NVIDIA GeForce MX350 And MX330 Entry-Level Discrete GPUs Benchmarked – Green Team Tackles Intel/AMD Integrated GPUs
NVIDIA will soon be launching updated entry-level discrete GPUs known as the GeForce MX350 and the GeForce MX330 for notebooks. The mobility GPUs will be the most entry-level chips that will be pitted against integrated graphics solutions from AMD and Intel.
NVIDIA's GeForce MX350 and GeForce MX330 Mobility Discrete GPUs Benchmarked - Green Team Shows Why You Still Want a Discrete GPU Inside Your Notebook
The world of mobility graphics has been moving at an accelerated pace recently. Both Intel & AMD are investing in better graphics architecture for their integrated GPUs which power notebooks and laptops. AMD recently unveiled its Renoir Ryzen 4000 lineup which offers a big leap in GPU power by utilizing 7nm Vega cores while Intel has been doing an outstanding work updating its entire line of mobility processors with the new Xe graphics core which would be featured on Tiger Lake CPUs this year, delivering a 2x performance uplift over existing Ice Lake Gen 11 GPUs.
NVIDIA's discrete GPU lineup on the high-end and mainstream spectrum isn't going to worry much about these integrated GPUs as they offer a performance of a whole different level (while costing more and consuming more power, of course). But at the same time, NVIDIA's entry-level GPUs that are part of the MX series definitely feel threatened but integrated graphics coming close to their performance tier.
There are a few reasons for that, firstly, at the pace at which integrated graphics are evolving, they would end up being almost as fast as NVIDIA's entry-level mobility GPUs, and that ends up eating up their market share in the entry-level notebook segment. Plus, while there's always the added benefit of having a discrete GPU onboard a laptop since it can offload much of the tasks from an integrated GPU over to itself, the discrete GPUs require additional power input and that would lead to lower battery times, which is a crucial factor on mobility devices. So for many, a laptop with an integrated GPU having similar or even slightly slower performance would be a better option than buying one with a discrete GPU.
NVIDIA GeForce MX300 Mobile GPU Family Specifications
As such, we knew it from last month that NVIDIA will be updating its GeForce MX lineup. The lineup would still utilize the Pascal GPU core but would get two variants, the MX350 and the MX330. It looks like the two chips will be using different Pascal GPUs, the MX350 would make use of GP107 and the MX330 would make use of GP108.
There's also the GeForce MX310 which is the entry-level variant in the MX 300 series family. The GeForce MX350 would feature 640 CUDA cores and 7 Gbps memory (up to 4 GB) along a 64-bit bus, the GeForce MX330 would feature 384 CUDA Cores and 7 Gbps memory (up to 4 GB) along a 64-bit bus while the MX310 would feature 256 CUDA cores and the same memory configuration.
The interesting part is that the GeForce MX350 would come in two variants, a 25W, and a 15W SKU. The placement of these SKUs would depend upon the OEM's decision and the design that they will be targetting the GPUs at. The 25W GeForce MX350 will feature clocks of up to 1468 MHz while the 15W variant will feature clocks of up to 937 MHz. That's a large gap which would result in a huge performance variation between the two chips. The GeForce MX330 would feature clocks of up to 1594 MHz while the GeForce MX310 would feature clocks of up to 1379 MHz.
NVIDIA GeForce MX Series GPU Family Specifications:
|GPU Name||GPU Architecture||CUDA Cores||GPU Clock||Memory Speed||Memory Bus||TDP|
|GeForce MX570||Ampere GA107||2048||TBA - 1155 MHz||12 Gbps GDDR6||64-bit||25W|
|GeForce MX550||Turing TU117||1024||TBA - 1320 MHz||12 Gbps GDDR6||64-bit||25W|
|GeForce MX 450||Turing TU117||896||540 - 1575 MHz||10 Gbps GDDR6||64-bit||25W|
|GeForce MX 430||Turing TU117||896||TBA - TBA MHz||10 Gbps GDDR6|
7 Gbps GDDR5
|GeForce MX 350||Pascal GP107||640||1354-1468 MHz||7 Gbps GDDR5||64-bit||25W|
|GeForce MX 350||Pascal GP107||640||746-937 MHz||7 Gbps GDDR5||64-bit||15W|
|GeForce MX 330||Pascal GP108||384||1531-1594 MHz||6/7 Gbps GDDR5||64-bit||25W|
|GeForce MX 310||Pascal GP108||256||1341-1379 MHz||6/7 Gbps GDDR5||64-bit||25W|
|GeForce MX 250||Pascal GP108||384||1518-1582 MHz||6/7 Gbps GDDR5||64-bit||25W|
|GeForce MX 250||Pascal GP108||384||937-1038 MHz||6 Gbps GDDR5||64-bit||10.5W|
|GeForce MX 230||Pascal GP108||256||1518-1531 MHz||6/7 Gbps GDDR5||64-bit||25W|
|GeForce MX 150||Pascal GP108||384||1468-1531 MHz||6 Gbps GDDR5||64-bit||25W|
|GeForce MX 150||Pascal GP108||384||937-1038 MHz||5 Gbps GDDR5||64-bit||10W|
|GeForce MX 130||Maxwell GM108||384||1122-1242 MHz||6 Gbps GDDR5||64-bit||25W|
|GeForce MX 110||Maxwell GM108||256||963-993 MHz||1.8 Gbps DDR3||64-bit||10W|
NVIDIA GeForce MX300 Mobile GPU Family Benchmarks
According to Zhuanlan (via Videocardz), the performance of the GeForce MX350 was measured in various Synthetic benchmarks. The results show that the GeForce MX 350, as it is, would be at least 40% faster than the Intel Core i7-1065G7 which rocks an Intel Gen 11 GPU. The surprising thing is that the chip can be further overclocked to reach speeds of up to 1.8 GHz on the core and 8 Gbps on the memory, leading to almost 90-95% performance level of the GeForce GTX 1050 mobility which is impressive considering the latter is a 40-50W TDP chip.
The extra overclock would result in more power consumption but it would not be hitting the same numbers as the mobility GTX 1050. The stock chip is almost 35% faster than the Ryzen 7 3750H with 640 cores but with Ryzen 4000 near the horizon, we can see the lead would drop down to 20-25%, depending on the optimizations for Ryzen 4000 GPU.
According to this table (CC to https://t.co/Vg27I5PX0H) 4800U's 8CU iGPU is either matching GTX 960M or just below MX350 at 1.6GHz in 3dmark 11 Performance https://t.co/vQWTuuaGdi pic.twitter.com/YRJI2KKbf3
— _rogame (@_rogame) February 11, 2020
In one comparison made by our friend _rogame, it is shown that the Ryzen 7 4800U is able to score up to 6121 graphics score in 3DMark 11 performance benchmark which puts it right next to the MX350 with 95% performance level of NVIDIA's discrete level GPU. This is the thing that we have been saying all along that if users will be getting similar performance in an integrated GPU, they won't feel the need to buy a discrete GPU option which would also cost more.
NVIDIA GeForce MX350 Synthetic Benchmarks (Courtesy: NotebookCheck)
Compared to its predecessor, MX350 is up to 20% faster on average than the MX250. But in most scenarios, the GeForce MX300 series will take a huge hit due to its bandwidth starved design and lower ROP counts which put them in much pressure in graphics demanding workloads. Additional benchmarks have been posted by NotebookCheck which shows that the GeForce MX350 is much faster than the GeForce GTX 960M mobility GPU and blazes past the older MX250.
NVIDIA GeForce MX350 Synthetic and Gaming performance benchmarks from a 10th Gen Lenovo laptop. (Image Credits: Expreview)
Rounding it all up, it feels like NVIDIA has definitely felt the heat from all of the situation in the mobility GPU department. While they won't face much competition in the high-end department, they might not retain the same grip on the entry-level or mass majority of the market in a few years unless they start outing more performance efficient chips. It looks like NVIDIA is already planning ahead and has a Turing based GeForce MX successor planned, as per the reports by NotebookCheck.
The information says that a TU117 GPU based GeForce MX400 lineup is being considered and would be positioned against AMD's Renoir Ryzen 4000 and Intel's Tiger Lake Xe GPU powered integrated graphics.
The new lineup would feature several variants and possibly move over the GDDR6, eliminating much of the bandwidth woes that we are looking at today. NVIDIA also plans to introduce PCIe Gen 4.0 support for these discrete GPUs and would offer performance close to a GeForce GTX 1650 in 15W and 25W tiers. The GeForce GTX 1650 itself is likely to get an updated mobility outing soon along with SUPER series RTX standard and Max-Q parts by the second quarter of 2020.