Carzy Zarx’s PC Gaming Emporium - Catch Up on All the Latest PC Gaming Related News

[Jizz_Beard_thePirate]

Bofferbrauer2 said:

Pemalite said:

The 7600M XT/7700S and 7600M/7600S looks to be a decent budget friendly mid-range chip, should prove to be a decent part in "cheap" gaming notebooks like the Acer Nitro series.
The 7600M XT/7700S has better bandwidth than the desktop 6600XT (288GB/s vs 256GB/s)
The 7600M/7600S has better bandwidth than the desktop 6600 (224GB/s vs 256GB/s)
So that is already a good start...

Although Ray Tracing is where things will either make it or break it.

I was at first concerned seeing that the CU count stayed the same between the 6600M and 7600M models, but this larger bandwidth along with architectural changes and lower TDP (6600M is 100W, 7600M is 50-90W) should give them a small boost in performance and a bit bigger boost in battery life.

Speaking of mobile GPUs, did you guys notice that the 4070 and 4060 actually have less SMs and GPU shaders than their 30xx counterparts? This doesn't really bode very well for their later desktop counterparts, which should be the 4060 and below.

The 4070 mobile comes with 36 SM and 4608 Shaders, while the 3070 mobile had 40 and 5120 respectively.

The 4060 mobile comes with 24 SM and 3072 Shaders, while the 3060 mobile had 30 and 3840 respectively.

The 4050 mobile comes with 20 SM and 2560 Shaders, which just matches those of the 3050Ti and is only 4 SM more than the 3050 mobile.

Also, the memory bus is practically halved, being a 128bit bus for the 4070 and 4060 and only 96 bit for the 4050, resulting in very low bandwidths especially for the 4070.

While the paper specs do look horrid, comparing cross architectures is pretty pointless which you should know considering how horrid the paper specs was with RDNA 2 compared to Ampere. Lovelace is similar to RDNA 2 where it does do more with less. For example, a 4070 Ti has only 7680 shaders and 192 bit memory bus with a bandwidth of 504.2 GB/s vs 3090 Ti that has 10752 shaders and 384 bit memory bus and a bandwidth of 1,008 GB/s. Yet according to techspot aka hardware unboxed, at 1440p, the 4070 Ti is 4% faster than a 3090 Ti:

But similar to RDNA 2, the bandwidth differences comes into affect as you go up the resolutions where a 3090 Ti is now 4% ahead of the 4070 Ti:

But all of which is to say that comparing paper specs across architectures is pretty pointless, especially as Lovelace has a large L2 cache pool which is something that you are leaving out in your comparisons along with being on a much better node and etc.

Last edited by Jizz_Beard_thePirate - on 06 January 2023