Intel Iris Plus Graphics G7 iGPU gives AMD RX Vega 10 a run for its money

Cal Jeffrey

TS Evangelist
Staff member

Update 10/2/19: Questions in the comments were wondering about of memory configuration of the Intel benchmark. According to Notebook Check, the Core i7-1065G7 in the Yoga C940 takes advantage of a dual-channel configuration with 16GB LPDDR4X-3733 onboard. This allowed the full potential of the Iris Plus iGPU to shine through with a maximum "Turbo Boost" of 1100MHz.

They made no mention of the configuration in the Ideapad S540.

Intel's Core i7-1065G7 Ice Lake processor features its Gen11 Iris Plus Graphics G7 iGPU. Notebook Check benchmarked the chip using 3DMark 11 and found a 15- to 25-percent boost in performance when compared to AMD's Ryzen 7 2700U processor, which sports its Radeon RX Vega 10 iGPU.

The rigs used for testing were a Lenovo Ideapad S540-14API with the AMD chipset, and a Lenovo Yoga s940-14IIL equipped with the i7. Both are ultraslim notebooks with the Ideapad going for under $1,000. The Yoga S940 debuted at CES this year and starts at about $1,500.

Along with a few game comparisons, they found that Intel's iGPU has caught up with AMD's RX Vega 10, while surpassing it in the synthetic benchmarks. In terms of actual gaming, the performance margins were much narrower than the 3DMark tests but still just barely gave the G7 the edge.

As mentioned before, a dedicated gamer is going to want a dedicated graphics card. However, for portable gaming on compact notebooks, it is an option for less demanding games running in 720p.

Since the iGPU is built-in to the processor, the price point remains affordable, but, of course, you get what you pay for. In this regard, you are still getting more bang for your buck with AMD's less expensive chipset as far as practical gaming is concerned.

Still, it's nice to see the two companies competing on the integrated graphics front. In the long run, it just means our laptops are going to be better at running those casual games that some of us play in the coffee shop to kill time.

Permalink to story.

 
Last edited:

neeyik

TS Guru
Staff member
One problem with testing iGPUs on ‘ultra’ detail settings is that the graphics used involve a lot of memory read/writes. This means the test ends up being more of a test of the memory subsystem, rather than the GPU itself. I’d prefer to see such tests done on ‘medium’ settings to reduce the dependency on the memory.
 
Yup, the DOTA tests are OK as ~30 FPS is usable in that type of game but really they should be targeting 40-60fps settings to compare. Where an actual user might want to game. And how about a game well-suited to these GPUs like Rocket League or Overwatch? Hell, my kid cut his gaming teeth on an Iris Plus Skylake i5 that could run SW Battlefront (2015) with decent framerates at low settings except when there was too much of a physics demand (asteroid missions).
 

hay fizzy

TS Rookie
One problem with testing iGPUs on ‘ultra’ detail settings is that the graphics used involve a lot of memory read/writes. This means the test ends up being more of a test of the memory subsystem, rather than the GPU itself. I’d prefer to see such tests done on ‘medium’ settings to reduce the dependency on the memory.
I agree to a point no mention of memory configurations was this single channel or duel channel for AMD.
 

Mr Majestyk

TS Maniac
But what about Vega 11 APU's, why are they comparing it to Vega 10. Vega 11 is a fair bit faster and should beat the Intel I would think. Or are they a higher power CPU?
 
But what about Vega 11 APU's, why are they comparing it to Vega 10. Vega 11 is a fair bit faster and should beat the Intel I would think. Or are they a higher power CPU?
Yup, higher power CPU. Vega 11 is in a desktop class APU and Vega 10 is in a laptop-class CPU, like these new Intel chips. Advantage is big for the Vega11 over the 10 as the laptop Vegas are limited by DDR4-2400, while the Desktops can take DDR4-3200 in a decent MoBo.
 
Here's an example using a benchmark which is appropriate for the 2 Vega GPUs being compared (3DMark Night Raid):

https://www.3dmark.com/compare/nr/14172/nr/96683#

This is from a pair of top 10 (but not #1) machines. The biggest difference is 3466 vs. 2400 MHz RAM. The Graphics scores are 61% higher for the Vega 11 (with 3466 RAM) over the Vega 10 (with 2400 RAM). This suggests that memory speed is contributing to much of the difference though I'd love to test the R5 2400G w/Vega 11 restricted to 25W to match the R7 2700U. More data is always better.
 

Mr Majestyk

TS Maniac
Here's an example using a benchmark which is appropriate for the 2 Vega GPUs being compared (3DMark Night Raid):

https://www.3dmark.com/compare/nr/14172/nr/96683#

This is from a pair of top 10 (but not #1) machines. The biggest difference is 3466 vs. 2400 MHz RAM. The Graphics scores are 61% higher for the Vega 11 (with 3466 RAM) over the Vega 10 (with 2400 RAM). This suggests that memory speed is contributing to much of the difference though I'd love to test the R5 2400G w/Vega 11 restricted to 25W to match the R7 2700U. More data is always better.
Thanks for that, yes that would be interesting comparison.
 
  • Like
Reactions: Charles Olson

Evernessince

TS Evangelist
Here's an example using a benchmark which is appropriate for the 2 Vega GPUs being compared (3DMark Night Raid):

https://www.3dmark.com/compare/nr/14172/nr/96683#

This is from a pair of top 10 (but not #1) machines. The biggest difference is 3466 vs. 2400 MHz RAM. The Graphics scores are 61% higher for the Vega 11 (with 3466 RAM) over the Vega 10 (with 2400 RAM). This suggests that memory speed is contributing to much of the difference though I'd love to test the R5 2400G w/Vega 11 restricted to 25W to match the R7 2700U. More data is always better.
It makes you wonder why OEMs don't spend an extra buck or two and get the faster RAM. DDR4 3000 is essentially the same price as 2400 anyways. Heck some OEMs are so daft as to run the system in single channel. It seems to mostly be a thing with OEMs based in the US. I know Huawei have done a good job with their Ryzen APU laptops. They are also one of the few companies using AMD in their more premium laptops. Dell certainly refuses to put AMD into their high-end laptop products (or their servers for that matter as well). Then again we all know what dell has done in the past so this is not surprising.
 
Last edited:

JohnnyStone

TS Enthusiast
Here's an example using a benchmark which is appropriate for the 2 Vega GPUs being compared (3DMark Night Raid):

https://www.3dmark.com/compare/nr/14172/nr/96683#

This is from a pair of top 10 (but not #1) machines. The biggest difference is 3466 vs. 2400 MHz RAM. The Graphics scores are 61% higher for the Vega 11 (with 3466 RAM) over the Vega 10 (with 2400 RAM). This suggests that memory speed is contributing to much of the difference though I'd love to test the R5 2400G w/Vega 11 restricted to 25W to match the R7 2700U. More data is always better.
It makes you wonder why OEMs don't spend an extra buck or two and get the faster RAM. DDR4 is essentially the same price as 2400 anyways. Heck some OEMs are so daft as to run the system in single channel. It seems to mostly be a thing with OEMs based in the US. I know Huawei have done a good job with their Ryzen APU laptops. They are also one of the few companies using AMD in their more premium laptops. Dell certainly refuses to put AMD into their high-end laptop products (or their servers for that matter as well). Then again we all know what dell has done in the past so this is not surprising.
Yes, Dell is just as corrupt as Intel. It is just sad that the majority of people are sheep.
 

Roberto R

TS Rookie
According to AMD’s numbers, the Ryzen 3780U scores 5124 and 1126.5 in 3DMark11 Performance and 3D Mark Timespy.

AMD Ryzen™ 7 3780U Microsoft Surface® Edition Processor
Graphics Model Radeon™ RX Vega 11 Graphics
# of CPU Cores 4
# of Threads 8
Max Boost Clock Up to 4GHz
Base Clock 2.3GHz
Default TDP / TDP 15W
 

Roberto R

TS Rookie
................................................................3DMARK 11 Score / 3DMARK Time Spy Score
AMD Ryzen 7 3780U (Surface Edition) : 5124 / 1123.5
AMD Ryzen 7 3700U : 4432.3 / 969
Intel i7-1065G7 (Ice Lake) : 4910 / 957
Intel i7-8665U : 2019.5 / 455.3
Intel i7-8565U : 2237.8 / 485.8
 

ubronan

TS Rookie
One problem with testing iGPUs on ‘ultra’ detail settings is that the graphics used involve a lot of memory read/writes. This means the test ends up being more of a test of the memory subsystem, rather than the GPU itself. I’d prefer to see such tests done on ‘medium’ settings to reduce the dependency on the memory.
Absolute agree on this statement I actually tested it in my intel nuc and indeed if I can get faster memory to run in the nuc at higher clockspeeds you see a instant better performance. Simply because it runs fully on the memory provides by your systems ram.
I simply did another test and lowered the memory speed to the lowest stable running speed and poof de performance went to the drain. So the internal graphics is heavily impacted by what memory can be used, if its locked at a certain speed you have no choice but if you can run faster memory I would certainly see if its not too expensive todo so.
I say this because you never can be sure that the memory will run faster, example I bought a asus laptop according to the manufactor it would be able to run faster memory but I was surprised to see it only allowed running at the standard 2133 mhz instead of the advertised 4000 mhz. Actually this behaviour can be detected on normal pc as well.
The only answer you will get from the factory support we do not support overclocking.
Funny right because all of these factories use overclocking as their selling point.
 
Last edited: