Forspoken implements Microsoft's DirectStorage API, offers faster load times but lowers frame rate

pegasus1

2023-01-27 08:48

The trade-off i spose depends on how much the load times are reduced compared to the frame rate drop. A 4090 can afford to lose 10% in FPS but only if there is a significant drop in load times. 10% loss for 10% improvement doesn't seem like a good tradeoff.

#6097073

Ghosty

2023-01-27 09:09

pegasus1:

The trade-off i spose depends on how much the load times are reduced compared to the frame rate drop. A 4090 can afford to lose 10% in FPS but only if there is a significant drop in load times. 10% loss for 10% improvement doesn't seem like a good tradeoff.

Useful for the Series X though.

#6097076

Crazy Joe

2023-01-27 09:19

That must be hands-down the most useless graph ever created: all the entries are marked with the same text, so there is no way to tell which graph corresponds to which SSD type used.

#6097080

Spets

2023-01-27 09:31

They collected the data wrong, sata has higher loading times and was recording empty frame black screens. CB will more than likely edit their results soon.

#6097082

Ghosty

2023-01-27 09:39

Spets:

They collected the data wrong, sata has higher loading times and was recording empty frame black screens. PCGH will more than likely edit their results soon.

Everyone knows SATA is slower than NVME for data transfer. Direct_Storage doesn't help bridge the gap, it just makes it worse for both.

#6097084

Spets

2023-01-27 10:03

Ghosty:

Everyone knows SATA is slower than NVME for data transfer. Direct_Storage doesn't help bridge the gap, it just makes it worse for both.

I'm talking about the avg fps recorded being invalidated. https://twitter.com/CapFrameX/status/1618636871445393409

Loading of the sequences causes blackscreens with high FPS which has an impact on the average FPS when measuring with CapFrameX. You can handle this by using the cutting feature.

#6097090

pegasus1

2023-01-27 10:12

Crazy Joe:

That must be hands-down the most useless graph ever created: all the entries are marked with the same text, so there is no way to tell which graph corresponds to which SSD type used.

I saw this article on another site last night and thought the same, percentages are fine when you know what is a percentage of what.

#6097098

Crazy Joe

2023-01-27 10:47

pegasus1:

I saw this article on another site last night and thought the same, percentages are fine when you know what is a percentage of what.

I see that Hilbert has updated the images now.

#6097099

PrMinisterGR

2023-01-27 10:49

Isn't this the exact tradeoff because the consoles have dedicated hardware to do this, but PCs have all the bus overheads (PCIe, memory etc), plus they need to use GPU resources (in the best case), to do constant asset streaming.

#6097100

Astyanax

2023-01-27 10:50

PrMinisterGR:

Isn't this the exact tradeoff because the consoles have dedicated hardware to do this, but PCs have all the bus overheads (PCIe, memory etc), plus they need to use GPU resources (in the best case), to do constant asset streaming.

theres no dedicated hardware involved, the results are faulty because one case has a longer period on a high framerate black screen than the other.

#6097104

pegasus1

2023-01-27 11:11

Crazy Joe:

I see that Hilbert has updated the images now.

It still makes no sense, the graph doesnt show frame rates with and without use of the API and doesnt show load times with and without the API

#6097107

asturur

2023-01-27 11:50

Ghosty:

Everyone knows SATA is slower than NVME for data transfer. Direct_Storage doesn't help bridge the gap, it just makes it worse for both.

I think the point was that with sata you can't have direct storage, so sata is the case of direct storage OFF

#6097109

rflair

Moderator

2023-01-27 11:54

DS incurs a CPU overhead isn't that what was shown to happen when they first released this for testing?

#6097113

Typhon Six Six Six

2023-01-27 12:06

I run my games off 8TB SSD. Long load times are like 15-20 sec and most games load in 5-10 seconds.

#6097115

MachinaEx

2023-01-27 12:10

This is quite unexpected! Well it seems that additional optimization's will have to be made before we see more implementations of this technology...ASUS TUF X3 GeForce 1660 Super user here

#6097116

TheDeeGee

2023-01-27 12:11

Saw a review of this game on YT, every 10 seconds it takes away control for a cutscene... ****ing aweful!

#6097117

southamptonfc

2023-01-27 12:11

pegasus1:

The trade-off i spose depends on how much the load times are reduced compared to the frame rate drop. A 4090 can afford to lose 10% in FPS but only if there is a significant drop in load times. 10% loss for 10% improvement doesn't seem like a good tradeoff.

It's a big NO from me for 10% loss in FPS. My 4090 manages less than 100fps in a lot of modern titles. I don't want to turn details down just because MS create inneficient software.

#6097123

Kaleid

2023-01-27 12:27

The game uses DS 1.0 and not 1.1 right? Which means the GPU is not assisting properly at this point with decompression.

#6097124

Horus-Anhur

2023-01-27 12:28

southamptonfc:

It's a big NO from me for 10% loss in FPS. My 4090 manages less than 100fps in a lot of modern titles. I don't want to turn details down just because MS create inneficient software.

There isn't a performance loss. The guys at PCGH made a mistake measuring frame rate with the CapX tool.

#6097128

pegasus1

2023-01-27 12:44

TheDeeGee:

Saw a review of this game on YT, every 10 seconds it takes away control for a cutscene... ****ing aweful!

That is a huge turn off for me in games, the scene in Metro EE where Sam and the Captain get drunk on the rooftop, ffs its about 15 minutes of cutscene bullsh1te.