@@MikeJones_OfficialUhh, Oculink adapters already use the m.2 port? When you modify a laptop as i have what it serves to do is expose those pcie lanes for unconventional use.
No one could have imagined an Intel GPU 10 years ago. Intel is probably still relevant because of arc and Iris brands, just like AMD made Gold with ATI.
@@james8449100 as a matter of fact: Laptops with OCuLink are getting more mainstream and more cheaper Laptops with Thunderbolt still Hilariously expensive. Mainly because the Thunderbolt protocol Royalty isn't cheap. It will speak for itself in less than 2 years
Not really, no. You're limited hardcore with OcuLink as it only really uses eGPUs. TB3/4/USB 4 allows you to also hook up super fast drives and docking stations as well as do other displays stuff through it
@@PixelatedWolf2077 raw PCIe signal of Thunderbolt 4 or USB4 is ridiculously slower than OCuLink Besides, Laptop with OCuLink also comes with USB4 alongside it anyway. But if you plug eGPU into USB4 instead of OCuLink , that's your mistake
@niezzayt3809 Ah, another thing you reminded me. The price isn't always worth it to run OcuLink, especially if you're going to run a handheld or you have a "business" PC build and want extra performance
I believe my account is shadow banned for using revanced, about 50% of my comments get deleted. But anyways all ive got right now is my I5-8250U Dell Inspiron 15 7573 that i installed oculink in and a E495 thinkpad
Give me money I'll mass produce a motherboard with 4 ram slots, 2 CPU slots, 2 mini PCIe slots, and 4 M.2 slots. I will call it the Decade+GF Ultra motherboard
I've watched a few B580 videos lately, so YT recommended this one. And I watched it. But it's time for some constructive criticism. First, there's way too much yapping at the beginning. Describe what you're doing. Give a few details about the test system, but only relevant stuff. Don't introduce even more unknowns and variables that you don't have to. What do you mean you don't know how many lanes the 8600G gives the x16 slot? Put the card in, open up GPU-Z, and find out. Record and publish only stuff you know. You don't need "I will find out how my system is even set up." in an overly long intro. Second, I was casually watching on my phone at the start while doing something else. By the time you finally got around to showing benchmark numbers, you had my attention again. But you didn't label which system and results were which. I've paused the video while typing this about 5 minutes in and I have absolutely no idea which system and which connection is related to which FPS results. Now that I've read everything in the wall of text the overlay I can see the OCuLink is on the left and the direct PCI-e is on the right. You should label the results so it's obvious. Next, your test setup is simply bad. Changing CPU, RAM, Windows install, drivers?, BIOS, available power, and who knows what else is too many variables. You should have tested multiple ways all on the 8600G system. If you're trying to apply the scientific method to get useful information, you have to limit your variables to as few as possible, ideally one. The eGPU dock you linked to includes all you need to do all the testing on the MSI B650-P board. And/or you can buy a PCI-e to OCuLink adapter for $20-30. Then you could reasonably test 4-6 scenarios: PCI-e 4.0 x8 (limited by the card) PCI-e 3.0 x8 (limited by the BIOS) PCI-e 4.0 x4 (limited by using PCI_E3 slot connected to the chipset, if the BIOS doesn't allow limiting PCI_E1 to only 4 lanes. You could also tape off lanes on the card, but I wouldn't recommend that.) PCI-e 4.0 x4 via OCuLink from M.2 Slot PCI-e 4.0 x4 via OCuLink from PCI-e Adapter Dock via Thunderbolt apparently people want to see. Then you could see what is causing the loss, if there actually is any without changing to much slower RAM. You'd be able to demonstrate bandwidth restrictions and possible overhead from either OCuLink or the various adapters.
Check Out The Merch! evo-tech-shop.fourthwall.com
B580 | amzn.to/40aXZGI
GMKtec M7 Pro | amzn.to/3ZScihV
eGPU Dock | s.click.aliexpress.com/e/_onLfAsn
Hmmm. Is resizable BAR enabled on the oculink PC? Thst might explain this difference too.
This is a really cool test. Looking forward to seeing the resulrs here. Thanks!
Any chance of trying with thunderbolt? There are like 5 computers with oculink.
oculink has historically been a very diy thing
Go m.2 it's 10% faster than oculink
@@MikeJones_OfficialUhh, Oculink adapters already use the m.2 port? When you modify a laptop as i have what it serves to do is expose those pcie lanes for unconventional use.
Look how the turns have tabled xD
Intel GPU with AMD cpu build haha. No one would have believed this 10 years ago
No one could have imagined an Intel GPU 10 years ago. Intel is probably still relevant because of arc and Iris brands, just like AMD made Gold with ATI.
OCuLink is forever better than Thunderbolt 3 or Thunderbolt 4 or USB4
Probs but how many people think about when buying a laptop
@@james8449100 as a matter of fact:
Laptops with OCuLink are getting more mainstream and more cheaper
Laptops with Thunderbolt still Hilariously expensive. Mainly because the Thunderbolt protocol Royalty isn't cheap.
It will speak for itself in less than 2 years
Not really, no. You're limited hardcore with OcuLink as it only really uses eGPUs. TB3/4/USB 4 allows you to also hook up super fast drives and docking stations as well as do other displays stuff through it
@@PixelatedWolf2077 raw PCIe signal of Thunderbolt 4 or USB4 is ridiculously slower than OCuLink
Besides, Laptop with OCuLink also comes with USB4 alongside it anyway. But if you plug eGPU into USB4 instead of OCuLink , that's your mistake
@niezzayt3809 Ah, another thing you reminded me. The price isn't always worth it to run OcuLink, especially if you're going to run a handheld or you have a "business" PC build and want extra performance
Resizable bar or lack off is hurting the B580
Yeah, it was hard to keep track of what was connected to what and what settings were on what
Intel ARC without Re-BAR obliterates it's performance - even AV1 encoding ist hampered by more than 50% on my A380 and A310.
I modified my best laptop to have an oculink port, and was gonna buy this gpu but... no rebar support :/
Which laptop is it
What cpu
@@matgaw123 I5-8250U Dell Inspiron 15 7573, its all ive got
I believe my account is shadow banned for using revanced, about 50% of my comments get deleted. But anyways all ive got right now is my I5-8250U Dell Inspiron 15 7573 that i installed oculink in and a E495 thinkpad
@@naomiwolf8944 there is a mod called
ReBarUEFI
It can enable rebar on most UEFI systems
Give me money I'll mass produce a motherboard with 4 ram slots, 2 CPU slots, 2 mini PCIe slots, and 4 M.2 slots. I will call it the Decade+GF Ultra motherboard
For stats.
Gonna buy B580 as soon as it becomes available again.
I've watched a few B580 videos lately, so YT recommended this one. And I watched it. But it's time for some constructive criticism.
First, there's way too much yapping at the beginning. Describe what you're doing. Give a few details about the test system, but only relevant stuff. Don't introduce even more unknowns and variables that you don't have to. What do you mean you don't know how many lanes the 8600G gives the x16 slot? Put the card in, open up GPU-Z, and find out. Record and publish only stuff you know. You don't need "I will find out how my system is even set up." in an overly long intro.
Second, I was casually watching on my phone at the start while doing something else. By the time you finally got around to showing benchmark numbers, you had my attention again. But you didn't label which system and results were which. I've paused the video while typing this about 5 minutes in and I have absolutely no idea which system and which connection is related to which FPS results. Now that I've read everything in the wall of text the overlay I can see the OCuLink is on the left and the direct PCI-e is on the right. You should label the results so it's obvious.
Next, your test setup is simply bad. Changing CPU, RAM, Windows install, drivers?, BIOS, available power, and who knows what else is too many variables. You should have tested multiple ways all on the 8600G system. If you're trying to apply the scientific method to get useful information, you have to limit your variables to as few as possible, ideally one.
The eGPU dock you linked to includes all you need to do all the testing on the MSI B650-P board. And/or you can buy a PCI-e to OCuLink adapter for $20-30. Then you could reasonably test 4-6 scenarios:
PCI-e 4.0 x8 (limited by the card)
PCI-e 3.0 x8 (limited by the BIOS)
PCI-e 4.0 x4 (limited by using PCI_E3 slot connected to the chipset, if the BIOS doesn't allow limiting PCI_E1 to only 4 lanes. You could also tape off lanes on the card, but I wouldn't recommend that.)
PCI-e 4.0 x4 via OCuLink from M.2 Slot
PCI-e 4.0 x4 via OCuLink from PCI-e Adapter
Dock via Thunderbolt apparently people want to see.
Then you could see what is causing the loss, if there actually is any without changing to much slower RAM. You'd be able to demonstrate bandwidth restrictions and possible overhead from either OCuLink or the various adapters.