3DMark Steel Nomad

[TASOS]

Hi everybody

Are we going to host both preset ?
Steel Nomad
Steel Nomad Light

... and what about the different setting inside the bench (DX12 and Vulkan) ?

It seems that Vulkan setting scores higher than DX12.

Hwbot thoughts ?

 

 

[Papusan]

One thing more. The rules for 3DMark Steel Nomad say nothing abour enabling/disabling ECC for nVIDIA RTX4090 cards. Will this be changed later in  same as for some other 3Dmark Suite benchmarks and your old scores being deleted? Thanks.

 

Edit. Added the rules for 3DMark Steel Nomad https://hwbot.org/rules?referenceId=5440

Edited May 22 by Papusan

[Leeghoofd]

We got some tests results in and its very memory sensitive, so ECC enabled for RTX 4090...

[NATA 58]

ECC enabled for RTX 4090... O.K  vulKan It's 10% times faster than dx12

Edited May 22 by NATA 58

[FM_Jarnis]

Should probably have DX12 and Vulkan separately for both Steel Nomad and Steel Nomad light as they will be separate tests in 3DMark Hall of Fame

NVIDIA DX12 vs Vulkan perf difference is most likely driver-related and could change.

[Leeghoofd]

We will add a Vulkan only version and rename the other one to DX12 only

[Leeghoofd]

9 hours ago, FM_Jarnis said:

Should probably have DX12 and Vulkan separately for both Steel Nomad and Steel Nomad light as they will be separate tests in 3DMark Hall of Fame

NVIDIA DX12 vs Vulkan perf difference is most likely driver-related and could change.

At the moment it is pretty hard to detect if DX12 or Vulkan was used as all result links seem to end up hidden.

HWBOT wil  need to enforce another rule as most users don't show the full score window to show "the Settings used" part. 

[Leeghoofd]

https://hwbot.org/benchmark/3dmark_-_steel_nomad_dx12/

https://hwbot.org/benchmark/3dmark_-_steel_nomad_vulkan/

Sample screenshot, showing the API rendering is required:

 

[NATA 58]

steel nomad and speedway horrible benchmarks

[Fasttrack]

On 5/23/2024 at 11:44 PM, Leeghoofd said:

https://hwbot.org/benchmark/3dmark_-_steel_nomad_dx12/

https://hwbot.org/benchmark/3dmark_-_steel_nomad_vulkan/

Sample screenshot, showing the API rendering is required:

 

On 4K screens it is only possible to get all the details required for the screenshot Albrecht. I mean a screenshot showing DX12 or Vulkan.

No other resolution has the screen space to deploy the entire benchmark run details. Not even 2K.

Should be mentioned as advice that a 4k screen is mandatory.

Trust me there is no way around

[Papusan]

1 hour ago, Fasttrack said:

On 4K screens it is only possible to get all the details required for the screenshot Albrecht. I mean a screenshot showing DX12 or Vulkan.

No other resolution has the screen space to deploy the entire benchmark run details. Not even 2K.

Should be mentioned as advice that a 4k screen is mandatory.

Trust me there is no way around

If so. Then UL should add an API settings box as suggested in the screenshoot. @FM_JarnisIs this possible change the GUI ? And you say this is a Time Spy replasement. Why not put in a CPU test same as for 3DM Time Spy and older benchmarks? The Cpu will be more and more important for gaming up to 4K. Thanks

Edited May 25 by Papusan

[Mr. Fox]

4 hours ago, Papusan said:

If so. Then UL should add an API settings box as suggested in the screenshoot. @FM_JarnisIs this possible change the GUI ? And you say this is a Time Spy replasement. Why not put in a CPU test same as for 3DM Time Spy and older benchmarks? The Cpu will be more and more important for gaming up to 4K. Thanks

Yes. That is the only logical, intelligent solution. Requiring people to waste money on a 4K display if they don't want one would be a truly great way of driving yet another nail into the coffin of our rapidly dying sport. That would be a very stupid decision. If anyone is easily butt-hurt and looking for ways to report scores that beat theirs they can always use the required validation link and visit 3dmark.com if they think the submitter is a liar.

Edited May 25 by Mr. Fox

[Leeghoofd]

if we have to click the UL link , open detailed settings for each sub, that is a pure waste of time..., hence why I preferred one ranking, 

This is 1920 x1080 res

 

[Mr. Fox]

Yeah, it would be extra work and not a great use of a moderator's time. But, you could require the person reporting the submission to provide a screenshot if they expect moderation to occur. That would shift the burden to the person reporting it.

[NATA 58]

no sooner can we eliminate this horrible test

[Papusan]

On 5/26/2024 at 6:47 AM, NATA 58 said:

no sooner can we eliminate this horrible test

LOOL. First... With the newer 3Dmark benchmarks you are at huge risk of getting uncorrectable errors stored in the inforom and cannot easly be erased when you running too high of memory overclock with "ECC enabled" for 4090 (HWbot's requirement). Now you get another gift from UL/Futuremark. Not sure whats up with the quality and QC they offer nowadays. @FM_Jarnisis there a real fix for the 4090 cards so we can use the graphics cards the way it was meant to be used (gaming and benchmarks). ECC Memory in NCP have never been meant to be Enabled for gaming graphics cards. I wonder how many 4090 owners that don't know that their 4090's is halfway destroyed after running their cards with ECC video memory Enabled. There has to be a real fix for UL benchmarks with 4090's. 

 

3DMARK STEEL NOMADU sers are annoyed by “bug runs”, UL is considering measures

“Bug runs” are still possible in 3DMark Steel Nomad. This annoys users, but it doesn’t have to stay that way: UL is considering countermeasures.

 

UL Solutions has “some ideas”

ComputerBase asked UL Solutions whether this will change in the future and received an answer. The developer explained that they are currently considering how to deal with the problem. They already have " a few ideas ". However, UL is not yet being more specific.

 

One possibility would be to create screenshots during the benchmark and submit them together with the validated result. In the cloud, UL could then (automatically) compare the screenshots created with reference screenshots. However, this could cost performance and thus have an impact on the result.

Edited May 28 by Papusan

[NATA 58]

isn't it better to delete steel nomad?

[Mr. Fox]

31 minutes ago, NATA 58 said:

isn't it better to delete steel nomad?

Probably so. UL has turned Futuremark into a huge joke. They used to be the industry standard, now they are just laughingstock.

And, no... I am not going to ruin the memory on either one of my 4090s just for the sake of submitting meaningless benchmark scores at 3DMark.com or here. That's just way too stupid to even consider the idea. I wish NVIDIA had not enabled the ECC crap on these cards.

57 minutes ago, Papusan said:

LOOL. First... With the newer 3Dmark benchmarks you are at huge risk of getting uncorrectable errors stored in the inforom and cannot easly be erased when you running too high of memory overclock with "ECC enabled" for 4090 (HWbot's requirement). Now you get another gift from UL/Futuremark. Not sure whats up with the quality and QC they offer nowadays. @FM_Jarnisis there a real fix for the 4090 cards so we can use the graphics cards the way it was meant to be used (gaming and benchmarks). ECC Memory in NCP have never been meant to be Enabled for gaming graphics cards. I wonder how many 4090 owners that don't know that their 4090's is halfway destroyed after running their cards with ECC video memory Enabled. There has to be a real fix for UL benchmarks with 4090's. 

 

3DMARK STEEL NOMADU sers are annoyed by “bug runs”, UL is considering measures

“Bug runs” are still possible in 3DMark Steel Nomad. This annoys users, but it doesn’t have to stay that way: UL is considering countermeasures.

 

UL Solutions has “some ideas”

ComputerBase asked UL Solutions whether this will change in the future and received an answer. The developer explained that they are currently considering how to deal with the problem. They already have " a few ideas ". However, UL is not yet being more specific.

 

One possibility would be to create screenshots during the benchmark and submit them together with the validated result. In the cloud, UL could then (automatically) compare the screenshots created with reference screenshots. However, this could cost performance and thus have an impact on the result.

Should be interesting to see what their "ideas" are since they haven't had any good ideas in a very long time. 3DMark turned to crap when UL took over. It's incredible how stupidity has taken the PC tech industry by storm. Intelligence has become rare and UL is leading the charge with their trashmark software.

 

Edited May 28 by Mr. Fox

[Papusan]

2 hours ago, Mr. Fox said:

Should be interesting to see what their "ideas" are since they haven't had any good ideas in a very long time.

And, no... I am not going to ruin the memory on either one of my 4090s just for the sake of submitting meaningless benchmark scores at 3DMark.com or here. That's just way too stupid to even consider the idea. I wish NVIDIA had not enabled the ECC crap on these cards.

With the much cheaper 4060-4060Ti-4070-4070Ti-4070Ti Super and 4080 and 4080 Super you are good. No ECC mem settings to Enable for Hwbot. This ECC corruption disaster only apply for the $2000 4090’s🤠 

Yup, I looking forwards to a possible fix. Both for the benchmark and hardware. 

Edited May 28 by Papusan

[Mr. Fox]

1 hour ago, Papusan said:

With the much cheaper 4060-4060Ti-4070-4070Ti-4070Ti Super and 4080 and 4080 Super you are good. No ECC mem settings to Enable for Hwbot. This ECC corruption disaster only apply for the $2000 4090’s🤠 

Yup, I looking forwards to a possible fix. Both for the benchmark and hardware. 

NVIDIA should have a warning that appears in NVIDIA Control Panel when ECC is enabled to inform the user that enabling ECC may result in unrecoverable damage to the GPU memory. Then neither HWBOT nor UL would need to worry about it or find solutions because nobody with a normal level of intelligence would enable ECC and there would be no 3DMark submissions for HWBOT moderators to have to worry about moderating. UL could continue producing the modern cartoon benchmark crap for the kiddos.

Edited May 28 by Mr. Fox

[Seby]

It is recoverable, you can flash the card back to new using nvflash. Instead of flashing VBIOS, you dump the infoROM, zero out the relevant sections, and reflash the infoROM.

Edited May 28 by Seby

[Papusan]

3 minutes ago, Seby said:

It is recoverable, you can flash the card back to new using nvflash. Not vbios it’s a different part of the gpu

In short. Damage the cards for benchmarks then repair it afterwards. Not so sure many know about this problem with 4090’’s and the fix🙂

Regardless. This shouldn’t be needed if we got benchmark software that worked properly. 

[Seby]

Not like the “damage” will ever be obvious, and it really is 2 minute fix if you care.

[Mr. Fox]

3 hours ago, Seby said:

It is recoverable, you can flash the card back to new using nvflash. Instead of flashing VBIOS, you dump the infoROM, zero out the relevant sections, and reflash the infoROM.

 

3 hours ago, Papusan said:

In short. Damage the cards for benchmarks then repair it afterwards. Not so sure many know about this problem with 4090’’s and the fix🙂

Regardless. This shouldn’t be needed if we got benchmark software that worked properly. 

Are there instructions how to do this posted online that you can share? Is it done using NVFLASH or a hardware programmer? This should not be necessary, and it wouldn't be had NVIDIA not enabled this. But, it would be helpful to know in case it ever needs to be done. Thanks in advance for any detail you can share.

Edited May 29 by Mr. Fox

[Seby]

nvflash64.exe --save oldifr.ifr

open file in hex editor, set everything past offset 0x5568 to just 00s. 

save as whatever.ifr

nvflash64.exe --flashinforom whatever.ifr

Edited May 29 by Seby