CPU Limited games people would like in the next CPU rebench.

[Melvis]

IF the game only uses 1 or 2 cores its not CPU bound its just shit, poorly optimised...looking at you Cities Skylines

 

[btk2k2]

That's my biggest issue with these tests that probably should be called "synthetic". Also is it reasonable to adjust the test suite to cherry pick workloads that show "interesting" numbers or even best-case results?

Depending on the amount of work you want to be doing you could have a core test suite and a section for 'interesting oddities' or something to deal with best case scenarios, worst case scenarios, things you thought would work but didn't, things you didn't think would work but did.

IF the game only uses 1 or 2 cores its not CPU bound its just shit, poorly optimised...looking at you Cities Skylines

Sure but it is a game that exists and that people play and struggle with performance wise for big cities. Probably more relevant to test games that struggle to hit 60 fps due to CPU limits than to test games that sail past 200 fps in a CPU benchmarking suite.

Nothing is going to change how Cities Skylines behaves so calling it poorly optimised is true but also irrelevant to people who play it. Same with a lot of other games in this list.

Then why use any game in a benchmark if it doesn't matter wether the result is indicative of real world performance or not?

You cannot test every game so for most people reviews are just indicative of performance based on similar ish titles. For example if you test Stellaris it is very likely that you will see similar performance deltas in other Paradox Grand Strategy titles like CK3, HoI4, EU4 etc.

Also how often do patches make real performance differences outside of the 1st few for a AAA title? It does happen for some games like Stellaris due to the long legs but updating for each expansion is probably a good balance between indicative performance and not too much workload.

 

[Ruru]

Lightning Returns (I can provide save file), go to fully populated Ruffian town. Expect a stutter fest which is primarily bottlenecked by rendering thread busy constantly dumping and reloading textures.

Ideally poorly optimised games should get more representation in these tests.

You mean the FF XIII sequel? Is that so badly optimized, that's a X360/PS3 era game..

 

[bug]

Then why use any game in a benchmark if it doesn't matter wether the result is indicative of real world performance or not?

The result is indicative of the real world performance. It's just not the performance of the latest version. It's the same for any game tested.

 

[Frick]

The result is indicative of the real world performance. It's just not the performance of the latest version. It's the same for any game tested.

Some time ago Stellaris got a patch that boosted the performance massively. Night and day difference.

 

[8tyone]

What about Ryujinx? Though not sure how hard it is on the cpu, as it runs great on my 12700k

The one that gives consistent results and is benchmarkable (Switch emulators), I am not knowledgeable enough to know which one (Yuzu, CEMU, Ryujinx ?). Also, have you tried rpcs3 with your cpu?

 

[chrcoluk]

You mean the FF XIII sequel? Is that so badly optimized, that's a X360/PS3 era game..

The second sequel.

Ironically it progressively got worse.

FF13 runs pretty well even at 4k 60fps rendering resolution, just crystarium lags, fully fixed with custom FF13 driver.
FF13-2 introduces stutters from asset loads, is weird as this is less demanding it no longer has enemies wondering the field as well as lower quality textures, but when a fight is about to spawn it will stutter as its streaming in the textures. It also has weird issues with custom 60fps driver.
Lightning Returns aka FF13-3, in empty areas its fine, but in populated towns it can get very stuttery, the graphical quality of the field area is the lowest of all 3 games, however NPCs and lightning herself are clearly a higher texture resolution, the enemies are also visibly higher quality. Its from the era you said and those consoles have a pittance of memory hence the agressive texture swapping, sadly the PC port was never optimised for the much higher VRAM available on PC, and for some reason the texture loading saturates powerful high end cpu cores (single threaded).

Maybe with 20th gen intel core cpu it might have been ok lol, but meltdown/spectre mitigations are really felt in this game so maybe in 2040 with mitigations cpu's can handle the game.

 

[HD64G]

With the upcoming GPUs from AMD and nVidia, many games that seem to be GPU limited will then become CPU limited. My 5c.

 

[Ruru]

The second sequel.

Ironically it progressively got worse.

FF13 runs pretty well even at 4k 60fps rendering resolution, just crystarium lags, fully fixed with custom FF13 driver.
FF13-2 introduces stutters from asset loads, is weird as this is less demanding it no longer has enemies wondering the field as well as lower quality textures, but when a fight is about to spawn it will stutter as its streaming in the textures. It also has weird issues with custom 60fps driver.
Lightning Returns aka FF13-3, in empty areas its fine, but in populated towns it can get very stuttery, the graphical quality of the field area is the lowest of all 3 games, however NPCs and lightning herself are clearly a higher texture resolution, the enemies are also visibly higher quality. Its from the era you said and those consoles have a pittance of memory hence the agressive texture swapping, sadly the PC port was never optimised for the much higher VRAM available on PC, and for some reason the texture loading saturates powerful high end cpu cores (single threaded).

Maybe with 20th gen intel core cpu it might have been ok lol, but meltdown/spectre mitigations are really felt in this game so maybe in 2040 with mitigations cpu's can handle the game.

Ah, never played that one much. Platinum'd XIII on PS3 (played it thrice; on PC, PS3 & X360) and finished XIII-2 twice, Lightning Returns felt so weird that haven't got into that yet.

I guess I'll try some day how it runs on a R5 3600.

 

[kapone32]

Give me some Strange Brigade or Sleeping Dogs

 

[Deleted member 24505]

Also, have you tried rpcs3 with your cpu?

Tried it earlier with fallout new vegas. Ran perfectly at 30fps

 

[AnotherReader]

Here is a writeup with some more CPU intensive games.

They also test Stellaris and make an important note that a fixed time period to test how many days pass is probably a bit unfair on faster CPUs because as each day passes the simulation gets more complex so each additional day takes slightly longer. It would probably therefor be better to do a time to X days and then you can average that out to create a days/s chart like.

It would be so interesting to see how these parts compare to ADL or how it handles P vs E cores and other such tests that someone without the hardware just cannot do.

That is a great find. The Civilization VI turn times are interesting as they buck the trend by not benefiting at all from the larger cache.

 

[R-T-B]

Kerbal Space Program is essentially the premiere single-thread CPU pig on the market, FWIW.

Even in stock, build a 200 part craft, turn on the engines and watch the physics engine chug along at low 30FPS with one core pegged 100%. Yes, that's with my 5950X

 

[ir_cow]

Rpcs3 is great emu. Massvie overhaul made a ton of games work. But how do you test a emu besides running a game.

 

[Ruru]

Rpcs3 is great emu. Massvie overhaul made a ton of games work. But how do you test a emu besides running a game.

Yeah, Skate 3 runs well too

 

[W1zzard]

But how do you test a emu besides running a game.

I would run a game and somehow measure the FPS, or is that the wrong approach for emulation?

 

[Ruru]

I would run a game and somehow measure the FPS, or is that the wrong approach for emulation?

IIRC Afterburner works with it fine with its OSD.

 

[Deleted member 24505]

IIRC Afterburner works with it fine with its OSD.

That is what i used, fps stayed at 30 fps.

 

[Ruru]

That is what i used, fps stayed at 30 fps.

Yep, same here when I've played Skate 3 with PS3 emulator.

 

[Deleted member 24505]

Yep, same here when I've played Skate 3 with PS3 emulator.

What does it run like with a lesser than 12700k CPU? is it hard to run? I don't know if it is easy or if the 12700k runs it well.

Does it use multiple cores?

 

[Ruru]

What does it run like with a lesser than 12700k CPU? is it hard to run? I don't know if it is easy or if the 12700k runs it well.

Does it use multiple cores?

Hard to say, all I can say is that R5 3600 runs it fine.

This is still more than a good CPU.

 

[Vayra86]

That's my biggest issue with these tests that probably should be called "synthetic". Also is it reasonable to adjust the test suite to cherry pick workloads that show "interesting" numbers or even best-case results?

Its interesting because it highlights differences between CPUs that can and usually do tell a story about which CPUs are faster. And with the newer and different architectures, big little vs chiplet and different core complexes, Id say those are all the more interesting to research. For most games the gap isnt visible because calculations are realtime and adjusted to graphics load but in the case of turn based and simulation, you can go way beyond those managed boundaries (big city, end game turns etc).

Perhaps this warrants not continuous testing but a single article once in a while much like the Pcie scaling ones?

 

[ir_cow]

I would run a game and somehow measure the FPS, or is that the wrong approach for emulation?

Not the wrong approach but what game and resolution? I was looking into playing my PS3 collection and found that the frame rate is wildly different per game. Games like MGS4 has its own special build for 4K60. Even things like AVX512 drastically increases the frame-rate. I guess what I'm getting at is one game doesn't represent the EMU as a whole.

 

[W1zzard]

I guess what I'm getting at is one game doesn't represent the EMU as a whole.

Indeed, but "one game doesn't represent PC gaming" too. I vaguely remember reading that RDR2 on PS3 is one of the most demanding titles? Ultimately it doesn't make much difference, pick anything, keep it constant, relative numbers should be fine?

AVX512

I've gotten a few requests for emulation+AVX512, which seems like the only use case for AVX512 for 99.99% of people

 

[Mussels]

Didn't CIV or Stellaris (I forget which) have a benchmark mode, which counts how many seconds til Turn X?

I Swear i saw it mentioned in the 5800x3D new posts somewhere, but it sounds like the ideal we're looking for - a repeatable, long-duration single threaded test, with gamer relevant results