News Cyberpunk 2077 update 2.2 claims to improve Arrow Lake performance by up to 33%, theoretically matching the Ryzen 7 7800X3D

[Admin]

Cyberpunk 2077 has received a surprise update that includes Arrow Lake optimizations that improve performance by up to 33%.

Cyberpunk 2077 update 2.2 claims to improve Arrow Lake performance by up to 33%, theoretically matching the Ryzen 7 7800X3D : Read more

 

[80251]

Kudos to CD Projekt Red for bothering to do this. Intel should give them a medal.

 

[atomicWAR]

I'll be curious to see if other titles follow suit with a similar uplifts. It might change my opinion, some, about Intel at the moment. Problem is after the unmitigated disaster that was Raptor Lake it will take a couple of gens before I trust Intel again to deliver a solid product that won't fail two days after the warranty runs out or sooner. Rewind to Sandy Bridge...I never would have thought Intel would be in such a bad spot but here we are. My how times change.

 

[Geef]

Now Intel is going to start advertising that their new chips give a up to 33% boost to gaming!

 

[AkroZ]

Kudos to CD Projekt Red for bothering to do this. Intel should give them a medal.

This update has probably been paid by Intel, you normally don't work on optimizing a game for a specific hardware when you have announced the end of support.
But TH didn't do a new benchmark, thoses changes should affect all CPU.
They most likely reduced the threads memory heap to not be higher than Intel 14th gen specifications (24MB L3 cache for Intel 14600K, 32MB for AMD 9700X), if this is the case then Ryzen 9800X should have similar performances than Ryzen 9800X3D with the new version of Cyberpunk 2077.

 

[rluker5]

1. The previous results aren't too bad for e-cores. and
2. I'm glad I turned off auto update so my mods don't stop working.

 

[thestryker]

This was one of those outlier titles where something seemed to just be wrong given the amount of regression there was. I wouldn't be surprised if other titles have similar issues with how threading is being done given the lack of HT on ARL. I wonder if any of this is going to be possible to control from Intel's side of things (or Microsoft) as many titles, especially big ones, effectively just get abandoned after a while.

It does seem like there are some fairly extreme outliers still:

Spoiler: Results screenshot from Latest HUB 45 Game 285K v 14900K Test

Taken from the following video:

View: https://www.youtube.com/watch?v=y4nC0e32n0s

 

[umeng2002_2]

This update has probably been paid by Intel, you normally don't work on optimizing a game for a specific hardware when you have announced the end of support.
But TH didn't do a new benchmark, thoses changes should affect all CPU.
They most likely reduced the threads memory heap to not be higher than Intel 14th gen specifications (24MB L3 cache for Intel 14600K, 32MB for AMD 9700X), if this is the case then Ryzen 9800X should have similar performances than Ryzen 9800X3D with the new version of Cyberpunk 2077.

Unless that specific "hack" is disabled on anything other than Arrow Lake CPUs.

 

[YSCCC]

And it would be interesting to see if the same patch just happen to boost RPL and maybe AMD also...

It could be that they've addressed an arrow lake specific bug, or just changed some codes that accidentally just boosted everything else also (maybe to a smaller degree), just happens to market it as an arrow lake fix

 

[tamalero]

Unless that specific "hack" is disabled on anything other than Arrow Lake CPUs.

I still remember when many programs back then where Pentium D was struggling to fight the Athlon X2..
were paid by intel to have optimizations that would stop multithreading if a non intel chip was present.

 

[thesyndrome]

This update has probably been paid by Intel, you normally don't work on optimizing a game for a specific hardware when you have announced the end of support.

It's possible, but CDPR have been very good about trying to update the game for various hardware over the years for various different companies, like adding SMT for AMD CPUs, or the update before this that added the latest version of FSR with frame gen.

Also this Arrow Lake boost doesn't seem to be the centrepiece of the update, but more of an added bonus they threw in; the rest of the update is stuff they teased before, like having Johnny appear in your car, and the ability to customise cars paintwork and "steal" patterns from any car you drive past on the street).

I will fully admit that I am biased towards CDPR after being a fan of theirs since 2009, but I'm too old and cynical to be a fanboy, so whilst I honestly believe that they optimised for Arrow Lake off their own backs without prompting from Intel (based on a history of this kind of optimisation and bug fixing they have done for various games over the years), I do agree that it is a possibility, though I do wonder if Intel would really pay CDPR to update a nearly 5 year old game to improve the FPS on it, rather than focusing their attention (and payments) for more recent games that might be experiencing the same issues.

 

[stakanov]

https://www.pcgamesn.com/intel/cyberpunk-2077-arrow-lake-benchmarks

As it turns out, 33% is an understatement of the effect of this update. After dropping down to 1080p at the High preset, with no ray tracing or DLSS, putting as much of a burden on the CPU as possible, the performance increase is an incredible 91%.

woah, insane. is this source credible?

 

[abufrejoval]

I wonder if at one point such CPU specific optimization will result in a regression on another architecture. And here we thought we only had different GPUs architectures to worry about...

It would be interesting to know just what sort of things they did and if that included playing specifically with the memory placement of data to fit cache lines etc.

Thinking on how even on a Raspberry PI using a NUMA library preset would influence RAM bandwidth in a system that theoretically has very uniform RAM, which internally naturally is organized in various hierarchical ways that make it in fact perform rather less uniform one would think.

 

[thestryker]

https://www.pcgamesn.com/intel/cyberpunk-2077-arrow-lake-benchmarks

woah, insane. is this source credible?

Yeah they're reliable on the less in depth side of things. If you look at the results it's obvious the game was hitting E-cores as the more CPU bound test saw a big jump and the other put it right behind the 14700K. This isn't particularly surprising as when Jay (of JayzTwoCents) did some testing found that overclocking E-cores could provide performance increase in some games. If I had to hazard a guess CDPR made adjustments to minimize what's running on the E-cores.

 

[umeng2002_2]

I think it's good that hardware companies pay software companies to use their hardware better, but I don't like it if the software devs lock off that optimization from other hardware, especially if it could improve performance on other hardware/

Ideally, the software team should be the responsible party here and make sure it runs well on most systems.