Overlord Gaming runs more Denuvo tests

D

Derangel

Fully [H]
Joined
Jan 31, 2008
Messages
20,331
Haven't gone through the entire video yet, but the title implies that the tests prove Denuvo harms performance. From what I've seen of the video so far, he is not wrong to make that pronouncement.



For those that don't have time for a half hour video or don't want to watch one I'll make a separate post talking about his findings and my thoughts on them.
 
What is it like 10% drop in performance and huge load time differences. Denuvo sucks more and more the more they update it.
 
Denumo Stealing your frames and giving you latancy... seems like a fair trade
 
They are just stealing the unnecessary frames. You can't see them all anyways so giving a few here and there for their peace of mind seems okay to me.

Well I guess you can't steal frames, as it isn't a tangible item to be taken. So maybe they are pirating them?
 
Darunion said:
They are just stealing the unnecessary frames. You can't see them all anyways so giving a few here and there for their peace of mind seems okay to me.

Well I guess you can't steal frames, as it isn't a tangible item to be taken. So maybe they are pirating them?
Click to expand...

peace of mind?

for whom?
 
Master_shake_ said:
peace of mind?

for whom?
Click to expand...

Any company deploying this in their games to protect their hard earned property. It is a small sacrifice we all are willing to make so the devs and publishers can make money on the products we enjoy without others taking it for nothing and enjoying it for free. If the extra frames mean so much probably should just upgrade that old potato :D
 
He doesn't mention his full specs in the video but he does say he is running a second generation i7 with a 1080. Given Denuvo's impact is likely largely CPU based a stronger CPU might reduce some of the Denuvo vs non Denuvo gaps.

F1 2018- Yikes, those load time differences. Even comparing the "fastest" Denuvo load time with the "slowest" non-Denuvo load time is a massive difference. The benchmark for this game seems pretty awesome. The fps differences on the Australia track are huge, but that max frametime difference is what really caught my eye. 37.93ms vs 13.99ms. Ouch. Denuvo is really putting pressure on that old i7. Adding heavy rain seems to give a boost to the performance of the Denuvo version, a huge boost even. The results are closer as well, interesting. Maybe adding rain reduces CPU load? Not going to summarize every single track he tests, but there are always differences. Denuvo was always slower in terms of loading, raw fps, and frame time to varying degrees. There are a lot of cases where the Denuvo version ends up going below 60fps, sometimes only by a little and sometimes by a decent amount.

Putting all the results together (thankfully he did this) it ends up with the Denuvo version averaging a 15.24% increase in load times. The non-Denuvo version has a 13.73% improvement on min frame rate and a 7.14% improvement on average fps. He didn't summerize the frametime differences, unforutently, and I'll let anyone else who wants to put all that data together figure it out, but I wouldn't be surprised if it showed similar differences if not even larger ones. There were 17 instances where the Denuvo version ended up with a min fps below 60 while the non-denuvo version remained above the threshold and a further 4 times where the Denuvo version dips below 30 with the non-Denuvo version not doing so.

Yakuza 0- Some fairly significant minimum differences here. 19 vs 62 in favor of non-Denuvo. The average fps is so high that unless you're running a 240hz monitor there won't be a noticeable difference but it is still 148.063 vs 160.012. With the max fps of both versions being the same its pretty clear that the Denuvo version was slower through-out the test. This test also marks the 18th time in this this series of tests where Denuvo ended up going below 60fps and the 5th below 30 all while the non-Denuvo version did not hit those numbers. Average frametimes aren't too far off (6.754 vs 6.249) but that maximum number is huge. 345.198 vs 90.758.

Sonic Mania- He reproduced the slow-down issue with time attack mode. Nice to have some confirmation about it being due to Sega's (supposedly) poor implementation of the tech. Once again, we see some big load time differences. No frame-rate tests reported so I'm going to assume that there were no differences between versions. This makes sense given the game and how light it seems to be on CPUs already.

Hitman 2- He had stability issues running the Miami map benchmark with the game so he only did Mumbai. Continuing the trend, there are significant load time differences. First run average FPS isn't a huge difference (42.79 vs 48.54) but in the second run he gets 42.08 vs 52.05, that's significant.

Devil May Cry 5- After the Digital Foundry test, this was a game I was hoping to see people poke at more. He had a lot of stuttering issues with the game using the Denuvo version. Game was run at 1440p, unlike the 720p testing Digital Foundry did.

First test (opening area to boss fight): Denvo's min framerate (due to the massive stutters) was 0 vs 35 for non-Denuvo. Averages weren't terribly different at 120.423 vs 123.201, but with those stutters the game is basically unplayable. Ooooh boy those frametimes. Fairly similar averages (8.3 vs 8.1) but the Denuvo version reaches a whooping 1179.158ms vs 353.822 on the non-Denuvo version. That's over an entire second to render a single frame, holy crap.

Second test (credits/first cinematic): Min fps was 55 vs 63 while average is 85.405 vs 91.75. 19th time Denuvo goes under 60 while non stays above. Interesting frametime results here. Averages show a near one ms difference while the Denuvo version actually has better max than non-Denuvo (96.466 vs 106.474). Interesting.

Third test (introduction of Devil Breaker): Another case where Denuvo ends up with a min fps of 0. Non-Denuvo's min is 83 (20th time for below 60 and 6th time for below 30 all while non-Denuvo remains above 60). Averages were 93.1114 vs 114.645. And here come the frametimes. Averages of 10.74 vs 8.72 and a max of 5182.646 (holy shit) vs 35.579. I think those numbers speak for themselves so I'm not even going to provide my thoughts on them.

Fourth test (evil plant fight): Aaaand third time Denuvo hits 0fps with non-Denuvo only going down to 67 (21st and 7th for the fps thresholds). 102.157 vs 115.295 averages. Not even going to bother with the averages here as the 1381.053ms max frametime for the Denuvo version is really all that matters.

Final test (Couldn't tell what he was saying for where this test is): Usual story with load times. 4th time Denuvo hits 0fps with non-Denuvo getting down the 34. Averages of 74.757 vs 89.778. Whatever mission this is it seems like a good place to test performance in DMCV. Denuvo's max frametime gets up to 6375.441ms. That's over 6 seconds.

Conclusion time- Not really much I can say after those tests. I'd still love to see more channels take on this kind of testing, but Overlord's results are still pretty conclusive. As I said above, a stronger CPU might reduce some of these gaps (especially as there aren't a ton of reports of that kind of crazy stutter in DMCV) but the CPU he's using, along with pushing high settings in the games, does a good job to show the impact Denuvo can have on CPU performance. I didn't mention it in the game summaries, but in all cases Denuvo drastically increased the size of the game exe. A couple hundred megabytes isn't a big deal, but it is another effect Denuvo has on games.
 
Derangel said:
He doesn't mention his full specs in the video but he does say he is running a second generation i7 with a 1080. Given Denuvo's impact is likely largely CPU based a stronger CPU might reduce some of the Denuvo vs non Denuvo gaps.

F1 2018- Yikes, those load time differences. Even comparing the "fastest" Denuvo load time with the "slowest" non-Denuvo load time is a massive difference. The benchmark for this game seems pretty awesome. The fps differences on the Australia track are huge, but that max frametime difference is what really caught my eye. 37.93ms vs 13.99ms. Ouch. Denuvo is really putting pressure on that old i7. Adding heavy rain seems to give a boost to the performance of the Denuvo version, a huge boost even. The results are closer as well, interesting. Maybe adding rain reduces CPU load? Not going to summarize every single track he tests, but there are always differences. Denuvo was always slower in terms of loading, raw fps, and frame time to varying degrees. There are a lot of cases where the Denuvo version ends up going below 60fps, sometimes only by a little and sometimes by a decent amount.

Putting all the results together (thankfully he did this) it ends up with the Denuvo version averaging a 15.24% increase in load times. The non-Denuvo version has a 13.73% improvement on min frame rate and a 7.14% improvement on average fps. He didn't summerize the frametime differences, unforutently, and I'll let anyone else who wants to put all that data together figure it out, but I wouldn't be surprised if it showed similar differences if not even larger ones. There were 17 instances where the Denuvo version ended up with a min fps below 60 while the non-denuvo version remained above the threshold and a further 4 times where the Denuvo version dips below 30 with the non-Denuvo version not doing so.

Yakuza 0- Some fairly significant minimum differences here. 19 vs 62 in favor of non-Denuvo. The average fps is so high that unless you're running a 240hz monitor there won't be a noticeable difference but it is still 148.063 vs 160.012. With the max fps of both versions being the same its pretty clear that the Denuvo version was slower through-out the test. This test also marks the 18th time in this this series of tests where Denuvo ended up going below 60fps and the 5th below 30 all while the non-Denuvo version did not hit those numbers. Average frametimes aren't too far off (6.754 vs 6.249) but that maximum number is huge. 345.198 vs 90.758.

Sonic Mania- He reproduced the slow-down issue with time attack mode. Nice to have some confirmation about it being due to Sega's (supposedly) poor implementation of the tech. Once again, we see some big load time differences. No frame-rate tests reported so I'm going to assume that there were no differences between versions. This makes sense given the game and how light it seems to be on CPUs already.

Hitman 2- He had stability issues running the Miami map benchmark with the game so he only did Mumbai. Continuing the trend, there are significant load time differences. First run average FPS isn't a huge difference (42.79 vs 48.54) but in the second run he gets 42.08 vs 52.05, that's significant.

Devil May Cry 5- After the Digital Foundry test, this was a game I was hoping to see people poke at more. He had a lot of stuttering issues with the game using the Denuvo version. Game was run at 1440p, unlike the 720p testing Digital Foundry did.

First test (opening area to boss fight): Denvo's min framerate (due to the massive stutters) was 0 vs 35 for non-Denuvo. Averages weren't terribly different at 120.423 vs 123.201, but with those stutters the game is basically unplayable. Ooooh boy those frametimes. Fairly similar averages (8.3 vs 8.1) but the Denuvo version reaches a whooping 1179.158ms vs 353.822 on the non-Denuvo version. That's over an entire second to render a single frame, holy crap.

Second test (credits/first cinematic): Min fps was 55 vs 63 while average is 85.405 vs 91.75. 19th time Denuvo goes under 60 while non stays above. Interesting frametime results here. Averages show a near one ms difference while the Denuvo version actually has better max than non-Denuvo (96.466 vs 106.474). Interesting.

Third test (introduction of Devil Breaker): Another case where Denuvo ends up with a min fps of 0. Non-Denuvo's min is 83 (20th time for below 60 and 6th time for below 30 all while non-Denuvo remains above 60). Averages were 93.1114 vs 114.645. And here come the frametimes. Averages of 10.74 vs 8.72 and a max of 5182.646 (holy shit) vs 35.579. I think those numbers speak for themselves so I'm not even going to provide my thoughts on them.

Fourth test (evil plant fight): Aaaand third time Denuvo hits 0fps with non-Denuvo only going down to 67 (21st and 7th for the fps thresholds). 102.157 vs 115.295 averages. Not even going to bother with the averages here as the 1381.053ms max frametime for the Denuvo version is really all that matters.

Final test (Couldn't tell what he was saying for where this test is): Usual story with load times. 4th time Denuvo hits 0fps with non-Denuvo getting down the 34. Averages of 74.757 vs 89.778. Whatever mission this is it seems like a good place to test performance in DMCV. Denuvo's max frametime gets up to 6375.441ms. That's over 6 seconds.

Conclusion time- Not really much I can say after those tests. I'd still love to see more channels take on this kind of testing, but Overlord's results are still pretty conclusive. As I said above, a stronger CPU might reduce some of these gaps (especially as there aren't a ton of reports of that kind of crazy stutter in DMCV) but the CPU he's using, along with pushing high settings in the games, does a good job to show the impact Denuvo can have on CPU performance. I didn't mention it in the game summaries, but in all cases Denuvo drastically increased the size of the game exe. A couple hundred megabytes isn't a big deal, but it is another effect Denuvo has on games.
Click to expand...
Ja, second gen would mean Sandy Bridge. Maybe it's time to upgrade that 8-year old CPU. Let's see what happens on Coffee Lake.
 
Darunion said:
Any company deploying this in their games to protect their hard earned property. It is a small sacrifice we all are willing to make so the devs and publishers can make money on the products we enjoy without others taking it for nothing and enjoying it for free. If the extra frames mean so much probably should just upgrade that old potato :D
Click to expand...

pirates don't deal with DRM though only paying customers.

paying customers get clobbered.
 
Master_shake_ said:
pirates don't deal with DRM though only paying customers.

paying customers get clobbered.
Click to expand...

Yea. i'm just trolling in this particular thread. I do agree though, same as with movies. You pirate a movie you don't have all the dumb ads or previews or fbi warnings. Only the people who pay for it have to deal with it.
 
Darunion said:
Yea. i'm just trolling in this particular thread. I do agree though, same as with movies. You pirate a movie you don't have all the dumb ads or previews or fbi warnings. Only the people who pay for it have to deal with it.
Click to expand...

Doesn’t really apply to Denuvo. Outside of cases like FFXV or DMCB where Denuvo free exes exist pirated copies still have Denuvo active. The cracks don’t disable Denuvo, they trick it into thinking the game is legit. So even if you pirate a Denuvo game you are still dealing with Denuvo.
 
Derangel said:
Doesn’t really apply to Denuvo. Outside of cases like FFXV or DMCB where Denuvo free exes exist pirated copies still have Denuvo active. The cracks don’t disable Denuvo, they trick it into thinking the game is legit. So even if you pirate a Denuvo game you are still dealing with Denuvo.
Click to expand...

This is true however pirates haven't had to deal with games refusing to launch when the publisher's denuvo server has gone down like legit users have so they still get an advantage. I'm not sure if pirate versions can use mods and utilities with scripthooks in games that denuvo prevents it but that's another potential advantage.
 
Doom without Denuvo takes 1/3 the loadtime it did. It dropped from 30 seconds to ten seconds.
 
DukenukemX said:
Wow the loading time is horrible. What pirates get to enjoy.
Click to expand...
It’s a pity pirated copies still have Denuvo for games that haven’t removed it. If they found a way to remove or deactivate the DRM I’d be getting pirated exes for all all the Denuvo games I own.
 
You must log in or register to reply here.
Back
Top