EVGA Left At the Right Time: NVIDIA RTX 4090 Founders Deep-Dive (Schlieren, 12-Pin, & Pressure)
video description
Date: 2022-10-13
Related videos
Comments and reviews: 15
NotLulzsec
(Sorry for poor english, Im high and drunk) Ive only watched about 3minutes of this video.
ehh as someone who actually knows electrical engineering, that power cable you showed will actually help with the power draw, like its basic logic, the more input load available and that can be drawn the more successful then overclock will be without a fire hazard... That 4 connector cable will actually do something, and it surprises me that corsair said it was fake. That cable isnt fake at all and is well backed up by basic science of electricity... The more power you draw the more power is required.. And to prevent a fire, you either need thicker cabling from your source input or more connections to distribute the load from the source. Which that 4connector cable is doing..
Its like when you want to knock someone's internet down. You can have a DoS with a thick bandwidth/wire or multiple bandwidth/DDoS/Multi connection cables.
who knew corsair dorsnt believe in logic when outside ofpc peripherals and cases.
reply
(Sorry for poor english, Im high and drunk) Ive only watched about 3minutes of this video.
ehh as someone who actually knows electrical engineering, that power cable you showed will actually help with the power draw, like its basic logic, the more input load available and that can be drawn the more successful then overclock will be without a fire hazard... That 4 connector cable will actually do something, and it surprises me that corsair said it was fake. That cable isnt fake at all and is well backed up by basic science of electricity... The more power you draw the more power is required.. And to prevent a fire, you either need thicker cabling from your source input or more connections to distribute the load from the source. Which that 4connector cable is doing..
Its like when you want to knock someone's internet down. You can have a DoS with a thick bandwidth/wire or multiple bandwidth/DDoS/Multi connection cables.
who knew corsair dorsnt believe in logic when outside ofpc peripherals and cases.
reply
fierydawn
Regarding the thing about plugging in only three connectors.
Der8auer cut off the sleeve and found that only two of the sensor pins on his FE card adapter, were connected to the actual 8pin connectors. Which, if that's true for all of the FE cards, then it's possible that depending on which of the four connectors you connect to, you might even get the full power limit with just two connectors.
So if you have time, maybe test to see if the power limit is decided by specific cables. By first connecting connectors 123, then 124, and 234. And testing to see if the powerlimit stays at 100%, or if some combination allows access to 133% with only three connectors.
Der8auer also mentioned that it's weird that the corsair provided a power connector has only two 8pins leading to the single 12pin. Which makes me think maybe corsair decided that's enough for the card, and the comment they made is based on that.
reply
Regarding the thing about plugging in only three connectors.
Der8auer cut off the sleeve and found that only two of the sensor pins on his FE card adapter, were connected to the actual 8pin connectors. Which, if that's true for all of the FE cards, then it's possible that depending on which of the four connectors you connect to, you might even get the full power limit with just two connectors.
So if you have time, maybe test to see if the power limit is decided by specific cables. By first connecting connectors 123, then 124, and 234. And testing to see if the powerlimit stays at 100%, or if some combination allows access to 133% with only three connectors.
Der8auer also mentioned that it's weird that the corsair provided a power connector has only two 8pins leading to the single 12pin. Which makes me think maybe corsair decided that's enough for the card, and the comment they made is based on that.
reply
Sebastian
Interesting results concerning LTT's CP2077 testing. Had the same thought, that there must have been something wrong.
Not only was their RTX4090 way above the expected, the 6950XT was also uncharacteristically low.
While all tests I have had a look at, reported relatively comparable results within a few frames, depending on the settings (DLSS, RT).
Not exactly comforting, when they particularly point out that they tripple checked the results...
But It's most definitely a good decision to clarify, that there was something wrong with the tests. So thumbs up for that.
On that note, with all the enhancement options and tweaks you can apply outside of basic rasterization,
it's no surprise that customers get confused and reviewers are drowning in tests and testing charts.
And the implementation of DLSS3 and frame insertion isn't going to make it any easier...
reply
Interesting results concerning LTT's CP2077 testing. Had the same thought, that there must have been something wrong.
Not only was their RTX4090 way above the expected, the 6950XT was also uncharacteristically low.
While all tests I have had a look at, reported relatively comparable results within a few frames, depending on the settings (DLSS, RT).
Not exactly comforting, when they particularly point out that they tripple checked the results...
But It's most definitely a good decision to clarify, that there was something wrong with the tests. So thumbs up for that.
On that note, with all the enhancement options and tweaks you can apply outside of basic rasterization,
it's no surprise that customers get confused and reviewers are drowning in tests and testing charts.
And the implementation of DLSS3 and frame insertion isn't going to make it any easier...
reply
ElectronAsh
Does the 4090 prevent the overclock if ANY one of the power cables is disconnected from the adapter?
If so, then the only way I can see that working is if there is a small amount of active logic in the adapter itself (which drives the two sense pins on the card), or the adapter simply connects a ground from each PSU socket through to separate ground pins on the card.
I think the latter is far more likely. I don't see any other way it would be able to tell if there is any one PSU plug not connected.
(I'm not arguing that your test is invalid, it clearly works. I'm just wondering how the adapter itself works.)
Is there any way you could do a pin map of the adapter cable. ie. all card pins vs the pins of all four PSU sockets?
reply
Does the 4090 prevent the overclock if ANY one of the power cables is disconnected from the adapter?
If so, then the only way I can see that working is if there is a small amount of active logic in the adapter itself (which drives the two sense pins on the card), or the adapter simply connects a ground from each PSU socket through to separate ground pins on the card.
I think the latter is far more likely. I don't see any other way it would be able to tell if there is any one PSU plug not connected.
(I'm not arguing that your test is invalid, it clearly works. I'm just wondering how the adapter itself works.)
Is there any way you could do a pin map of the adapter cable. ie. all card pins vs the pins of all four PSU sockets?
reply
Brezimenko
Nvidia is trying to cut their losses with forcing out other AIBs, since their flagship is in reality only 25-30% faster than 3090 (without fake DLSS), and AMD is comming strong this year around. Its expected to be at least 50% stronger than 4090, which will turn other buyers (including me) to switch to AMD, like I did with Ryzen.. And the design is , saving costs using the same design only bigger and thicker, which look like mini itx build, so ugly in the case. Their reviewers hyping the bigger card design, but the world is going in another direction (smaller but stronger), no one is using tower cases like in the 2000s, which are ment for servers nowdays.. RTX 4000 is a flop, just wait for RX 7000
reply
Nvidia is trying to cut their losses with forcing out other AIBs, since their flagship is in reality only 25-30% faster than 3090 (without fake DLSS), and AMD is comming strong this year around. Its expected to be at least 50% stronger than 4090, which will turn other buyers (including me) to switch to AMD, like I did with Ryzen.. And the design is , saving costs using the same design only bigger and thicker, which look like mini itx build, so ugly in the case. Their reviewers hyping the bigger card design, but the world is going in another direction (smaller but stronger), no one is using tower cases like in the 2000s, which are ment for servers nowdays.. RTX 4000 is a flop, just wait for RX 7000
reply
TheSteelRodent
It's sad you feel a need to demonstrate and explain your testing methods. It's the manufacturers who should be required to do that, especially when they keep making bogus statements like 33% faster or whatever without any reference to how they achieved or verified those numbers. But while you mentioned the air bouncing off the PSU shroud, I wonder if it would be possible for you guys to do some video about what difference that shroud makes? I for one still use my old Define R4 (because it's an expensive and very good case) which doesn't have one of those and I doubt I'm the only one who use a case from before those shrouds became a standard feature.
reply
It's sad you feel a need to demonstrate and explain your testing methods. It's the manufacturers who should be required to do that, especially when they keep making bogus statements like 33% faster or whatever without any reference to how they achieved or verified those numbers. But while you mentioned the air bouncing off the PSU shroud, I wonder if it would be possible for you guys to do some video about what difference that shroud makes? I for one still use my old Define R4 (because it's an expensive and very good case) which doesn't have one of those and I doubt I'm the only one who use a case from before those shrouds became a standard feature.
reply
Ryan
The thing about the sense pins is that it's a VERY simple system. Anyone with 1/4 of a brain and a 20 multimeter from Amazon can figure it out. Pins S3 and S4 on the 4 pin sideband connector are the pins that tell the GPU the power level. And they do this by other leaving the wires open or grounded.
Intel publishes the ATX 3.0 spec on their website, so it's easy to look up. All you have to do is set your meter to continuity mode and probe S3 and S4 continuity to ground. I verified this at work when dealing with a PSU manufacturer. I'm not an engineer, I don't work for a power supply manufacturer, but it was trivially easy to figure out.
reply
The thing about the sense pins is that it's a VERY simple system. Anyone with 1/4 of a brain and a 20 multimeter from Amazon can figure it out. Pins S3 and S4 on the 4 pin sideband connector are the pins that tell the GPU the power level. And they do this by other leaving the wires open or grounded.
Intel publishes the ATX 3.0 spec on their website, so it's easy to look up. All you have to do is set your meter to continuity mode and probe S3 and S4 continuity to ground. I verified this at work when dealing with a PSU manufacturer. I'm not an engineer, I don't work for a power supply manufacturer, but it was trivially easy to figure out.
reply
David
So we're going to see this from AMD and Intel as well at some point. AIB's specialize on just what they do, so they start out better, but eventually that edge will be reduced as the parent company improves their own processes with each generation. I don't think it means the end of AIBs - provided the main company still wants AIB participation - it just means they are no longer the automatic go to for end users. In the case of Nvidia, the warning sign will not be oh look their own coolers aren't crap anymore, it was inevitable they'd eventually catch up to AIBs, no it's when they actually start investing in the ability to produce at scale.
reply
So we're going to see this from AMD and Intel as well at some point. AIB's specialize on just what they do, so they start out better, but eventually that edge will be reduced as the parent company improves their own processes with each generation. I don't think it means the end of AIBs - provided the main company still wants AIB participation - it just means they are no longer the automatic go to for end users. In the case of Nvidia, the warning sign will not be oh look their own coolers aren't crap anymore, it was inevitable they'd eventually catch up to AIBs, no it's when they actually start investing in the ability to produce at scale.
reply
Thomas
I just stick with my 2070 Super and maybe switch to an AMD GPU in future. I don t use DLSS1 because it s garbage. DLSS2 also does not look good. And I only play one game with RTX And Corsair is a funny company. I like their SFF PSUs but their are now some alternatives. I can t talk about Corsair directly, but imagine there is a company and it has thousands of AIOs and nobody knows how to connect pump and LED to their software and it is before Christmas and they want to sell it before Christmas. So the company had to outsource reverse engineering everything to someone else to make their own things work together.
reply
I just stick with my 2070 Super and maybe switch to an AMD GPU in future. I don t use DLSS1 because it s garbage. DLSS2 also does not look good. And I only play one game with RTX And Corsair is a funny company. I like their SFF PSUs but their are now some alternatives. I can t talk about Corsair directly, but imagine there is a company and it has thousands of AIOs and nobody knows how to connect pump and LED to their software and it is before Christmas and they want to sell it before Christmas. So the company had to outsource reverse engineering everything to someone else to make their own things work together.
reply
Robinthefox88
Corsair's rep needs a strong talking to, it almost feels like they spouted off based on their own limited knowledge rather than seeking information from their engineers or from nvidia directly.
It feels like their information is partly based on the way old adapters used to work (molex to pcie) that really were dumb as bricks or the 3000 series adapter which didn't have the sense pins hooked up.
Whatever the case, they should apologise and put out a statement correcting the situation before the misinformation gets out of hand
reply
Corsair's rep needs a strong talking to, it almost feels like they spouted off based on their own limited knowledge rather than seeking information from their engineers or from nvidia directly.
It feels like their information is partly based on the way old adapters used to work (molex to pcie) that really were dumb as bricks or the 3000 series adapter which didn't have the sense pins hooked up.
Whatever the case, they should apologise and put out a statement correcting the situation before the misinformation gets out of hand
reply
sid
Oh, Why oh, why would Corsair, come for you, Steve??? They had to know like we know what they were going to get if they challenged you and you used an Nvidia engineer/representative to destroy Corsair's argument, priceless!!! This is how to have an egg on your face when you think you are smarter than you really are (Corsair). Way to go Steve and way to do the Gamers Nexus community and your name proud!!! Corsair you better put some RESPECT on the GAMERS NEXUS Name!!!
reply
Oh, Why oh, why would Corsair, come for you, Steve??? They had to know like we know what they were going to get if they challenged you and you used an Nvidia engineer/representative to destroy Corsair's argument, priceless!!! This is how to have an egg on your face when you think you are smarter than you really are (Corsair). Way to go Steve and way to do the Gamers Nexus community and your name proud!!! Corsair you better put some RESPECT on the GAMERS NEXUS Name!!!
reply
MSJI
If the design is done right, 3x8 pin cables would suffice. nVidia seems to have tracked the safer route (4x8pin cables) here through the implementation of the sense wires. Manufacturing a 12VHPWR (600 watt) with dual pin connectors is inviting risk of overheated wires/pin connectors. I saw this is Corsair implementation. I wonder if this cable will function in 4090 and that the card can detect that 2x8 pins is connected to the psu..
reply
If the design is done right, 3x8 pin cables would suffice. nVidia seems to have tracked the safer route (4x8pin cables) here through the implementation of the sense wires. Manufacturing a 12VHPWR (600 watt) with dual pin connectors is inviting risk of overheated wires/pin connectors. I saw this is Corsair implementation. I wonder if this cable will function in 4090 and that the card can detect that 2x8 pins is connected to the psu..
reply
Karol
I don't know what's with NVidia's FE cards here in Europe, I made my mind to already put my hands on it but in never appeared in any online store here in Germany. I have bought the cheapest AiB card I have found instead, Inno3d X3 OC, hope it will serve well as well. All the AiBs cards are now sold out so NVidia made them a big favor by not delivering their FE product on the premiere.
reply
I don't know what's with NVidia's FE cards here in Europe, I made my mind to already put my hands on it but in never appeared in any online store here in Germany. I have bought the cheapest AiB card I have found instead, Inno3d X3 OC, hope it will serve well as well. All the AiBs cards are now sold out so NVidia made them a big favor by not delivering their FE product on the premiere.
reply
Anon
4090 was planned to be a 660W card for the record so Corsair are a buncha clowns.
The reasons Nvidia chose not to do that in order to squeeze out the remaining 5-10% of performance was cause
- they're already getting clowned for the 450W imagine a 660W
- they realised in testing that CPU's were bottlenecking making it pointless
reply
4090 was planned to be a 660W card for the record so Corsair are a buncha clowns.
The reasons Nvidia chose not to do that in order to squeeze out the remaining 5-10% of performance was cause
- they're already getting clowned for the 450W imagine a 660W
- they realised in testing that CPU's were bottlenecking making it pointless
reply
Trenjeska
(more comments than likes atm? weird...)
Both fans at same 85 percentage speed is causing resonance.
They know this apparently as they said they purposefully let the fans spin at different values to avoid this resonance.
(like you pointed out when talking about the 42/48% split between the 2 in the default profile)
reply
(more comments than likes atm? weird...)
Both fans at same 85 percentage speed is causing resonance.
They know this apparently as they said they purposefully let the fans spin at different values to avoid this resonance.
(like you pointed out when talking about the 42/48% split between the 2 in the default profile)
reply
Add a review, comment
Other channel videos
