacture their own gaming laptops. Asus, Gigabyte, MSI, HP and Dell. Ever since, gaming laptops have been evolving from the quality perspective, up to performance level.

Various Laptops - Image Generated by AI

Hipothesis
Aside from the growing level of performance inside a laptop, the amount of  system memory, or RAM, inside laptops have also grown significantly.  We can easily see common amount of RAM listed as 8GB, 16GB, 24GB and even 32GB, installed in latest gaming laptops. And their technical datasheets also list maximum amount of RAM that can be installed. Different laptop, may have different amount of maximum RAM, as officially announced and listed by the manufacturer. And each of the laptop, carries 64-bit CPU, either its from Intel or AMD. And this, how we addressed it in huge curiosity.

As you may not know, 64-bit CPU can take as much as 16 exabytes of memory. That's millions times of Terabyte of RAM. Huge isn't it? So, a 64-bit CPU can easily take 64GB, 96GB or even 256GB, when the slots and memory modules are available. And this was our million dollar question: can our Asus ROG Flow X16 with AMD Ryzen 9 6900HS take more memory than  what stated on its datasheet?

Asus ROG Flow X16 2022 Model - Image Courtesy of Asus Website

The Preparation
Our Asus ROG Flow X16 2022 model technical specification is as following.

• AMD Ryzen 9 6900HS with 8-cores and 16-threads
• Radeon 680m APU with shared memory
• 32GB DDR5 (2x16GB) RAM, populating 2 SODIMM slots
• 2x2TB PCI NVME SSD
• Nvidia RTX 3060 with 6GB discrete GPU

Based on the technical specification, we're preparing 2x48GB of DDR5 SODIMMs, to replace the existing 2x16GB modules.

2x 48GB DDR5 SODIM from Crucial - Real Picture from the Modules

Installation and Memory Training (Post-Preparation)
SODIMM modules replacement was a breeze. The next thing was doing memory training first, before the system accepted the settings. Memory Training is common to latest systems as a low-level procedure during system's POST (Power On Self Test) phase. During this procedure, the system dynamically tested and adjusted memory signal timings, voltages up to data strobe alignments. It did huge read/write operations to address valid signal margins. This phase, will ensure data validity and long term reliability. The memory training took around 3 minutes to complete. Once it's finished, it's ready for further test.

ROG Flow X16 2022 model specification

AMD Radeon 680m APU expanded 16GB of VRAM

As shown, the total system memory is now 96GB with 79.8GB usable, as 16GB of system RAM is being allocated as VRAM for the Radeon 680m. The conversion required carefully settings and adjustments in order to make it properly functioning. However, an interesting reading is, the total video memory is now 56GB. Way beyond even very expensive GPUs that commonly carrying 8GB to 16GB of VRAM.

Now, the ROG Flow X16's ready to perform some tests using some common software being used by professionals.

ROG Flow X16 2022 Model - Ready for the Tests

The Tests
We test most hardwares using some widely available as free and paid software like Blender, FreeCAD, Rhinoceros and even Davinci Resolve Studio. However, this time, we're gonna use Blender and Rhinoceros.

BLENDER
With Blender, HIP support was shown in Preferences window inside Blender and Rendering was a breeze.

Viewport interactive is smooth with the above demo file. Further, we're testing the Flow X16 with Blender Benchmark as well, to see how much VRAM is being used.

Blender Benchmark Final Score

All 3 phases of Blender Benchmark results, show interesting numbers on GPU memory. For casual Blender users, you may consider not to use GPU with less than 8GB of VRAM. The 3rd test shows 7.6GB of VRAM usage, pretty much close to maximum VRAM available on common GPUs. Choosing GPU with 10GB and above would be wise.

And Blender Benchmark Score, shows higher score than any other integrated GPU like Intel Iris Xe Graphics.


RHINOCEROS
To know how it gives advantage to Rhinoceros or not, we decided to test it using 2 methods:

1. Real Rhinoceros viewport rendering 
2. Holomark2 using Rhino 7

REAL RHINOCEROS VIEWPORT RENDERING
16GB VRAM on this test, seems to show overkill amount of VRAM when using Rhinoceros photoreal viewport rendering. The process scored 2.6GB VRAM usage during 21 seconds of rendering task.

Rhinoceros photoreal viewport rendering

The following test is Holomark2, using Rhino 7, as Holomark2 benchmark test isn't 100% compatible yet with Rhino 8.

Holomark2 Benchmark Process

Holomark2 Total Score of Radeon 680m

Interesting result is shown by Holomark2 on Rhino 7. The not all 16GB of dedicated VRAM is recognized. Holomark only capture 4GB of VRAM. So, in order to compare the result, as the ROG Flow X16 is also carrying Nvidia RTX 3060, there's nothing wrong if we did another benchmark with RTX 3060 being used as the GPU.

Seems the RTX 3060 is almost twice as fast, compares to Radeon 680m in Holomark2 benchmark result. However, the amount of VRAM is also 4GB instead of 6GB it carries. But remember, this is Rhino 7, not Rhino 8. Rhino 7 architecture is different in graphics compared to Rhino 8.

Surprisingly, RTX A4000 with 16GB of VRAM is do different, being recognized to have 4GB VRAM. But, the performance is stellar. It scored double than the RTX 3060. Observing the result, it made us curious about the RTX A5000 performance. But we decided to do it sometime in the near future. However, since the RTX A4000 was still connected to ROG Flow X16, we concluded the test with Blender Benchmark, and were surprised with the result.

Blender Benchmark Score of RTX A4000 attached to ROG Flow X16 Ryzen 9 6900HS

A staggering more than 10x faster than the Radeon 680m. It brought us a big smile seeing the result.


Conclusion
Eventhough a single 64-bit CPU can address and utilize 16 EXABYTE of memory, the memory module is nowhere to be found. Many manufacturers list maximum of installable memory inside a computer may be based on the availability of the module itself at the time the laptop was designed. However, through this experiment, even the ROG Flow X16 2022 model is listed to have 64GB of memory top, installing more than that is possible. Also, with the integrated APU that uses system memory, through a careful settings, more system memory (RAM) can be allocated as VRAM to the APU. This will bring flexibility of choices should we require more VRAM when using some certain software.

And with 96GB of memory installed, high computing performance of the AMD Ryzen 9 6900HS, it's considered to be on par with common servers installed at datacenters. The NVME PCIe Gen 4 SSD, allows it to deliver more than 7GB/s bandwidth! That's huge!

So, with this kind of extreme computing power in a backpack, you could travel anywhere similar to bring a heavy and hefty size of a server, in your hand.