finally someone mention the Nvidia driver overhead, when using my 12400f for Warzone at competitive settings rx6600 gives better frames than 3060ti while costing almost half the price...
CPU limit with a 12400F? That's a rather powerful CPU. Is overhead reallyy that impactful? Keep in mind that MWII and WZ2 LOVES amd cards, they are more performant versus the Nvidia counterpart
Its good that you accurately acknowledge the pros and cons here. If the RX6400 had proper video encoding, it would have been a powerhouse GPU especially for light video work. However, it's clear the RX6400 was a laptop GPU designed to be paired with AMD APUs. It would have been nice if AMD designed a workstation card WITH encoding like the RX6400 with slightly more Vram and similar power/ form factor. It would have been a budget competitor to the A2000.
@@Miyconst I suppose that depends on a few things. For starters, what does the professional world look like at the $150-300 price point for small workstation GPUs. Also, would the card be at least PCIE x8. I suspect if it is better than a GTX 1650 and it's between $180-200 there might be a market for the card in both professional and enthusiast SFF builds. There is a community ready to gobble up efficient small form factor cards especially under 75W
@@greenprotag its not the power, its the feature set that sells in a professional setting, who needs 4k 240hz gpu power when your just using hardware acceleration and encoding
they did make systems that worked together with onboard... but people shat on them ... only to get a p4 and be like wow! when its still using the same hardware its just accelerating it and adds video acceleeation libraries too.
@@pazsion I have done quite a bit of testing on both the RX6400 and P4 in the time this video has been out. The P4s are a bit of a challenge and obviously the drivers are not as optimized for gaming but it DOES run VERY well. The RX6400 is limited and that is disappointing from a feature standpoint, but it made an EXCELLENT mini console GPU especially for indie games, old PC games, light 1080p, or even a little retro emulation. With lowered expectations and understanding of limitations, both cards work well, but for different uses or at least different hardware configuration. The P4 I think is best for server/homelab use. Video encoding and even gaming in a virtualized environment helps embrace its original hardware configuration, but using it in different ways.
I have used the Tesla M40 for a few months, and one major issue I found is that not all games and applications respect your Windows graphics settings. While you can force most games and apps to use the high-performance GPU, some, especially OpenGL games and emulators, will only use the GPU that the monitor they're started on is connected to. This means unless you set up a ghost monitor on the Tesla as your primary display and mirror it onto your real monitor, those programs will only use your display card or IGP - or nothing at all (i.e., CPU rendering), as in the case of StarCraft 2. But such a setup brings more problems than it solves; for example, you can't run anything in exclusive full screen, only borderless window, or you lose signal to your real monitor.
SOrry i am just a bit missing details - my head is a bit lost soo... for Tesla P4 i need modded drivers, and simply plugin my screen to motherboard, and then use the laptop trick of having display running off iGPU ports, but using eGPU power for computing - right? just like on laptops with GPU? got any idea if i can do it on Linux with official nvidia drivers?
You got it almost right. Tesla P4 works just like a regular GPU and to use it you do the same settings as with a laptop, but you need official, but hidden Nvidia drivers for that.
@@Miyconst official, hudden drivers? hahaha that sounds sick other question (if you don't mind) - do you know any Chinese LGA 1151 motherboard that actually supports E3 xeon CPUs? Officially runs only on C236 / C262 chipsets :/
@@Karti200 search for China branded LGA 1151 motherboards on AliExpress and ask sellers if it would work with Xeons, because it's not a hardware limitation, it's a BIOS lock.
I love hardware like this, thanks for your coverage of the p4! But I have one question: Do you know if there is the same limitation with the generation of the CPU like with the Nvidia mining GPUs? If I remember correctly, for using a mining GPU for gaming, you need a 4th Gen Intel Core or higher, or an AMD Ryzen with integrated graphics, Would be interesting to know because i've got a 3rd Gen SFF Intel PC for free and I might consider a P4 for this project.
@@Miyconst Thanks for your reply! Yea it's a realy old CPU at that point. And I might need to reconsider this idea, because the SFF PC is small and cooling the P4 would certainly be a challenge.
I've noticed tesla p40s are quite cheap too right now (around 200/250$). It has the titan Xp (slightly better than 1080ti) core, 24gb of vram and won't be power limited like the p4. It would be very interesting if you could test it. I doubt it will be very convenient for games when the rx 6600 is around, but for rendering workloads it could be quite nice. Anyway great video as always, I love seeing this strange cards tested.
You can actually use the P4 with a dedicated gpu, even when it's an amd dgpu on windows 10. The graphic option for it might not be available so you can trouble shoot it by manually settings the EnableMsHybrid value in registry instead. Navigate to Computer\HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\Class\{4d36e968-e325-11ce-bfc1-08002be10318}, then set EnableMsHybrid = 1 on one gpu and EnableMsHybrid = 2 on the other. Windows 11 have multiple selections for gpus so there is no need to jank with them. Occasionally some applications would be confused and used the P4 instead of your output gpu but that's rare. I use it along with a Radeon VII because some games refuse to fix visual bug on older AMD cards, and i don't want to use anything other than the vega VII. So i have to make due with a P4 and manually troubleshoot it. They works great :)
I was always curious about using one GPU for "gaming" and another one as simply the display adapter. My point is, being able to use a modern GPU that only has digital outputs with an older one that still has analogue VGA for a CRT monitor. I have a GTX 960, which is the last generation to support analog, and I wondered if installing a newer GPU, say, a GTX 1660S, and setting it as the main one would introduce any lag when displaying the image through the 960. I've read some threads telling that eventually you wouldn't be able to use either the older or the newer GPU if they were the same brand, because of driver updates, but that wouldn't be case with different brands or iGPUs (unless they're AMD, maybe), because there are new motherboards that still comes with a VGA output, but it seems it wouldn't be a good solution given your results... HDMI to VGA or even DisplayPort to VGA adapters are an option, but I always thought it was an "ugly" option XD.
Why would you want an analogue monitor in 2022? Anyway, answering your question - it's possible, but the input lag will of course be higher. There is no way around it, if you process data using one GPU and then sending it to another GPU, it's extra work and extra data traveling.
@@Miyconst Simply because CRTs are the best, my friend, end of story, 69/420 XD. No, seriously, I'd much rather use a janky 17" CRT monitor than a blurry "high end" LCD, been doing so for at least 6 years now. But I already looked around and found a DisplayPort to VGA adapter that is a good solution, which is basically just a DAC (digital to analogue converter), which directly translates the signal. The main thing about CRTs on modern setups is having the system recognize the monitor (better with the monitor driver), so it has the correct EDID modelines and such, and most adapters don't do that like a direct connection to the GPU does, so I really thought having 2 GPUs would be better...
Доброго времени суток. Скажите ,а ее можно использовать с другими карточками от нвидии или от амд по типу как с теслами м40 ? Как выбрать приоритет чтобы изображение было с одной карточки а рабочая нагрузка на тесле ? Со встроенной графикой интел вроде понятно ,а с другими карточками что скажете ? Спасибо за ответ.
Hi, Miyconst i was wondering do you recommend an x99 motherboard named Machinista mr9a pro max/E5 mr9a Pro Max/x99 mr9a pro? I saw your other videos about the non-pro mr9a x99 and you've recommend it. So, i'm curious if you would recommend this Motherboard too. Or should i go am4 some r5 3600 right now are dropping prices and a320m would be a good idea.
As far as I know, the Pro version is the same as the original one, just some cosmetic changes. Anyway, I would pick R5 3600 over X99, unless you really need the X99 features.
@@Miyconst I really appreciate your reply despite being busy creating content thanks dude. Stay safe out there. Its not that i don't like to build a Ryzen system... X99 is way cheaper in my country.
AMD Radeon software has a cool feature that lets you set target fps range that can really smooth frame times. I’ve been using single slot RX 550 in a SFF pc and this feature lets me run games at smooth 45 fps without big dips and better temps too.
I've been thinking about getting a tesla P4 for a dedicated video encoder, but haven't made the decision yet because I thought if I ever needed a backup GPU to game on it wouldn't be a valid option, but seeing that there is a way to run it through an IGPU if I needed to makes the idea more palatable. I don't want to spend a whole lot on a dedicated hardware encoder for my server, and now knowing that I could use it as a backup in case my Vega fails, I think it might just be the right choice.
@@tanmaypanadi1414 I don't believe the Rx 6400 has encoding capabilities though, and am limited not necessarily half height but having no power connections
For your cooling solution, better would be to leave the top cover on, cut a hole in the cover and blank the side inside the case. That way you create a tunnel and cooling will be more efficient. This card could be a good solution for video encoding for plex servers etc.
I do 3D rendering using cuda cores. the tesla brings 2560 CUDA cores . so i wonder if it would be usefull as a pure render GPU on a budget. nvidia Iray can use more then one GPU for rendering.
Hi, do you think the p4 will Play yuzu Switch Emulator well? I try to build a gaming Emulator from a Lenovo sff PC with i5 9400. It have only one Slot for GPU.
I am late for that but - how you did used 80mm fan? You sure you didn't used 40mm fan? card is like 50mm on 1 axis only, i used 3d printed bracket for 60mm fan and fan itself is much bigger than the card :D
Hello, Can you do a video on tesla P40 gpu. You will find it on ebay around 200 euro. Most likely it will perform somewhere like quadro p6000. It have 24gb vram. Draw back will the no video output.
Have you ever tested that Tesla P4 card on system that has fully legacy bios, and no above 4g decoding, and no Resizable bar? Because I want to buy that card for my homelab workstation Dell R5500(dual LGA 1366 xeon), but it has fully legacy bios, no above 4g decoding and no Resizable bar, and I am wondering if I will be able to boot with my gtx 1660 Ti and that card in other slot, because gtx 1660 Ti working perfectly, but I heard Tesla cards need above 4g decoding, UEFI.
if anyone is following this in 2024, the most recent driver version that will install in windows is 16.7 (538.78) the 17+ drivers don't seem to be compatible
I saw on another p4 vid, someone said to disable the ecc on the p4 vram, as the ecc on the p4 vram is not needed for gaming, and makes the card have lower clocks, and makes it so you can only use 4gb of the 8gb on the p4.
@@Miyconst Im gonna just copy and paste the comment so I dont mess it up. - "He's not talking about ECC for system memory. He's talking about ECC for the onboard memory for the GPU. Gaming cards never have it. So, unless you've used a Tesla card before, you've never used this type of ECC before. ECC was introduced with the Tesla cards because some of them have a lot more memory than their gaing counterparts, but also most importantly because they are targeted at GPU compute applications. If you're a scientist running a complex simulation that takes days, weeks, or even months to run, the last thing you want to do is get erroneous results because at some point during that time a bit got flipped by a cosmic ray. If that happens your entire result could be bogus and you might not even know it. For gaming, it's a different story because frames are rendered in milliseconds or less, which means there's less time for something like that to happen and even if it did, it'd probably just mean the color of one pixel on the screen might be off a bit or if it happened in the z buffer maybe there would be a very slight rendering error on one pixel. In essence, it's highly unlikely you'd ever even notice it. In fact, it'd probably be almost impossible to notice it, but for compute application, and cypto work, it would be a huge deal. Therefore, the Tesla cards have ECC support and they usually have it enabled by default. The downside is the way ECC is done on the cards, it takes away a portion of the memory. It also takes away a bit of memory bandwidth and decreases performance a bit on applications where bandwidth matters more. The way system meory does ECC, it usually has extra bits and those extra bits have extra connections on the bus so that the ECC is done in parallel. That's not the case here, so the ECC bits take up memory and narrow the bandwidth. So, for gaming it makes perfect sense to disable it and it's highly unlikely to affect system stability. Anyway, I believe to toggle it on or off, you must use either the Nvidia SMI or the GPUmodeswitch utility."
I didn't know about Nvidia Tesla, is it a good choise for my case? my pc just acepts low profles gpu. I thought about Rx6400, Gtx1650 and Quadro T600, but I found a T4 8GB more cheaper thant the others.
@@Miyconst Yes. answered to me. T600 ~170 euros, RX6400 ~200 euros and GTX 1650 ~230 euros... P4, used of course, but just 90 euros, we have the winner. Thanks for your video. Just a question, Can I still continue using my GT 1030 and yours videos output without problems? right?
I am not sure if anyone is paying attention but I went to the NVIDIA web site and they have 511.65 ? Is this modified to still not work or did they release it again?
@@Miyconst Awesome. I contemplated getting one of these, but didn't decide on it until I saw your video on how to set it up. I plan on installing it in a dell 7040.
Hope this comment reach you. I have an old dell 7010 sff usingg as nas. I wanna add the P4 for some 3d stuff but when i install the card. it slow down the whole system and task manager said that 100% CPU used system interupt. withou the p4 it work just fine. I try to use adjust in bios such as video Intel. UEFI and so on.
The comment has reached me but I have no clue what to suggest you as the provided description is way too vague. It can be software error or hardware error.
@@Miyconst thanks for your reply. i did figure out. so i did try 2 setup igpu and P4 - the Igpu is only help me to few resolution - is there any change that i can fix it gt 710 and p4 - it allow me to choose resolution to my screen but the p4 driver is being uninstall cause i thing the system not allow 2 driver . Is there any change that i can have the 2k output with any of my setup Thank you in advance for help.
the pcie2-4 all would be x16 unless you put other pcie cards in the system. if those cards dont limit themselves to x1 than you may lose performance as most boards at this time dont have more than 16 parhways to the cpu and its shared across the slots.
Hi, I registered on Discord, I posted my hardware information, in the #nvidia tesla section, still no solution, I followed these steps again, I reinstalled Windows 10, nothing to do, the driver installs, in device manager it appears detected with an exclamation point, it doesn't enable correctly, I tried both the drover version of the video and the latest version indicated in the link, nothing to do.😢
Recently purchased a Tesla K80 because the deal seemed too good to be true (63 USD on Ebay)... After days of tinkering i've finally managed to cool it down to stable temps with two crazy loud Xbox 360 S fans and three 80mm server fans on the back of the board because the VRMs heated up too much. With all of that hassle, I haven't noticed an important performance gain in video editing applications over my GTX 1050 Ti. While having 24GBs of VRAM might sound cool, in practice unless you're running some heavy computing task like physics simulation or AI, having such a power-hungry yet not very efficient beast in your system kinda sucks. Now I think about selling the card and getting either a 2060/1070ti or a Tesla P4, as it shares the same chip with the 1050Ti but has the additional 4GB of VRAM that allows the chip to unleash its full performance in video editing and AI processing tasks. The lesson from this... Do your research before indulging into the cheap server hardware mess.
I don't understand how you notice the input lag. I use a Tesla M40 for gaming and it passes the image through another gpu and I cant notice anything. By your logic I should have double the input lag because the image travels from the gpu to my cpu and then back to my disp out gpu.
I can't say that the input lag will be double as much, but it's surely significantly bigger. How do I notice it? Well.. in my case it was so noticeable that I can feel it by just moving mouse while observing game's reaction.
@@Miyconst I have experimented with it and for some reason using igpu has extra input lag. My bad. This is kinda stupid, maybe the processor graphics aren't good enough I guess. Also noticed that the performance with no igpu ( using another cheaper GPu as a display out) is way better, sometimes using igpu meant I had to turn down my graphics from high to low because of the performance impact.
Саме та ситуація коли напрягався робив налаштування щоб ні яке за Богор не лайно не було ви фірі знайшов попроси характеристики потрібних цікавих мені відеокарти тут хера все те саме тільки навпаки навпаки зовсім не на тій мові і зовсім нічого не зрозумілааа
finally someone mention the Nvidia driver overhead, when using my 12400f for Warzone at competitive settings rx6600 gives better frames than 3060ti while costing almost half the price...
I have not seen the video yet but my guess is it's the scheulder. Amd has a dedicated hardware scheduler while Nvidia uses the CPu
CPU limit with a 12400F? That's a rather powerful CPU. Is overhead reallyy that impactful? Keep in mind that MWII and WZ2 LOVES amd cards, they are more performant versus the Nvidia counterpart
@@Gurka9999 yes, warzone is super cpu intensive, i tested max fps numbers with 1080p low settings, gpu is not maxed out
that only because you are cpu limited, the price of the graphocs cars is meaningless.
omg FINALY!! with the link to the driver i was able to use my P40! :DD
thats worth a like a sub
Its good that you accurately acknowledge the pros and cons here. If the RX6400 had proper video encoding, it would have been a powerhouse GPU especially for light video work. However, it's clear the RX6400 was a laptop GPU designed to be paired with AMD APUs. It would have been nice if AMD designed a workstation card WITH encoding like the RX6400 with slightly more Vram and similar power/ form factor. It would have been a budget competitor to the A2000.
They would gladly do that for $200+, but at that price nobody would buy it.
@@Miyconst I suppose that depends on a few things. For starters, what does the professional world look like at the $150-300 price point for small workstation GPUs.
Also, would the card be at least PCIE x8. I suspect if it is better than a GTX 1650 and it's between $180-200 there might be a market for the card in both professional and enthusiast SFF builds. There is a community ready to gobble up efficient small form factor cards especially under 75W
@@greenprotag its not the power, its the feature set that sells in a professional setting, who needs 4k 240hz gpu power when your just using hardware acceleration and encoding
they did make systems that worked together with onboard... but people shat on them ... only to get a p4 and be like wow! when its still using the same hardware its just accelerating it and adds video acceleeation libraries too.
@@pazsion I have done quite a bit of testing on both the RX6400 and P4 in the time this video has been out. The P4s are a bit of a challenge and obviously the drivers are not as optimized for gaming but it DOES run VERY well. The RX6400 is limited and that is disappointing from a feature standpoint, but it made an EXCELLENT mini console GPU especially for indie games, old PC games, light 1080p, or even a little retro emulation. With lowered expectations and understanding of limitations, both cards work well, but for different uses or at least different hardware configuration. The P4 I think is best for server/homelab use. Video encoding and even gaming in a virtualized environment helps embrace its original hardware configuration, but using it in different ways.
I have used the Tesla M40 for a few months, and one major issue I found is that not all games and applications respect your Windows graphics settings.
While you can force most games and apps to use the high-performance GPU, some, especially OpenGL games and emulators, will only use the GPU that the monitor they're started on is connected to. This means unless you set up a ghost monitor on the Tesla as your primary display and mirror it onto your real monitor, those programs will only use your display card or IGP - or nothing at all (i.e., CPU rendering), as in the case of StarCraft 2. But such a setup brings more problems than it solves; for example, you can't run anything in exclusive full screen, only borderless window, or you lose signal to your real monitor.
Yeah, a headless GPU is coming with all sorts of issues.
It's insane how power efficient the rx 6400 is.
Might also be part of why the RX6400 is getting beat out by the P4 though.
@@KiraSlithI'll take the 6400's stability over the p4s numbers any day😂
SOrry i am just a bit missing details - my head is a bit lost
soo... for Tesla P4 i need modded drivers, and simply plugin my screen to motherboard, and then use the laptop trick of having display running off iGPU ports, but using eGPU power for computing - right? just like on laptops with GPU?
got any idea if i can do it on Linux with official nvidia drivers?
You got it almost right. Tesla P4 works just like a regular GPU and to use it you do the same settings as with a laptop, but you need official, but hidden Nvidia drivers for that.
@@Miyconst official, hudden drivers? hahaha that sounds sick
other question (if you don't mind) - do you know any Chinese LGA 1151 motherboard that actually supports E3 xeon CPUs? Officially runs only on C236 / C262 chipsets :/
@@Karti200 sadly no, Intel has locked E3 Xeons from the consumer LGA 1151 motherboards.
@@Miyconst darnit, i managed to get my hands on 1240v5 and 1270v5 but i cannot find any motherboard that could use them :<
oh well
@@Karti200 search for China branded LGA 1151 motherboards on AliExpress and ask sellers if it would work with Xeons, because it's not a hardware limitation, it's a BIOS lock.
The quality of your videos is so great!
I love hardware like this, thanks for your coverage of the p4!
But I have one question: Do you know if there is the same limitation with the generation of the CPU like with the Nvidia mining GPUs?
If I remember correctly, for using a mining GPU for gaming, you need a 4th Gen Intel Core or higher, or an AMD Ryzen with integrated graphics,
Would be interesting to know because i've got a 3rd Gen SFF Intel PC for free and I might consider a P4 for this project.
I am not sure about the compatibility, but the iGPU is certainly too weak in those old 3/4th generation Core CPUs, thus I wouldn't recommend it.
@@Miyconst Thanks for your reply!
Yea it's a realy old CPU at that point. And I might need to reconsider this idea, because the SFF PC is small and cooling the P4 would certainly be a challenge.
I've noticed tesla p40s are quite cheap too right now (around 200/250$). It has the titan Xp (slightly better than 1080ti) core, 24gb of vram and won't be power limited like the p4. It would be very interesting if you could test it. I doubt it will be very convenient for games when the rx 6600 is around, but for rendering workloads it could be quite nice. Anyway great video as always, I love seeing this strange cards tested.
I also tempted to buy a P40 for testing, but the prices are not that good, still.
@@Miyconst I would like to see the p40 and the a2000 even the 6 gb compared for low profile solutions which I have seen at 219 and 250 respectively.
@@jamesdavis8726 where did you see A2000 for just $250? In Sweden they are more like $450.
For a while they were available for half that (2 months ago; eBay)
Cheers! Thanks to your great video! I will save up and buy the rx 6400 instead, was eyeing the p4 for months. Might still get one just for shiggles!
You can actually use the P4 with a dedicated gpu, even when it's an amd dgpu on windows 10. The graphic option for it might not be available so you can trouble shoot it by manually settings the EnableMsHybrid value in registry instead.
Navigate to Computer\HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\Class\{4d36e968-e325-11ce-bfc1-08002be10318}, then set EnableMsHybrid = 1 on one gpu and EnableMsHybrid = 2 on the other. Windows 11 have multiple selections for gpus so there is no need to jank with them.
Occasionally some applications would be confused and used the P4 instead of your output gpu but that's rare.
I use it along with a Radeon VII because some games refuse to fix visual bug on older AMD cards, and i don't want to use anything other than the vega VII. So i have to make due with a P4 and manually troubleshoot it. They works great :)
Total g for the link for the drivers ✊😑
Hi . What do you think about those chinese rx 580 like the szmz varietes and the maxsun
What do recommend
I would rather stay away from those.
@@Miyconst thank you, i share the same opinion, i think they are re-using old rx ships on a new pcb
I was always curious about using one GPU for "gaming" and another one as simply the display adapter. My point is, being able to use a modern GPU that only has digital outputs with an older one that still has analogue VGA for a CRT monitor. I have a GTX 960, which is the last generation to support analog, and I wondered if installing a newer GPU, say, a GTX 1660S, and setting it as the main one would introduce any lag when displaying the image through the 960. I've read some threads telling that eventually you wouldn't be able to use either the older or the newer GPU if they were the same brand, because of driver updates, but that wouldn't be case with different brands or iGPUs (unless they're AMD, maybe), because there are new motherboards that still comes with a VGA output, but it seems it wouldn't be a good solution given your results... HDMI to VGA or even DisplayPort to VGA adapters are an option, but I always thought it was an "ugly" option XD.
Why would you want an analogue monitor in 2022? Anyway, answering your question - it's possible, but the input lag will of course be higher. There is no way around it, if you process data using one GPU and then sending it to another GPU, it's extra work and extra data traveling.
@@Miyconst Simply because CRTs are the best, my friend, end of story, 69/420 XD. No, seriously, I'd much rather use a janky 17" CRT monitor than a blurry "high end" LCD, been doing so for at least 6 years now. But I already looked around and found a DisplayPort to VGA adapter that is a good solution, which is basically just a DAC (digital to analogue converter), which directly translates the signal. The main thing about CRTs on modern setups is having the system recognize the monitor (better with the monitor driver), so it has the correct EDID modelines and such, and most adapters don't do that like a direct connection to the GPU does, so I really thought having 2 GPUs would be better...
Can you re-upload the drivers? Download link gives 404
Доброго времени суток. Скажите ,а ее можно использовать с другими карточками от нвидии или от амд по типу как с теслами м40 ?
Как выбрать приоритет чтобы изображение было с одной карточки а рабочая нагрузка на тесле ? Со встроенной графикой интел вроде понятно ,а с другими карточками что скажете ? Спасибо за ответ.
Hi, Miyconst i was wondering do you recommend an x99 motherboard named Machinista mr9a pro max/E5 mr9a Pro Max/x99 mr9a pro? I saw your other videos about the non-pro mr9a x99 and you've recommend it. So, i'm curious if you would recommend this Motherboard too. Or should i go am4 some r5 3600 right now are dropping prices and a320m would be a good idea.
As far as I know, the Pro version is the same as the original one, just some cosmetic changes. Anyway, I would pick R5 3600 over X99, unless you really need the X99 features.
@@Miyconst I really appreciate your reply despite being busy creating content thanks dude. Stay safe out there. Its not that i don't like to build a Ryzen system... X99 is way cheaper in my country.
@Miyconst Can you please link the Hardware unboxed video you mentioned in the video deep diving into the nvidia driver overhead.
It's already in the video description: ruclips.net/video/JLEIJhunaW8/видео.html
Hi Brown, Will we be able to unlock the turbo boost for the v4 line in the future?
No, most likely it won't be possible.
AMD Radeon software has a cool feature that lets you set target fps range that can really smooth frame times. I’ve been using single slot RX 550 in a SFF pc and this feature lets me run games at smooth 45 fps without big dips and better temps too.
Yes, it works similar to Nvidia's anti-lag.
I've been thinking about getting a tesla P4 for a dedicated video encoder, but haven't made the decision yet because I thought if I ever needed a backup GPU to game on it wouldn't be a valid option, but seeing that there is a way to run it through an IGPU if I needed to makes the idea more palatable. I don't want to spend a whole lot on a dedicated hardware encoder for my server, and now knowing that I could use it as a backup in case my Vega fails, I think it might just be the right choice.
if you have a newer Amd chip or any recent 8 th gen or newer Intel Igpu they have plenty of horsepower for encoding
@@tanmaypanadi1414 I don't believe the Rx 6400 has encoding capabilities though, and am limited not necessarily half height but having no power connections
For your cooling solution, better would be to leave the top cover on, cut a hole in the cover and blank the side inside the case. That way you create a tunnel and cooling will be more efficient. This card could be a good solution for video encoding for plex servers etc.
Yes, an air tunnel is a better cooling solution, but why would you damage the beautiful P4 cover if the card is staying cool either way?
Or you can simply cut any cheap plastic to the size of the cover and create the air tunnel yourself.
I downloaded the 3d printing design for the air tunnel funnel.
I do 3D rendering using cuda cores. the tesla brings 2560 CUDA cores . so i wonder if it would be usefull as a pure render GPU on a budget. nvidia Iray can use more then one GPU for rendering.
Yes, you can use it as a rendering device.
Thanks a lot
Thanks for review! Nice english ))
Could you test the Tesla P4 again when using an EGPU case? I wonder if the input leg would be better...
Will an external GPU case the input lag will be even worse.
Love your shirt man!!!
Hi, do you think the p4 will Play yuzu Switch Emulator well? I try to build a gaming Emulator from a Lenovo sff PC with i5 9400. It have only one Slot for GPU.
Do yourself a favor and to with RX 6400 or Arc A380.
I am late for that but - how you did used 80mm fan? You sure you didn't used 40mm fan?
card is like 50mm on 1 axis only, i used 3d printed bracket for 60mm fan and fan itself is much bigger than the card :D
Will this work in a PciE x4 slot? Thinking of adding a second GPU for video editing/acceleration and since my main is an AMD I have no CUDA stack.
Honestly I don't know, but it should work.
THANKS FOR THE VIDEO!!!!
Did you disable ECC ? If it's enabled it will cause latency issues. Disabling it also allows full use of all 8gb of ram.
No, I had ECC enabled.
Hello,
Can you do a video on tesla P40 gpu. You will find it on ebay around 200 euro. Most likely it will perform somewhere like quadro p6000. It have 24gb vram. Draw back will the no video output.
Thanks, saw the card and wanted to see if the P4 would be good for a prodesk G5.
You have to keep in mind that your prodesk G5 most likely has 180w power supply and pci-e x16 is limited to 35W instead of 75W
did u end up doing it? could you give me an update?
Have you ever tested that Tesla P4 card on system that has fully legacy bios, and no above 4g decoding, and no Resizable bar?
Because I want to buy that card for my homelab workstation Dell R5500(dual LGA 1366 xeon), but it has fully legacy bios, no above 4g decoding and no Resizable bar, and I am wondering if I will be able to boot with my gtx 1660 Ti and that card in other slot, because gtx 1660 Ti working perfectly, but I heard Tesla cards need above 4g decoding, UEFI.
Man awesome gpu, if you want a small form factor gpu for console 45-60fps its absolutely amazing
if anyone is following this in 2024, the most recent driver version that will install in windows is 16.7 (538.78) the 17+ drivers don't seem to be compatible
Thanks for the extra info.
I got my hands on an E5-2666 v3 for free through a friend of mine and I was wondering what motherboard you would go for?
This video has all the answers. m.ruclips.net/video/AW0QgQNbBwA/видео.html
Just one question, these tesla Gpus are compatible with Intel 3rd gen cpus or they only work with 4th and over?
There are many different Tesla GPUs. The P4 should work with the 3 gen as well, but obviously I can't guarantee that.
@@Miyconst thanks for the help, I could not find this type of information anywhere
My understanding is that you need 4th gen or newer to pass through video on the iGPU, but maybe there are other ways on older devices
Very nice shirt, where did you bought it?
Sadly I can't post links in comments, if you want to buy a shirt like that - ping me on Discord.
I saw on another p4 vid, someone said to disable the ecc on the p4 vram, as the ecc on the p4 vram is not needed for gaming, and makes the card have lower clocks, and makes it so you can only use 4gb of the 8gb on the p4.
What would be the benefit of disabling 4 GB of vRAM?
@@Miyconst Im gonna just copy and paste the comment so I dont mess it up. - "He's not talking about ECC for system memory. He's talking about ECC for the onboard memory for the GPU. Gaming cards never have it. So, unless you've used a Tesla card before, you've never used this type of ECC before. ECC was introduced with the Tesla cards because some of them have a lot more memory than their gaing counterparts, but also most importantly because they are targeted at GPU compute applications. If you're a scientist running a complex simulation that takes days, weeks, or even months to run, the last thing you want to do is get erroneous results because at some point during that time a bit got flipped by a cosmic ray. If that happens your entire result could be bogus and you might not even know it. For gaming, it's a different story because frames are rendered in milliseconds or less, which means there's less time for something like that to happen and even if it did, it'd probably just mean the color of one pixel on the screen might be off a bit or if it happened in the z buffer maybe there would be a very slight rendering error on one pixel. In essence, it's highly unlikely you'd ever even notice it. In fact, it'd probably be almost impossible to notice it, but for compute application, and cypto work, it would be a huge deal. Therefore, the Tesla cards have ECC support and they usually have it enabled by default. The downside is the way ECC is done on the cards, it takes away a portion of the memory. It also takes away a bit of memory bandwidth and decreases performance a bit on applications where bandwidth matters more. The way system meory does ECC, it usually has extra bits and those extra bits have extra connections on the bus so that the ECC is done in parallel. That's not the case here, so the ECC bits take up memory and narrow the bandwidth. So, for gaming it makes perfect sense to disable it and it's highly unlikely to affect system stability. Anyway, I believe to toggle it on or off, you must use either the Nvidia SMI or the GPUmodeswitch utility."
If I pass through the graphics through a cheap GPU like the Radeon HD 7570 would it reduce the input latency?
No, most likely it won't.
@@Miyconst thanks for the reply! Love the content btw!
Another great video myconst. ❤️
Is there a performance drop if the Nvidia card/driver can't get a grid license?
The other drivers didn't work with P4 at all, thus I can't compare.
I didn't know about Nvidia Tesla, is it a good choise for my case? my pc just acepts low profles gpu. I thought about Rx6400, Gtx1650 and Quadro T600, but I found a T4 8GB more cheaper thant the others.
Just watch the video, it answers your question.
@@Miyconst Yes. answered to me. T600 ~170 euros, RX6400 ~200 euros and GTX 1650 ~230 euros... P4, used of course, but just 90 euros, we have the winner. Thanks for your video.
Just a question, Can I still continue using my GT 1030 and yours videos output without problems? right?
I am not sure actually, GT 1030 might be incompatible with the P4 drivers.
@@Miyconst ok, I'll research for it, anyway my cpu hás Intel gpu, uhd 630
If rx6400 sells for $99 i might grab one for my itx build
Coincidence i hv i7 4770+h81 itx for $100 combo
Yup, that would be a good package.
@@Miyconst thx for ur great effort for testing on both i7👍
I am not sure if anyone is paying attention but I went to the NVIDIA web site and they have 511.65 ? Is this modified to still not work or did they release it again?
Driver for Tesla or the GTX/RTX driver?
The link to the grid drivers no longer works. Can I use the 511.65 desktop drivers?
No, the desktop drivers won't install with Tesla inside. I have added another link to the video description, please try it.
@@Miyconst Awesome. I contemplated getting one of these, but didn't decide on it until I saw your video on how to set it up. I plan on installing it in a dell 7040.
Goodnight friend. When possible, bring a review of the i9-11900h interposer to us. The component has a good price on Aliexpress.
Are you asking about this one? m.ruclips.net/video/NgXYs4DeAKs/видео.html
I watched this video. I'm thinking of replacing the 2640v3 with this i7.
@@pacatatu7660 if you have to replace entire platform, go with i3-12100 / i5-12400 or AM4.
He looks like he's gonna review eating them like a drumbstick
Thx :)
Would you say performance is similar to a gtx 1060 3GB?
Should be better this is the 1080 chip and you have 8GB of Memory
The performance is a bit better than GTX 1060 6 GB.
Hope this comment reach you. I have an old dell 7010 sff usingg as nas. I wanna add the P4 for some 3d stuff but when i install the card. it slow down the whole system and task manager said that 100% CPU used system interupt. withou the p4 it work just fine.
I try to use adjust in bios such as video Intel. UEFI and so on.
The comment has reached me but I have no clue what to suggest you as the provided description is way too vague. It can be software error or hardware error.
@@Miyconst thanks for your reply. i did figure out. so i did try 2 setup
igpu and P4 - the Igpu is only help me to few resolution - is there any change that i can fix it
gt 710 and p4 - it allow me to choose resolution to my screen but the p4 driver is being uninstall cause i thing the system not allow 2 driver .
Is there any change that i can have the 2k output with any of my setup
Thank you in advance for help.
@@quocvutien9918 do you have Intel iGPU drivers installed? If the resolution is not selectable it means you don't have drivers properly installed
@@Miyconst yes it installed but there are 3 4 option and the maximum is 720p i believe
@@quocvutien9918 then I am out of ideas but it sounds like something is really off.
the pcie2-4 all would be x16 unless you put other pcie cards in the system. if those cards dont limit themselves to x1 than you may lose performance as most boards at this time dont have more than 16 parhways to the cpu and its shared across the slots.
You are amazing
GPU makers would make bank if they released more low-profile/low-power gpus. I don't get why the rx 6600 is pcie4.0 though.
Does It Works on cpu with no onboard graphic? As Ryzen X séries and Intel F séries?
It works, but you need another GPU for video output.
Could use HDMI/dp/vga from motherboard?
@@gabrielsdmonly if your CPU has an iGPU.
Hi, I registered on Discord, I posted my hardware information, in the #nvidia tesla section, still no solution, I followed these steps again, I reinstalled Windows 10, nothing to do, the driver installs, in device manager it appears detected with an exclamation point, it doesn't enable correctly, I tried both the drover version of the video and the latest version indicated in the link, nothing to do.😢
I can't monitor discord 24/7, but I have already answered there.
The link to the driver does not work. Do you have a working link to these?
Maybe there are some geolocation restrictions, but the link works just fine for me.
Can i use grid driver on quadro m1200? Also recommend passive cool t4 or stock 6400?
Tesla T4 is very overpriced and untested at the moment, thus I can't comment on it. Regarding the drivers - give it a try.
@@Miyconst atm it is cheaper than 6400 USD. t4 is 100. 6400 is 140+
@@forgottenxd5588 where do you get a Tesla T4 for $100? I want two!! Are you sure it's T4, and not P4?
Recently purchased a Tesla K80 because the deal seemed too good to be true (63 USD on Ebay)... After days of tinkering i've finally managed to cool it down to stable temps with two crazy loud Xbox 360 S fans and three 80mm server fans on the back of the board because the VRMs heated up too much. With all of that hassle, I haven't noticed an important performance gain in video editing applications over my GTX 1050 Ti. While having 24GBs of VRAM might sound cool, in practice unless you're running some heavy computing task like physics simulation or AI, having such a power-hungry yet not very efficient beast in your system kinda sucks. Now I think about selling the card and getting either a 2060/1070ti or a Tesla P4, as it shares the same chip with the 1050Ti but has the additional 4GB of VRAM that allows the chip to unleash its full performance in video editing and AI processing tasks.
The lesson from this... Do your research before indulging into the cheap server hardware mess.
I totally support the conclusion, you always have to do some research before buying weird (or standard) PC hardware.
K80 is an old Kepler card. Most kepler cards today struggle to run modern games even at low settings due to outdated drivers.
@@suiken3149 I don't really game that often, so didn't notice. But 4k video editing definitely became an easier task for my system.
Is it possible to undervolt the Tesla P4?
No, unfortunately everything is fully locked with Tesla P4.
I think the P4 shines as a cloud gaming card for a vm.
Cloud gaming is an interesting hobby project, nothing more.
MSI Radeon RX 6600 ARMOR 8G would have been a better comparison (and it still really cheap)
I don't understand how you notice the input lag. I use a Tesla M40 for gaming and it passes the image through another gpu and I cant notice anything. By your logic I should have double the input lag because the image travels from the gpu to my cpu and then back to my disp out gpu.
I can't say that the input lag will be double as much, but it's surely significantly bigger. How do I notice it? Well.. in my case it was so noticeable that I can feel it by just moving mouse while observing game's reaction.
@@Miyconst I have experimented with it and for some reason using igpu has extra input lag. My bad. This is kinda stupid, maybe the processor graphics aren't good enough I guess. Also noticed that the performance with no igpu ( using another cheaper GPu as a display out) is way better, sometimes using igpu meant I had to turn down my graphics from high to low because of the performance impact.
Its always nice to see these so called obselete mining GPUS being used for something other than just being e-waste. truly enjoying
Tesla P4 is not a mining card, it's a professional accelerator for video encoding / decoding.
what is the best brand of rx 6400
Anything works, Sapphire or XFX is my pick.
Bro, please test Dual Xeon e5 2699 v4. It's better than m1 mac!
Better in which scenario? What would be the purpose of such testing?
Is the rtx a2000 6gb much better(gaming)
Yes.
@@Miyconst they sell for 150 currently
What if I put them together
Then you get a system with RX 6400 and Tesla P4 😀
R9 350 2GB 45$
The performance is far from RX 6400 / Tesla P4.
Саме та ситуація коли напрягався робив налаштування щоб ні яке за Богор не лайно не було ви фірі знайшов попроси характеристики потрібних цікавих мені відеокарти тут хера все те саме тільки навпаки навпаки зовсім не на тій мові і зовсім нічого не зрозумілааа
so sad rx 6400 only have 4gb ddr5 kek
Yup, with 8 GB it would have been a much more attractive product.
gddr6 for the rx 6400, not gddr5.