Vega iGPU Gaming Comparison – 12nm vs 7nm (4900HS vs 3750H)
AMD noted that, rather than using their latest Navi architecture with Ryzen 4000, they would instead use an improved 7nm version of Vega. AMD, determined that Navi graphics would not Have the necessary optimizations ready in time, so they decided to stick with the Vega architecture and improve it.. One of the big differences with 7nm Vega, of course outside of the benefits going from 12nm to 7nm, is that they actually cut back the number of compute units, but instead boosted up the clock speed. According to AMD. These improvements result in 25, higher peak graphics, clocks, 2 times wide data fabric interface for power, efficient data transfer, graphics, low power state for transition, optimization and 77 higher peak memory bandwidth. As the Vega graphics, essentially utilize. The laptops DDR4 memory, this improved memory bandwidth, would in part be from AMDs move from DDR4 2400 with Ryzen 3000 to DDR4, 3200 or LPDDR4X 4266 with Ryzen 4000.. All things considered, the end result is that per compute unit. The new 7nm Vega ends up being 59 faster when compared against 12nm Vega from the last generation. According to AMD., Lets put this to the test by comparing two laptops. Ill, be using the MSI Alpha 15, with the Ryzen 7 3750H. A 35 watt part, which represents the best mobile chip AMD, were offering from the previous generation, and this has older, 12nm Vega. Ill. Be comparing with the ASUS G14 with the Ryzen 9 4900HS again a 35 watt part and is one of the best Ryzen processors from the new generation and makes use of the newer 7nm Vega.
. I thought this would be an interesting comparison, because both processors have the same 35w TDP that and I dont have a 4900H at the moment. However, as we can see here, both the H and HS have the same number of compute units and graphics frequency anyway., With all of that in mind, lets get into the gaming benchmarks.. Both laptops were tested with the discrete GPU disabled through device. Manager. Lets start out with CSGO, as this is a game you can actually get away with playing on the Vega graphics. At 1080p., Ive got the older 12nm Vega represented by the red bars and the newer 7nm Vega shown by the purple bars. At all setting levels. Its a clear win for the newer chip. However, the performance difference is larger at lower settings where the 4900HS was 45 faster than the 3750H. At lower settings. There will likely be a greater difference, as these tests are also CPU bound and the 4900HS destroys the 3750H in CPU bound tasks. At max settings. Where were hopefully more GPU bound, the new Vega graphics were 38 faster.. Dota 2 was also performing a fair bit better with the new 7nm Vega graphics, from the 4900HS. At ultra settings, where were likely to be more GPU bound, the new Vega graphics were 18 faster than the older 3750H. At low settings, though the new chip was a Much higher 38 faster, but again this is more likely to be showing differences between CPU performance rather than Vega.
Specific performance. Fortnite was tested using the replay feature with the exact same replay on each laptop. At minimum settings. The newer 4900HS was a massive 67 faster. Even the 1 low performance was dominating the average from the 3750H, but yeah again, thats, probably more of a CPU focussed test. At epic settings. Neither were offering a playable experience, however. The new graphics were still 33 faster.. The new graphics played quite well at medium settings, which was averaging above 60 FPS and for context. The old graphics were similar with low settings. Overwatch was tested in the practice range, so the FPS is a little above actual gameplay, but it allows me to perform the exact same test pass on each laptop.. In this test. I found low and medium settings to play. Well enough with the 3750H. Meanwhile, though, the 4900HS was outperforming both even at high settings. At epic, where were presumably most GPU bound, though the newer Vega graphics were 32 faster.. I found it particularly interesting that even the 1 low performance from 7nm Vega was higher than the average FPS from 12nm Vega at all, but medium settings where it was very close.. Ive also tested a few other games. As I wanted some more data, however, I only tried these at the lowest setting presets, as in many cases they were hardly even playable there. Rainbow. Six Siege was tested with the games built in benchmark using Vulkan. Ive only tested low settings here, but the 1 low from 7nm Vega was ahead of even the average from 12nm Vega, while the average FPS was 49 faster but again lower settings.
So a chunk of this would be down to CPU performance differences, rather than explicitly Vega. PUBG was tested using the same replay on each machine with very low settings. The 4900HS laptop was a massive 55 faster in terms of average FPS. But again, a lot of that is probably not specific to the Vega difference. Shadow of the Tomb Raider was tested with the games benchmark at lowest settings where the 4900HS was 35 faster, but I wouldnt say either are playable with these results.. The Witcher 3 also saw a very similar difference with the 4900HS 34, faster in average frame rate when compared to the 3750H.. These are the differences over all 8 games tested with the minimum setting preset. On average. The Ryzen 9 4900HS was 44 faster than the Ryzen 7 3750H in terms of average FPS., As mentioned throughout the video, as this is with the lowest possible presets. A fair bit of the performance difference is likely down to other differences, such as the improvements with Zen2 over the older Zen Heres. A quick look at the 1 low difference on average results were similar, granted its not directly comparable, as this graph doesnt account for Shadow of the Tomb Raider, because I cant get 1 low data out of that test.. These are the performance differences between the four esports titles, with the highest setting presets in use. In theory, were more GPU bound here, so were probably looking more at the differences between 7nm and 12nm Vega here.
, Assuming so on, average 7nm Vega is almost 30 faster when Compared to 12nm Vega, unfortunately, I dont have more data, as most other games are honestly too heavy to be run at max settings 1080p.. I could have tried 720p, but I didnt test it out prior to sending the G14 off to Hardware Unboxed.. I only had a brief look at power draw from the wall in CSGO with max settings. The G14 was using around 83 watts, while the Alpha 15 was using 78 watts. So 6 more power seems fair for the large performance boost in that game, though there could also be other differences at play as they are different laptops.. I really doubt that well see a laptop with a 4900HS that doesnt have some form of Radeon or Nvidia, discrete graphics, so its probably fairly unlikely youll be using the Vega graphics in the 4900HS for gaming. So I guess that makes this more of a for science test to see what the differences are between best case last gen and best case current gen.. I think with Vega graphics, less demanding titles are more likely to be played anyway, and I think a 30 performance jump in one generation is fairly impressive., Assuming the gains are similar on lower end laptops, where Vega is actually more useful. This could certainly offer a playable experience for thinner and lighter laptops in many titles.. It will be interesting to see how the lower powered Vega parts stack up.
Hopefully I can get my hands on all the variants and do a big performance comparison in future. Let me know if youd be interested in something like that Im honestly, not sure how many of you guys only game on the integrated graphics compared to discrete graphics..