Starfield CPU & GPU limited, once again DX12/Vulkan fails to deliver

Page 3 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Tup3x

Golden Member
Dec 31, 2016
1,008
996
136
Also the reason why you see loading screens constantly. The basics haven't really changed since Morrowind and it shows.
 
Reactions: Tlh97 and ZGR

BFG10K

Lifer
Aug 14, 2000
22,709
2,979
126
Generally speaking we expect designers of any sort to take what was good from an earlier design and improve on it, right? DirectX was released in 1995, no doubt DX12 is "based on" it yet no-one is complaining about that, are they?
Yep, exactly. You may as well say Call Of Duty MW3 from 2023 uses the Quake 3 engine because the original COD did.

Bethesda talked about the Creation engine in a past interview and basically very little of original Gamebryo remains. They also mentioned there's no other engine capable of doing what it does: large open worlds with so many interactive objects, scripting customization, and ease of extensive modding.

The basics haven't really changed since Morrowind and it shows.
That's...not true. Morrowind couldn't come close to handling the draw distance of later games. Even OpenMW has trouble if you push cells out too far, especially with shadows. This is still true even after recent improvements which specifically target the problem with batching optimizations.
 
Last edited:

Tup3x

Golden Member
Dec 31, 2016
1,008
996
136
That's...not true. Morrowind couldn't come close to handling the draw distance of later games. Even OpenMW has trouble if you push cells out too far, especially with shadows. This is still true even after recent improvements which specifically target the problem with batching optimizations.
It's very true. Of course they have made improvements but it is very obvious to see what the foundation really is.
 

DeathReborn

Platinum Member
Oct 11, 2005
2,755
751
136
The fact that they have removed things that were in the previous engine tells me that this isn't just CE1 wrapped in DX12. They may have used parts of it but it is fundamentally changed. I posted this video in the PC Games section but here it is for you to see how Fallout 4 (CE1 & Starfield (CE2) compare in Details & Physics (not complete comparison).

 

Iron Woode

Elite Member
Super Moderator
Oct 10, 1999
30,935
12,436
136


A few years ago in this very forum, people were telling us Mantle would allow AMD to dominate the Android platform (LMAO), and that low-level APIs would provide automatic performance gains to even unknown Indie developers just by flipping a simple switch to DX12, with zero effort (LOL).

Instead what we got is upscalers are now mandatory to make games playable, and this is a terrible place to be. This is the result of the collective failure of low-level APIs, along with the push to ray tracing (though the latter doesn't apply to Starfield).

Almost none of these games look significantly better than the best rasterized DX11 titles. And aside from a few outliers like Doom Eternal and Tomb Raider Shadow, the former perform vastly worse than the latter.

We were also told in this very forum "oh, once game engines are rebuilt from the ground up, we'll see the true performance of low-level APIs!" Yet DX12 is now nine years old, and even the brand new Baldur's Gate 3 can recommend switching to DX11 because it often runs better than Vulkan.

I said back then, anyone who thought the average game developer can optimize code better than GPU driver programmers backed by hardware engineers (ya' know, the people that actually build the things) was delusional, and I was right.

This is also exactly why VRAM requirements are ballooning. Because once again, it's lunacy to expect game developers to manage GPU memory better than AMD/NV engineers. In virtually every game that allows DX11 alongside Vulkan/DX12, DX11 uses far less VRAM.
It appears Bethesda asked for AMD's help in June of 2023 for Starfield. The Vulkan code was problematic and needed help to use DX12. Starfield is a bloated mess and it's no wonder it has performance issues.

Nvidia failed to have proper drivers and with an up coming patch from Bethesda it will support DLSS for better performance.

Give the developer some time to fix what should have been done pre-launch. This brings up an annoying trend of studios just releasing titles that aren't ready for prime-time.
 

Mopetar

Diamond Member
Jan 31, 2011
8,004
6,446
136
I think most people are just happy the game isn't a big-riddled mess. After Fallout 76, Starfield looks immaculate and even by usual Bethesda standards it's solid.

Their games have never had particularly good performance though. Fallout 4 ran poorly when it was released. There's even a thread about draw call performance in the CPU forum that still has people post in it.

I'd have to guess that it's just not well optimized from how it stores and accesses data, so unless someone fixes it algorithmically, they're never going to get the level of performance that people are expecting.
 

PJVol

Senior member
May 25, 2020
619
549
136
It's funny that the latest patch 1.7.29 seems to push CPU even more. When running my ref. segment in Akila (very CPU bound) I see 1-2 fps less at 1080p max settings, compared to 1.7.23 version, that is ~ 72 vs 74 fps.

And also, it's quite unusual to say the least to see a difference in only ONE fps when switching to 1440p ))
 
Last edited:

ZGR

Platinum Member
Oct 26, 2012
2,054
661
136
NV has a rather large advantage for Skyrim and Fallout 4 so this is a pretty big deal for Bethesda fans. Great news.

I find Vulkan to have far lower CPU usage than DX11. All the Ubisoft games I own that support vulkan see far lower CPU usage. This is a big deal for people with really slow/old CPU's. We first saw this with Mantle; FX CPU's doing really well for their time.

Then there is X4 Foundations. No other game heats my CPU up to its limit like X4 does. That is a Vulkan game and it is multithreaded really well now. With unlocked power limit, X4 heats up my CPU as much as Cinebench does.

I think it is a bad idea to focus on Starfield as representation of a DX12/Vulkan title. I have never played a Bethesda game that is good out of the box. I find them all horrifically buggy.

I'd revisit Starfield in 3+ years. Fallout 4 is still nowhere near as polished as Skyrim because of the incredible mod community Skyrim has.
 

CP5670

Diamond Member
Jun 24, 2004
5,524
602
126
The graphics in this title are very mixed. The outdoor terrain and spaceship interiors look amazing, while the New Atlantis area looks like a mid-2000s game, especially the foliage and water. It somehow looks worse than the Fallouts. The game runs at 100+fps at most times for me with 85% scaling and the DLSS mod.
 

psolord

Platinum Member
Sep 16, 2009
2,015
1,225
136
I just uploaded a multiple run, on the age old 2500K and a quite newer rx6600. 1080p staight high preset. System for testing mostly, not a proper config. The card works at 1/4 of the bandwidth, plus worse 8b/10b encoding due to pcie 2.0.

(non monetized channel, just for fun)


Quick sum up, for people not liking youtube vids.

Quadruple test, quick space roaming, outdoors fight, indoors fight, a lengthy Akila run through.

1) Space roaming 60fps easy.
2) Outdoors fight. 40-45 fps, high load on both chips.
3) Indoors fight. 60fps with about equal load on both, maybe a little higher on the gpu at times.
4) Akila run. Yeah right. Good luck 2500K. I think I heard it screaming something about my quality as a human being. xD Just a little above 30fps, with 100% cpu load throughout. That must be the worst gaming load I have seen on the 2500K.


As a side note, here are hardware unboxed findings on Akila at high.



The fact that the 4/4 2500k, from January 2011, three years inside PS3's life, manages to come close to the 8/16 Ryzen 1700, impresses me immensely, from the chip side of things and disappoints me immensely from the multithreading side of things.

The fact however, that the 2500K (OC) manages to provide Gen 9 console like quality, even on Akila, while giving X2 performance at indoor fights, I don't know how ridiculous that is.

PS 8GBs vram is enough xD

edit grammar
 
Last edited:

BFG10K

Lifer
Aug 14, 2000
22,709
2,979
126
I think it is a bad idea to focus on Starfield as representation of a DX12/Vulkan title.
I'm not. Starfield is just the latest example of a long list of games that fail to use low-level APIs properly.
 

Panino Manino

Senior member
Jan 28, 2017
846
1,061
136
Interesting article from chips and cheese about starfield GPU utilisation between NV and AMD.

Link.

Seems the game just utilities AMD hardware really well rather than under utilising NV hardware.

People were discussing this at the Beyond3D forum, and I hoped to learn something reading the discussions.
No one can agree with anything, anything from that article, it's conclusions, what happens with NV GPUs.
 

Timorous

Golden Member
Oct 27, 2008
1,727
3,152
136
People were discussing this at the Beyond3D forum, and I hoped to learn something reading the discussions.
No one can agree with anything, anything from that article, it's conclusions, what happens with NV GPUs.

1 person cannot agree and someone from DF talked about lower power consumption on NV cards.

Seems like a lot of people there think MS heavily optimised it for Xbox series consoles and that benefits RDNA3 a lot as well.

Also talk about higher bandwidth L2 in RDNA3 helping a lot.

Not sure your summary that nobody can agree is accurate in the slightest.

Link to B3D discussion. Here is the link to the post where the article was shared so others can read and form their own opinions without needing to find the bit of the thread it was in.
 

ZGR

Platinum Member
Oct 26, 2012
2,054
661
136
I'm not. Starfield is just the latest example of a long list of games that fail to use low-level APIs properly.

If a game runs better on lower end CPU's because of Vulkan, isn't that a success? I only see the benefits of using Vulkan if you have a really slow CPU.

Of all the games I tested, I don't see any performance gains on Vulkan vs DX. If I was using an FX 8350 though, I'd choose Vulkan every time.

Getting 99% CPU usage in Fallout 4 or Skyrim SE is impossible. You will be CPU limited at around 60% usage. This shows a nice improvement in CPU utilization for Starfield.
 

EXCellR8

Diamond Member
Sep 1, 2010
3,982
839
136
I've been running this game on a "temporary" 1st gen Ryzen machine with a 1700 and oof, some of the areas like Akila and NA are super choppy. Smaller indoor areas and space battles are like butter, though.

3440x1440 still doesn't max out my GPU, so definitely looking forward to moving the game to one of the better systems.

EDIT: Just realized this old board has beta BIOS for Ryzen 5000 series... the hell am i even doing with my life? 5700X ordered lol
 
Last edited:

Mopetar

Diamond Member
Jan 31, 2011
8,004
6,446
136
I've been running this game on a "temporary" 1st gen Ryzen machine with a 1700 and oof, some of the areas like Akila and NA are super choppy. Smaller indoor areas and space battles are like butter, though.

3440x1440 still doesn't max out my GPU, so definitely looking forward to moving the game to one of the better systems.

EDIT: Just realized this old board has beta BIOS for Ryzen 5000 series... the hell am i even doing with my life? 5700X ordered lol

I don't know how much of a cost difference it is these days, but I would have tried to get a 5800X3D to drop in since those do so well for gaming.

Of course a 5700X will still be a pretty big upgrade, but if you wanted something long term the v-cache will give the system slightly longer legs.

But yeah, OG Zen is pretty ancient by CPU standards. The cores weren't individually the most powerful (particularly for gaming) but you could get a lot of them for the asking price.
 

coercitiv

Diamond Member
Jan 24, 2014
6,387
12,812
136
I don't know how much of a cost difference it is these days, but I would have tried to get a 5800X3D to drop in since those do so well for gaming.
$190-200 vs. ~$320

We're talking ~60% higher price for ~15% higher perf. It would only be worth it for someone with specific needs, like high refresh gaming or simulators where 1% lows dip bellow 60 FPS.
 

PJVol

Senior member
May 25, 2020
619
549
136
We're talking ~60% higher price for ~15% higher perf
The performance is not just fps numbers in a reviews. I swapped my 5600X for a 5800X3D a couple of months ago (just for the sake of science) and have noticed almost no fps increase in games, except for maybe two or three titles, which is not surprising given the res I play is 1440p.

But what I didn't quite expect which was clearly noticeable, are better UI responsiveness from desktop apps to games, and faster app launch, such as Steam, EGS, etc. Although +2 cores could also contribute to the overall positive experience.
 
Last edited:
Reactions: Grazick

Mopetar

Diamond Member
Jan 31, 2011
8,004
6,446
136
I can't think of a good reason for faster app launching, but the more responsive transition between apps makes perfect sense as the X3D can keep more apps in cache which would make it feel snappier given the CPU doesn't have to load everything from RAM.

I probably should have asked about resolution though. Gaming above 1080p would make any difference between the two processors negligible unless it's something like Flight Simulator or Factorio where the X3D has disproportionately large gains.
 

DAPUNISHER

Super Moderator CPU Forum Mod and Elite Member
Super Moderator
Aug 22, 2001
28,805
21,542
146
Going from a 5800X@5050MHz boost to 3D was definitely not because I was being fiscally prudent.

Nevertheless, gaming on a 1080 60 TV using a frame cap and RX 6800, there have been a couple of games with notable improvement. Spiderman Miles Morales with RT and Gotham Knights are silky smooth now. Was not expecting anything really, since I use a frame cap, but I'll take it.
 

CP5670

Diamond Member
Jun 24, 2004
5,524
602
126
Interestingly, the bad looking parts (cities with lots of NPCs) run poorly due to single thread bottlenecks, while the good looking parts (outdoor landscapes) run well. I'm using one of the DLSS mods and a performance tweaking mod, with a mix of high/ultra settings. The performance drops to 70-80 in areas with lots of NPCs and is generally over 110 otherwise. It's much better than what the reviews show.

Also, the HDR profile in this game is totally messed up by default. Blacks get crushed and look like dark blue. I used Reshade to correct the black level, but the gamma still looks off. Metro Exodus had this problem too.
 
Last edited:

EXCellR8

Diamond Member
Sep 1, 2010
3,982
839
136
Big performance gain moving from the 1700 to the 5700X (not surprisingly). The areas the system struggled with before are much smoother, and the newer CPU also allowed me to run whatever bs memory kit I'm using with this first gen board to run faster, so go figure. I'm gonna play around with some of the display settings a bit more, but the overall experience is so much better now.
 

Genx87

Lifer
Apr 8, 2002
41,095
513
126


A few years ago in this very forum, people were telling us Mantle would allow AMD to dominate the Android platform (LMAO), and that low-level APIs would provide automatic performance gains to even unknown Indie developers just by flipping a simple switch to DX12, with zero effort (LOL).

Instead what we got is upscalers are now mandatory to make games playable, and this is a terrible place to be. This is the result of the collective failure of low-level APIs, along with the push to ray tracing (though the latter doesn't apply to Starfield).

Almost none of these games look significantly better than the best rasterized DX11 titles. And aside from a few outliers like Doom Eternal and Tomb Raider Shadow, the former perform vastly worse than the latter.

We were also told in this very forum "oh, once game engines are rebuilt from the ground up, we'll see the true performance of low-level APIs!" Yet DX12 is now nine years old, and even the brand new Baldur's Gate 3 can recommend switching to DX11 because it often runs better than Vulkan.

I said back then, anyone who thought the average game developer can optimize code better than GPU driver programmers backed by hardware engineers (ya' know, the people that actually build the things) was delusional, and I was right.

This is also exactly why VRAM requirements are ballooning. Because once again, it's lunacy to expect game developers to manage GPU memory better than AMD/NV engineers. In virtually every game that allows DX11 alongside Vulkan/DX12, DX11 uses far less VRAM.

I would had agreed with you on this topic. The whole low level API thing was not going to work. Game developers want to spit out games, not optimize graphic card performance. That is best left to the card manufacturer. Also I can't believe DX12 is nine years old!
 

BFG10K

Lifer
Aug 14, 2000
22,709
2,979
126
Another DX12 game with mandatory upscaling that runs like utter garbage. ~540p medium needs 6700 / 3070 for 60fps even without ray tracing:


"Just flip a simple switch to DX12 and all games will automatically run faster and get a benefit!" LMAO.
 
sale-70-410-exam    | Exam-200-125-pdf    | we-sale-70-410-exam    | hot-sale-70-410-exam    | Latest-exam-700-603-Dumps    | Dumps-98-363-exams-date    | Certs-200-125-date    | Dumps-300-075-exams-date    | hot-sale-book-C8010-726-book    | Hot-Sale-200-310-Exam    | Exam-Description-200-310-dumps?    | hot-sale-book-200-125-book    | Latest-Updated-300-209-Exam    | Dumps-210-260-exams-date    | Download-200-125-Exam-PDF    | Exam-Description-300-101-dumps    | Certs-300-101-date    | Hot-Sale-300-075-Exam    | Latest-exam-200-125-Dumps    | Exam-Description-200-125-dumps    | Latest-Updated-300-075-Exam    | hot-sale-book-210-260-book    | Dumps-200-901-exams-date    | Certs-200-901-date    | Latest-exam-1Z0-062-Dumps    | Hot-Sale-1Z0-062-Exam    | Certs-CSSLP-date    | 100%-Pass-70-383-Exams    | Latest-JN0-360-real-exam-questions    | 100%-Pass-4A0-100-Real-Exam-Questions    | Dumps-300-135-exams-date    | Passed-200-105-Tech-Exams    | Latest-Updated-200-310-Exam    | Download-300-070-Exam-PDF    | Hot-Sale-JN0-360-Exam    | 100%-Pass-JN0-360-Exams    | 100%-Pass-JN0-360-Real-Exam-Questions    | Dumps-JN0-360-exams-date    | Exam-Description-1Z0-876-dumps    | Latest-exam-1Z0-876-Dumps    | Dumps-HPE0-Y53-exams-date    | 2017-Latest-HPE0-Y53-Exam    | 100%-Pass-HPE0-Y53-Real-Exam-Questions    | Pass-4A0-100-Exam    | Latest-4A0-100-Questions    | Dumps-98-365-exams-date    | 2017-Latest-98-365-Exam    | 100%-Pass-VCS-254-Exams    | 2017-Latest-VCS-273-Exam    | Dumps-200-355-exams-date    | 2017-Latest-300-320-Exam    | Pass-300-101-Exam    | 100%-Pass-300-115-Exams    |
http://www.portvapes.co.uk/    | http://www.portvapes.co.uk/    |