Yeah, i've got an RTX 4080 Gigabyte WindForce V2 and i've got the same FPS with my Desktop with an 7900X, i am surprised that you have such good performance with an eGPU, i may order a Mini PC to low my electricity bill. Good review
Yes, it's strange since this sollution only runs on 4 pcie lanes, instead of 16 as it does hooked up to the motherboard. Seen several others testing oculink with about 30% loss of performance, which is reasonable due to the considerably lower bandwidth of the oculink. Maybe this dude didn't even put the gpu in the PCIe x16 4.0 slot on his motherboard. That would explain the results he gets. I don't trust this "review" at all after seeing so many others showing different results/numbers.
@@MrMandelll No way Oculink is losing 30%, you are getting confuse with Thunderbolts PCIE 3, I thought that the Oculink would lose between 15% to 5%, most optimistic reviews i saw were talking about 3%, i knew Oculink was good but i didn t know it was that good until i watch this review, unless this channel is faking results and using desktop for testing, i swear you won t get better FPS with a Desktop or at least with my setting level ii got my RTX 4080 S Gigabyte WindForce running with my 7900X, coz i got the same FPS ranging in Cyberpunk and the Witcher ( don t know about other games he tried ), as the guy was testing setting maxed with DLSS and FR, he can t have faked settings, but could fake MSI component name and use desktop and rename to mobile, however it s interesting, i would like to see more test from other channel with my cards, but there are too few videos on Oculink with high ends cards and mobile CPU.
@@MrMandelllOculink can Be PCIE 4.0 or 3.0. Ryzen 5000U and 5000H series are PCIE 3.0. Ruzen 6000-8000U/H are PCIE 4.0. The performance loss in PCIE X4 4.0 is minimal. PCIE 4.0 X4 = 3.0 X8. Look videos comparing RTX4090 PCIE 3.0 X8 vs PCIE 4.0 X16
Hi, thank you for the review, as someone who's on a split between having a laptop(movement to different places from time to time), building a pc (planning to play AAA games with no compromise), or best of both worlds: a minipc, this review is really helpful. My only worries on the oculink setup is the latency, and the performance penalty coming from the inherent limitations of the oculink cable, cheers , greetings from the Philippines
Hi, thanks you for your comment, you should not be worry about latency, because the Oculink port is PCIE 4.0 X 4 and its taking the full power of the 4 lanes of PCIE 4 and is taking directly information from the GEM 12 CPU, just like a GPU would on a Desktop Motherboard unlike Thunderbolt cables You can compare my CyberPunk FPS results with with the ZwormZ Cyberpunk RTX 4080 SUPER (YT Video) results. There are no performance lost at all, even with an RTX 4080 SUPER, that's why the Oculink is a class of its own in the Mini PC world. Note that my resolution during the test were mostly made on a 3440 X 1440P display monitor.
@@aekprm435 Hello again😃 Yes the K8 Should have similar performances than the GEM 12 as long as the K8 has as good as GEM 12 RAM. I will try to get the K8 Plus ASAP. I am also interested in testing the Beelink GTI 14 with its PCIE 4 X 8 lanes GPU Dock station.
@@aekprm435 i am glad you to hear that you have ordered the K8 plus too, in my review i will compare the K8 plus performance against the GEM 12 as these 2 Mini PC's model have the same specs. Have a nice day friend
I have a Laptop with Thunderbolt 4, can i use and external 4090? Should i buy an enclosure or are there already packed external 4090 Just tò connect tò the pc? How much performance drop on TB4? Compared tò oculink? Also, does a 4090 loose likes 40% or Power because It needs more bandwidth and Speed, while 4080 looses maybe only 10%, meaning that its more convenient a 4080 instead of a 4090? Or the performance spread Is enough? Like 4090 Time Spy 27000 and 4080 Time Spy only 22000? Anyway, best eGPU solution for a TB4 Laptop? Is there a ready solution with enclosure + card? If there Is a ready solution, how does it perform compared tò a self made best enclosure + best card?
Hi, i have released a video with the Aoostar AG02 eGPU Dock, the video deals with Performance loss ratio between Oculink and Thunderbolt as the AG02 eGPU Dock have both Oculink and Thunderbolt port,. An RTX 4090 won t lose 40% performance via Oculink with an AMD 8000 or 7000 series, i ve seen a loss of 8% in Time Spy with an RTX 4090 with a Minisforum UM780 XTX via Oculink port. I don t know about Thunderbolt loss tough. Usually eGPU enclosure don t support large Graphic Cards, i would add that different eGPU Dock would get better results than Others, OCUP4 eGPU Dock belong to the Best eGPU Dock regarding Performance Loss.
@@aekprm435 it's true, but the price difference is significant, it is also normal to witness better performances on the GEM 12 compared to the M7 as the GEM 12 Specs are pretty much the best mobile hardware made in early 2024. Also note that OCUP4 eGPU Dock is slightly giving better Performances with the RTX 4080 S than the AG02 eGPU Dock tested with the M7. Thanks you for your comment.
@@aekprm435 Hi, this is an honest Price, you essentially get what you paid for. and it s 19$ Cheaper than the Aoostar official website price for this model, so it can be concidered as a good deal.
@@vlad_8011 Hi, Tks you for your comment, This video purpose is not "World Record FPS in Cyberpunk", i did a bunch of game testing and also several 3DMARKS Bench to give a maximum of reference for comparison regarding this Mini PC Model + eGPU set up. Concerning Cyberpunk I was lazy playing throught the all endless Cyberpunk Dialogue to reach the City part so i picked the fastest story to get to car driving. But yes you are right the FPS will be lower but not destroyed at all, i am preparing a new video with a new Mini PC with Thunderbolt EGPU and even with this much lower Hardware set up i am way above 100 FPS at Max setting with RT Max, it s not even close to be called destroyed, it s still in the excellent user gaming zone experience with Max Setting.
Hello, same setup here (7900 Gre for graphic car). Impossible to detect the graphic car with the ocup4v2 EGPU. Oculink works great with GDP G1 v2. I have the green led start when I launch the GEM12. Si I think it's a good think. But nothing detected in windows (and the graphic card doesn't move). OCUp4 on M2 mode (small switch on the side) Any idea ? Somthing to change in the bios (trying disable igpu not working)? thanks
@@anakinskywalker7079 Hi, No, you don t have to change anything in the BIOS, if you have the same set up then it should work just fine, Rx 7900 GRE should work. 1- Check that ypur HDMI cable is connected to your monitor port, and unplug your HDMI cable on your GEM 12 if it s plugged. 2- Uninstall AMD Adrenaline followed by DDU to uninstall your previous GPU AMD drivers. 3- Reinstall AMD Adrenaline. 4- disable your igpu in Device manager in Monitor display option I hope those tip would fix your issue
@@anakinskywalker7079 Do you have an NVME M2 adaptator? Why have you mentioned "OCUP4 on M2 Mode (small switch on the side)"??? With the GEM 12 you are not supposed to use the NVME M2 adaptator.
@@anakinskywalker7079 You don t have the same setting than me obviously, there is not such a thing as TGX position in my setting, i am sorry i couldn t help you to fix your issue.
I know it's a review...but is raytracing supposed to look so ugly? Every surface is shiny which looks totally unnatural. If this is what the ppl want, then i don't want that crap because it looks like garbage. Or with other words said, normal surfaces (dry surfaces doesn't look shiny) appear to be unrealistic.
@@BaamAlex Hi, it depends of the game and people, personally i like RayTracing, i wouldn t enjoy Fallout 4 without heavy Raytracing, or the witcher 3 Water ain t look natural without Raytracing, you won t be able to run Minecraft SEUS PTGI HRR3 without RayTracing, but for Cyberpunk or Akan Wake 2 or Forza Horizon 5 RayTracing Off don t bother me. Also maybe the encoding of the video capture ain t look as good as it is when you watch it live on your wide monitor.
@@Ghost-Keyboard That and I first played Mario in 1987, 2 years after my birth, at least from what I can remember. The first console I owned that I remember was the NES, which is where Chain Chomp made its first appearance in Super Mario Bros. I started young with Atari, Commodore 64 and the first console that I owned that I remember was the NES. Mario debuted 6 months and a day after I was born. Born March 12, 1985, Mario debuted September 13, 1985 in Japan and sometime in October, 1985 in the USA
@@乾淨核能 unfortunately RUclips isn't allowing me to edit or post links as my channel is to small. and not allowing me too post link either in this comment, sorry for the inconvenient, i ll advice you to type "Ocup4" in the Aliexpress search bar and pick the seller with the fastest delivery and the most sold item
Is it possible to connect two GPU to the Gem12? Let's say you convert one of the free M.2 PCIE4x4 slots into another Occulink port with 64gbps bandwidth. If one were to connect two separate external Nvidia GPU's to each Occulink port... would it be able to function in the same way as if you connected two separate GPU to 4x4 PCIE slots on a regular motherboard? The purpose for my question is for running large AI / LLM models, which require LOTS of VRAM. PyTorch and other AI frameworks typically allow for distributed processing across multiple GPU, pooling their VRAM together.
Yeah, i've got an RTX 4080 Gigabyte WindForce V2 and i've got the same FPS with my Desktop with an 7900X, i am surprised that you have such good performance with an eGPU, i may order a Mini PC to low my electricity bill. Good review
Yes, it's strange since this sollution only runs on 4 pcie lanes, instead of 16 as it does hooked up to the motherboard. Seen several others testing oculink with about 30% loss of performance, which is reasonable due to the considerably lower bandwidth of the oculink. Maybe this dude didn't even put the gpu in the PCIe x16 4.0 slot on his motherboard. That would explain the results he gets. I don't trust this "review" at all after seeing so many others showing different results/numbers.
@@MrMandelll No way Oculink is losing 30%, you are getting confuse with Thunderbolts PCIE 3, I thought that the Oculink would lose between 15% to 5%, most optimistic reviews i saw were talking about 3%, i knew Oculink was good but i didn t know it was that good until i watch this review, unless this channel is faking results and using desktop for testing, i swear you won t get better FPS with a Desktop or at least with my setting level ii got my RTX 4080 S Gigabyte WindForce running with my 7900X, coz i got the same FPS ranging in Cyberpunk and the Witcher ( don t know about other games he tried ), as the guy was testing setting maxed with DLSS and FR, he can t have faked settings, but could fake MSI component name and use desktop and rename to mobile, however it s interesting, i would like to see more test from other channel with my cards, but there are too few videos on Oculink with high ends cards and mobile CPU.
@@MrMandelllOculink can Be PCIE 4.0 or 3.0.
Ryzen 5000U and 5000H series are PCIE 3.0.
Ruzen 6000-8000U/H are PCIE 4.0.
The performance loss in PCIE X4 4.0 is minimal.
PCIE 4.0 X4 = 3.0 X8.
Look videos comparing RTX4090 PCIE 3.0 X8 vs PCIE 4.0 X16
@@MrMandelll woaw my comment was deleted, all i said was no way Oculink is losing as much as 30%,
30% loss is what Thunderbolt PCIE 4 eGPU would get.
Maybe this little monster be really good with 3d design and cad programs. Nice review!
Hi, thank you for the review, as someone who's on a split between having a laptop(movement to different places from time to time), building a pc (planning to play AAA games with no compromise), or best of both worlds: a minipc, this review is really helpful.
My only worries on the oculink setup is the latency, and the performance penalty coming from the inherent limitations of the oculink cable,
cheers , greetings from the Philippines
Hi, thanks you for your comment, you should not be worry about latency, because the Oculink port is PCIE 4.0 X 4 and its taking the full power of the 4 lanes of PCIE 4 and is taking directly information from the GEM 12 CPU, just like a GPU would on a Desktop Motherboard unlike Thunderbolt cables
You can compare my CyberPunk FPS results with with the ZwormZ Cyberpunk RTX 4080 SUPER (YT Video) results. There are no performance lost at all, even with an RTX 4080 SUPER, that's why the Oculink is a class of its own in the Mini PC world.
Note that my resolution during the test were mostly made on a 3440 X 1440P display monitor.
K8 PLus is Out!!! should be same performance GEM12 8845 awaiting ur review
@@aekprm435 Hello again😃 Yes the K8 Should have similar performances than the GEM 12 as long as the K8 has as good as GEM 12 RAM. I will try to get the K8 Plus ASAP. I am also interested in testing the Beelink GTI 14 with its PCIE 4 X 8 lanes GPU Dock station.
Update info I confirm that i will receive the GMKtec K8 Plus in the following week, my next publication will be a review of the GMKtec K8 Plus.
U r the man!!
@@Ghost-Keyboard my K8+ will arive next week
@@aekprm435 i am glad you to hear that you have ordered the K8 plus too, in my review i will compare the K8 plus performance against the GEM 12 as these 2 Mini PC's model have the same specs. Have a nice day friend
Thank you for the review
My pleasure!
Thanks for the review.
How long can the PSU be on the OcuP4, or does it mount over the 24p connector?
@@MrMandelll hi, sorry but i don t understand your question, maybe you could watch my eGPU video to help you visualize the whole eGPU setting.
I have a Laptop with Thunderbolt 4, can i use and external 4090? Should i buy an enclosure or are there already packed external 4090 Just tò connect tò the pc? How much performance drop on TB4? Compared tò oculink? Also, does a 4090 loose likes 40% or Power because It needs more bandwidth and Speed, while 4080 looses maybe only 10%, meaning that its more convenient a 4080 instead of a 4090? Or the performance spread Is enough? Like 4090 Time Spy 27000 and 4080 Time Spy only 22000?
Anyway, best eGPU solution for a TB4 Laptop? Is there a ready solution with enclosure + card? If there Is a ready solution, how does it perform compared tò a self made best enclosure + best card?
Hi, i have released a video with the Aoostar AG02 eGPU Dock, the video deals with Performance loss ratio between Oculink and Thunderbolt as the AG02 eGPU Dock have both Oculink and Thunderbolt port,.
An RTX 4090 won t lose 40% performance via Oculink with an AMD 8000 or 7000 series, i ve seen a loss of 8% in Time Spy with an RTX 4090 with a Minisforum UM780 XTX via Oculink port. I don t know about Thunderbolt loss tough.
Usually eGPU enclosure don t support large Graphic Cards, i would add that different eGPU Dock would get better results than Others, OCUP4 eGPU Dock belong to the Best eGPU Dock regarding Performance Loss.
SO better Gem 12 than M7
@@aekprm435 it's true, but the price difference is significant, it is also normal to witness better performances on the GEM 12 compared to the M7 as the GEM 12 Specs are pretty much the best mobile hardware made in early 2024.
Also note that OCUP4 eGPU Dock is slightly giving better Performances with the RTX 4080 S than the AG02 eGPU Dock tested with the M7.
Thanks you for your comment.
@@Ghost-Keyboard GEM12 8845 for 450usd is ok price?
@@aekprm435 Hi, this is an honest Price, you essentially get what you paid for. and it s 19$ Cheaper than the Aoostar official website price for this model, so it can be concidered as a good deal.
@@Ghost-Keyboard k8+
:-)
Youre testing Cyberpunk with 4080 Super ON DESERT? Entering the city would destroy this framereate
@@vlad_8011 Hi, Tks you for your comment, This video purpose is not "World Record FPS in Cyberpunk", i did a bunch of game testing and also several 3DMARKS Bench to give a maximum of reference for comparison regarding this Mini PC Model + eGPU set up.
Concerning Cyberpunk
I was lazy playing throught the all endless Cyberpunk Dialogue to reach the City part so i picked the fastest story to get to car driving. But yes you are right the FPS will be lower but not destroyed at all, i am preparing a new video with a new Mini PC with Thunderbolt EGPU and even with this much lower Hardware set up i am way above 100 FPS at Max setting with RT Max, it s not even close to be called destroyed, it s still in the excellent user gaming zone experience with Max Setting.
@@Ghost-Keyboard OK, i just want you to know, in Cyberpunk best setting to test is japanese area (city) with lots of neon lights.
Hello, same setup here (7900 Gre for graphic car). Impossible to detect the graphic car with the ocup4v2 EGPU. Oculink works great with GDP G1 v2. I have the green led start when I launch the GEM12. Si I think it's a good think. But nothing detected in windows (and the graphic card doesn't move).
OCUp4 on M2 mode (small switch on the side)
Any idea ?
Somthing to change in the bios (trying disable igpu not working)?
thanks
@@anakinskywalker7079
Hi, No, you don t have to change anything in the BIOS, if you have the same set up then it should work just fine, Rx 7900 GRE should work.
1- Check that ypur HDMI cable is connected to your monitor port, and unplug your HDMI cable on your GEM 12 if it s plugged.
2- Uninstall AMD Adrenaline followed by DDU to uninstall your previous GPU AMD drivers.
3- Reinstall AMD Adrenaline.
4- disable your igpu in Device manager in Monitor display option
I hope those tip would fix your issue
@@Ghost-Keyboard no, tested. Thanks
@@anakinskywalker7079 Do you have an NVME M2 adaptator? Why have you mentioned "OCUP4 on M2 Mode (small switch on the side)"???
With the GEM 12 you are not supposed to use the NVME M2 adaptator.
@@Ghost-Keyboard No I have not. I will test again with switch on TGX position.
@@anakinskywalker7079
You don t have the same setting than me obviously, there is not such a thing as TGX position in my setting, i am sorry i couldn t help you to fix your issue.
I know it's a review...but is raytracing supposed to look so ugly? Every surface is shiny which looks totally unnatural. If this is what the ppl want, then i don't want that crap because it looks like garbage.
Or with other words said, normal surfaces (dry surfaces doesn't look shiny) appear to be unrealistic.
@@BaamAlex Hi, it depends of the game and people, personally i like RayTracing, i wouldn t enjoy Fallout 4 without heavy Raytracing, or the witcher 3 Water ain t look natural without Raytracing, you won t be able to run Minecraft SEUS PTGI HRR3 without RayTracing, but for Cyberpunk or Akan Wake 2 or Forza Horizon 5 RayTracing Off don t bother me.
Also maybe the encoding of the video capture ain t look as good as it is when you watch it live on your wide monitor.
10:13 That is a chain chomp
@@timclark_420 thanks you, you are a man of Culture
@@Ghost-Keyboard That and I first played Mario in 1987, 2 years after my birth, at least from what I can remember. The first console I owned that I remember was the NES, which is where Chain Chomp made its first appearance in Super Mario Bros. I started young with Atari, Commodore 64 and the first console that I owned that I remember was the NES. Mario debuted 6 months and a day after I was born. Born March 12, 1985, Mario debuted September 13, 1985 in Japan and sometime in October, 1985 in the USA
the egpu link is broken
@@乾淨核能 unfortunately RUclips isn't allowing me to edit or post links as my channel is to small. and not allowing me too post link either in this comment, sorry for the inconvenient, i ll advice you to type "Ocup4" in the Aliexpress search bar and pick the seller with the fastest delivery and the most sold item
@@Ghost-Keyboard thank you so much!!
@@乾淨核能 you are welcome, 😀
I hate the OCulink ports on the front
I agree, but "hate "maybe is a too strong statement for me.
This feels like an AI review, also come on man cropping out the Afterburner screen on 4080 setup side by side
@@Simulati0n i am sorry you didn t like this review, i will try to do better if i ever do an other review, thank you for watching
Is it possible to connect two GPU to the Gem12? Let's say you convert one of the free M.2 PCIE4x4 slots into another Occulink port with 64gbps bandwidth. If one were to connect two separate external Nvidia GPU's to each Occulink port... would it be able to function in the same way as if you connected two separate GPU to 4x4 PCIE slots on a regular motherboard?
The purpose for my question is for running large AI / LLM models, which require LOTS of VRAM. PyTorch and other AI frameworks typically allow for distributed processing across multiple GPU, pooling their VRAM together.
@@s.alizaidi6635 Hi, theorically the answer is "Yes" as the GEM 12 has 3 NVME M2 slot.
And the Oculink you used is no longer avaliable on AliExpress
@@timclark_420 Hi , How about Ocup4 V2?