Starfield discussion thread

[Solaris17]

I'm still flabbergasted how people have decided that video is the best to look at graphs.

Why read an article when you can watch a still image at 60fps!

 

[gffermari]

I'm still flabbergasted how people have decided that video is the best to look at graphs.

Yes, something's not right in the graphs.

Starfield GPU Benchmarks & Comparison: NVIDIA vs. AMD Performance - YouTube

It seems there is a massive bottleneck in the HU graphs. Is it the 7800X3D??

 

[Frick]

Yes, something's not right in the graphs.

Starfield GPU Benchmarks & Comparison: NVIDIA vs. AMD Performance - YouTube

It seems there is a massive bottleneck in the HU graphs. Is it the 7800X3D??

View attachment 311873
View attachment 311874

Different systems, plus it's not a benchmark, it's just running around in the game meaning the results will vary. The GPUs are still in the same order though.

 

[Lew Zealand]

Yes, something's not right in the graphs.

Starfield GPU Benchmarks & Comparison: NVIDIA vs. AMD Performance - YouTube

It seems there is a massive bottleneck in the HU graphs. Is it the 7800X3D??

View attachment 311873
View attachment 311874

HUB specifically searched for and tested the most GPU-demanding scene in the first few hours of the game, to see what your GPU's worst-case scenario will be. GN very likely tested in a different area. IOW same as every other time one site's benchmarks don't exactly replicate another's.

 

[Zyll Goliat]

Yes, something's not right in the graphs.

Starfield GPU Benchmarks & Comparison: NVIDIA vs. AMD Performance - YouTube

It seems there is a massive bottleneck in the HU graphs. Is it the 7800X3D??

View attachment 311873
View attachment 311874

Nah it's just a different maps where they did benchmarking test because of that they have totally different results.....also when you are in base or when you are in ship and closed areas FPS are much higher compared to the certain open areas with lot of vegetation then game will hit much harder on your GPU...

 

[yzonker]

It's crazy how demanding the game is. I was walking around Neon and noticed my CPU usage was almost 100%. I do have HT and 8 ecores disabled, which tends to be the best performance in most games, but I haven't tested this one.

I was getting 110-120 fps with DLSS at 67% in 4k max settings (4090). Close to holding the 117 fps limit I have set (LG C1 monitor).

 

[Avro Arrow]

Looking at the Hardware Unboxed video this morning made me look at my collection of video cards in a whole new way because only one is good (7900 XTX), one is decent (RX 6800 XT) and two are barely usable (RX 5700 XT, RX 6600). The rest are just paperweights when it comes to Starfield (although anything below the RX 580 2048SP isn't meant for 3D gaming anyway):

(Absent are my two HD 4870s, my HD 5870 and my HD 7970 because they're too packed away to bother getting and my RX 6500 XT is in my mom's PC.)

 

[Dr. Dro]

Really? No matter what I do, im GPU bound. Im currently at 720p with 50% resolution scale on top of that, still 99% gpu usage. I don't know man, it's extremely GPU heavy

It's reading wrong. Look closely it reports mega high gpu usage but no power consumption to match

No, cyberpunk was not and is not "just bad".

I played it from day 1 and had no issues bar a couple of visual glitches, and the game is immense.

 

[Super Firm Tofu]

@Super Firm Tofu care to confirm with your 13900k? I'm just curious. If you don't care that's fine too.

I trust @fevgatos when it comes to knowing Alder Lake and Raptor Lake.

There's an apparent difference in how 12th and 13th gen handle this game. I can't really swap cards to see if this is down to a difference in GPU drivers, or possibly it's a difference in the Intel thread director, but look how the 13900k handles the P-Core load. It's purely single-threaded, as to where the 12700k spreads the load equally across all available P-Core threads.

13900k/4090:

12700k/6900XT:

 

[Dr. Dro]

I trust @fevgatos when it comes to knowing Alder Lake and Raptor Lake.

There's an apparent difference in how 12th and 13th gen handle this game. I can't really swap cards to see if this is down to a difference in GPU drivers, or possibly it's a difference in the Intel thread director, but look how the 13900k handles the P-Core load. It's purely single-threaded, as to where the 12700k spreads the load equally across all available P-Core threads.

13900k/4090:

View attachment 311887


12700k/6900XT:

View attachment 311888

This difference is likely due to the different behavior of the AMD and the Nvidia graphics drivers, if you swap gpus the outcome should be the same

 

[Super Firm Tofu]

This difference is likely due to the different behavior of the AMD and the Nvidia graphics drivers, if you swap gpus the outcome should be the same

It's possible, but I'm interested in what @fevgatos saw on the 12900k and 4090 combo as they said there was a performance improvement by disabling HT.

 

[Frick]

It's crazy how demanding the game is. I was walking around Neon and noticed my CPU usage was almost 100%. I do have HT and 8 ecores disabled, which tends to be the best performance in most games, but I haven't tested this one.

I was getting 110-120 fps with DLSS at 67% in 4k max settings (4090). Close to holding the 117 fps limit I have set (LG C1 monitor).

How does it look compared to FSR and native?

 

[Dr. Dro]

It's possible, but I'm interested in what @fevgatos saw on the 12900k and 4090 combo as they said there was a performance improvement by disabling HT.

I don't have concise numbers but it ran at least just as well on a pure P-core no HT scenario with my 13900KS, sky high bad speculation rate though

 

[FoulOnWhite]

I trust @fevgatos when it comes to knowing Alder Lake and Raptor Lake.

There's an apparent difference in how 12th and 13th gen handle this game. I can't really swap cards to see if this is down to a difference in GPU drivers, or possibly it's a difference in the Intel thread director, but look how the 13900k handles the P-Core load. It's purely single-threaded, as to where the 12700k spreads the load equally across all available P-Core threads.

13900k/4090:

View attachment 311887


12700k/6900XT:

View attachment 311888

I have a 12700k, still have E cores enabled.. Seems to be running ok with my 3080 at 3440x1440, kinda surprised really. Should i disable the E cores for this game?

 

[Easy Rhino]

duh, i just realized i should use the research bench on the ship to convert a bunch of the crap in storage to useful stuff to be used in mods, etc.

 

[FoulOnWhite]

duh, i just realized i should use the research bench on the ship to convert a bunch of the crap in storage to useful stuff to be used in mods, etc.

Not looked at anything like that yet, just done a couple of the artifact q's

 

[Super Firm Tofu]

I have a 12700k, still have E cores enabled.. Seems to be running ok with my 3080 at 3440x1440, kinda surprised really. Should i disable the E cores for this game?

I haven't tested it myself, but there were some posts in here that said performance was better with them left on.

duh, i just realized i should use the research bench on the ship to convert a bunch of the crap in storage to useful stuff to be used in mods, etc.

I just finally figured that out last night.

 

[Easy Rhino]

I just finally figured that out last night.

Same, after I finished a space battle I could not loot any more from the debris and was like huh? So then it dawned on me I can either jettison a bunch of the stuff I have been pack-ratting around or turn it into actual good stuff. I'm pretty slow.

edit: also on a side note can i say it is hard to know which loot to keep and which really isnt any good. especially the space suits. no idea... i guess you need to have a couple of different kinds depending on the conditions of the planet.

 

[Super Firm Tofu]

Same, after I finished a space battle I could not loot any more from the debris and was like huh? So then it dawned on me I can either jettison a bunch of the stuff I have been pack-ratting around or turn it into actual good stuff. I'm pretty slow.

Not sure if you've seen it, but you can also store the things you've picked up in the ships storage as well. It's accessed by a small panel on the left hand side as you're about to sit in the pilot's seat.

I'm taking it pretty slow too. I'm about 11 hours in and it's all starting to 'click'.

 

[JustBenching]

It's possible, but I'm interested in what @fevgatos saw on the 12900k and 4090 combo as they said there was a performance improvement by disabling HT.

First big city, New Atlantis is called, worst case scenario with default configuration was around 90 fps. That's the lowest fps recorded (not 1% lows). Turning off HT boosted that to around 107-110 fps. Turning off ecores + HT also gave me similar results but 1% lows suffered a lot cause the CPU was running at 100% utilization

 

[FoulOnWhite]

I look at stats on the suits and helmets, and try a keep what "seems" to be the best.

Also where is the merchant in atlantis to sell some crap?

 

[Easy Rhino]

Am I dumb or does the cutter not "cut" into all resources?

 

[FoulOnWhite]

Am I dumb or does the cutter not "cut" into all resources?

I thought that cutter was just a mining tool

 

[Super Firm Tofu]

Am I dumb or does the cutter not "cut" into all resources?

Some of them you have to walk up to and press 'E' to harvest. (water, some plants, etc.)

 

[yzonker]

Well maybe not very memory sensitive unless they were still GPU limited (might be with no upscaling)...

Starfield PC Performance Review and Optimisation Guide - OC3D

Memory Performance Scaling A lot of games benefit from fast memory, and in our demanding city test area, we have found that there is some performance to be gained by playing Starfield with faster DDR5 memory on our i9-13900K and RTX 4090 based test system. That said, these performance gains are...

www.overclock3d.net