@@teflonstestbenchBut bro what I don't understand is how come the newer generation CPUs and GPUs that consume low power than their counterparts run quite hotter than them. For example the ryzen 7800X3D runs 15 - 20 percent hotter than its i9 counterpart
Well it's different in GPUs and CPUs. In GPUs it depends on the cooling of each card, here the 3070 I have has a massive heatsink so it runs very cold, while the rtx 4060 is an entry level gigabyte so it has a crappy heatsink and thus runs hotter. For the cpus it depends on a lot of things but the Ryzen thing you are saying is because the Ryzen 7000 has a really thick IHS so it conducts less heat than other cpus meaning that even with a good cooler it will be a little hotter than other generations.
Loved the benchmarks and the progress bar for the still images. As a constructive critique though, the white text on white background of the test results was really hard to figure out 😄
Thank you so much, and yeah, the end results I usually edit them before adding the stock footage behind and I didn't realized it was white. I usually don't bother about that because all the other text in the video has some black outline to prevent this, will try to fix it for the next one, thanks :)
Ive been thinking of upgrading my 2080 Super but a couple hundred for barley 10+ fps on some games isnt worth it. I get power consumption but people dont build racecars worrying about gas mileage
Thinking of buying a gaming pc with these Specifications: - Intel i7 9700k - Cooler Master AIO Cooler - 16GB HyperX Fury RGB - Asus TUF Z390 - Asus RTX 2080 Super - Cooler Master 650w - Huntkey Closed Case Thoughts?
I mean, surely it is power efficient. But is it worth it? I'd say no, just wait for 5000-series release. The 4000 series nvidia cards only rely on smaller process manufacturing (4nm) and dlss 3. If only they use 5nm or even 6nm, it wouldn't have been cut down this far. The die sizes, bandwidth, cores, bus width, all of those get reduced within 4000 series....truly 1 step forward 2 steps backward, at least for gaming pc consumer, Since they only care about their enterprise market
I feel like the 3070 still has way better value. I've already seen the 3070 matching the same price as the 4060 on ebay (from the UK) and you're also getting more performance. Yeah the 4060 might be more power efficient and have DLSS 3 but the 3070 does better overall on 1080p and 1440p too. I don't really play games that demand high VRAM and seeing the 3070 prices drop alot shows that I can finally make the upgrade.
The thing about the RTX 3070 value is that in the second hand market it can be a lot cheaper than a new 4060. It's true the DLSS frame gen is a selling point but I even think for raw ray tracing the RTX 3070 is better, tho I will test that soon.
4050- i mean 60 looks like the Prius of nvidia gpus rn. Not a whole lot of power, but not whole lot of consumption either. Its just good enough to get the job done. Looks like a 2080 super is still able to leave it in the dust though. Thats not a good thing.
Yup, Nvidia could have done a perfect entry level GPU if they priced it and named it like a 4050. It would be perfect for new entry level builds but at that price tag it's a lot more attractive to go second hand
My 3070 is a gigabyte oc version the power consumption was 260w at 100% gaming usage! I undervolt it to 960mv the power usage went down to 190w and only lost 2fps but the temperature went down from 69c to 61c at max usage , overall 3070 is a very strong graphics card but sadly the vram holding it in some games specifically if you use ray tracing because the vram is not enough the power consumption gonna cut to half and the fps gonna drop so much
Sadly yeah, the VRAM is a huge problem with it, it's so tight and can not even be enough for some games at 1080p with rt and maximum settings, even if the core has enough power for it. Tho for the power consumption, the default is 220w, but on my EVGA I could go up to 270w so as these cards are really power hungry with some oc you can use that headroom. But yeah, probably undervolting is the best option for these cards.
@teflonstestbench it's possible to add more 8gb vram to it but sadly I don't have a professional guy that can do that thing in my city I mean it's going to only cost 30$ so getting 16gb will solve everything from what I saw in youtube some guys did it , Before I did the undervolt I tried overclocking and I reached 2050mhz on core and 7500 on memory But after undervolt it's 1960mhz on core without overclocking the memory , In games there is no difference at all At max 2 fps so I don't mind keeping the card cool with undervolt
@iikon- the adding more ram thingy does sound good, but in practice it means soldering new memory chips and modifying the BIOS, so even if it's only 30$ for the memory chips you need someone to do all that so it might be more expensive, also depending on the board and bios it might not work correctly so it's not a risk free mod to do
@teflonstestbench from what I saw the 3070 don't need a bios modification because the 3070 have a 16gb vram ready something like this I didn't until it correctly but they said when adding the 8gb to it the nvidia normal bios will recognize it as 16gb card already
I mean it depends, I think there was a RTX 3070 16gb version in the works so that's why it works out of the box, but you can't assume that is the case with every single card out there as different vendors might have edited the bios and there are different versions of it too, so I wouldn't just assume it will work out of the box for all, tho maybe it does.
Mod that 8 GB 2080 Super with 16 or 24 GB VRAM and it works like a 4070 Ti Super , sometimes even better or close to 4080 Super , Only if you are playing in 4k Though
interesting that i dont see any comments on the image quality. plane to see that the 4060 has what it takes in the sharp imagine, lighting ,shadows and reflections. cant wait to see the image fidelity of the 5090. Or even apples new GPU when it is available.
My 150w rtx 3070ti mobile is the same GPU as the desktop 3070. Usually holds 1700 to 1740mhz in all games and the vram holds a +1500mhz OC solidly at 70 to 75 degrees. It should be within 1-3fps of the desktop 3070. So the 4060, while efficient, it isn't crazily more efficient as one might suspect based on the comparison here. Of course, if I repaste it with PTM7950 (slightly worse than liquid metal) I could prolly push the clocks even more. Point is, 100 to 150mhz more ain't gonna make much of a difference. Nvidia couldn't even do the bare minimum and give it a 160bit bus with 10gb of vram. At least then it could've matched the 3060ti. And could use ray tracing due to the extra vram.
The thing is that the RTX 4060 is a card with the specs of a low cost but with the price of a mid range one. It should have been an RTX 4050. That's why while it is efficient the performance is bad and gets bottlenecked in a lot of points, while older cards with a llmore solid configuration can be tunned or down clocked or downvolted to be a lot more efficient and still faster. Specially the Ampere ones scale really bad with high clocks so if you run the 3070 with the clocks you are saying or even a bit less I bet you can get to even less than 150w of power while not loosing much performance.
And funnily enough as you mention the ray tracing I tried this cards in my latest video with ray tracing and the RTX 4060 is horrible, at the levels of the RTX 2080 super in ray tracing, that card is literally not designed to do ray tracing and get choked hard. I'm guessing by the memory or the lack of rt cores, as the architecture should be way better.
@@teflonstestbench Well, mine is a laptop with a 150w 3070ti. Shares the same specs as the desktop 3070 as they're the same GPU. Usually laptop gpu's are better binned than desktop GPU's. So for the desktop 3070, something like 1740mhz cap + UV would lead it to use around 150 to 180w or so. Yeah, the 4060 is literally using an AD107 die. The xx7 die is reserved for the xx50 class GPU's. Thats why the 4060 is so poor in performance and specs. Realistically the 4060ti should've been a 192bit 12gb GPU and called the 4060. Likely would've performed ~10% slower than a rtx 3080. And the current 4070 super should've been a 256bit 16gb gpu. Would've likely performed around a desktop 3090 or 3090ti even.
@@teflonstestbench My personal theory is nvidia and AMD were hoping to get gddr7 for lovelace and rx7000. It would've covered for the narrow bus widths of these gpu's. Like how the 1650 got a decent 10% boost over the gddr5 version due to higher bandwidth. That was despite slightly lower clockspeeds. GDDR7 would've also let nvidia and amd use 2 or 3gb vram chips, so 4060 would've likely gotten more vram. Of course, gddr7 won't fix the fact the 4060 is a AD107 die. A die usually reserved for xx50 class GPU's. The issue is the 3060 wasn't exactly that much faster than the 2060. And the 4060 is even worse uplift than the 3060 was. Thats 2 generations in a row the xx60 class hasn't seen much perf uplift. So the rtx 5060 needs to be an above average uplift in perf to compensate. Otherwise we're screwed.
I don't think gddr7 would have been the case there, I think Nvidia moved all the naming up to squeeze more money, as if you calm it a 60 series you can price it like one. Also about gddr7 i think it's far still from being massively produced, not because it may not ready but because of production problems. So I wouldn't be surprised even if we don't see it on the next generation or only on some cards.
I have the 3070 aorus master beautiful card but that thing uses a lot of power, I think I've reached 310W at 111% power limit with some overclocking, managed to get 2025mhz core clock speed and 8500mhz memory speed with temps maxing at 75 I think, default it reaches a good 280-290W but it's an absolute beast, I had the choice between a 2080ti and the 3070, I would've picked the 2080ti because it has more ram but it didn't look as good and was apparently more power hungry than the 3070.
I haven't had a 2080 ti yet, but the ampere architecture seems to be a lot more power hungry than the previous RTX 2000 one, so yeah, once you go past the sweet spot in core clock going further uses a ton of power, specially on the high end of the series that have big chips. I haven't messed around overclocking it much but it seems the best call is to undervolt the card as it seems the performance is really close if not the same and you can almost halve the power consumption as 310w for that is brutal xD
@teflonstestbench I have to agree, there was an improvement in performance and it wasn't that bad but the power draw is insane, I should've undervolted but rn I decided to just lower the power limit to 60% because I got a new cpu and I want to play it safe until I get a new psu, even tho it's at 60% power limit, it barely lost any performance and now draws between 190 to 210w which is impressive
I don't think it's a good idea to lower the power that much while having an overclock, maybe you should under lock and undervolt by curve and that might give you more stability and better performance with 60% the power. Tho that being said the regular tdp is 220w so you now basically have something really close to the regular 3070 funnily enough xD
@teflonstestbench already removed the oc when I set it to 60% to avoid those issues, it runs really well and the performance loss was like 5 fps from 65 to 60 in the benchmarks I ran with no stability issues, only issue I've seen is nvidia doesn't want to monitor the card anymore so it won't show me fps or other stats from it which is kinda weird but doesn't matter, will definitely look into an undervolt soon
Is the amount of power the card is using measured externally, the other power value is the power consumption the card reports, but in some cases it is not very accurate.
2 reasons: The cyberpunk preset activates FSR by default, even in Nvidia cards, so to not change the preset I just use what the game uses. Also all my other tests on other videos use FSR on that game for that same reason, so to make all the results comparable I just use the same settings for all the cards, even if it doesn't make the most sense on some cases.
@@teflonstestbenchyou are right but getting the 4060 for brand new and having no risk associated with if the card works and if it doesn’t you at least have a support network willing to help you get a new one or refund
If you mean the RTX 3000 series then yeah, some good products in performance and price but miners ruined that at the time and the VRAM of some cards was a joke
4060 is using half the power of a 3070
Yeah, ampere is a really power hungry generation, specially at higher clocks it gets wild
@@teflonstestbenchBut bro what I don't understand is how come the newer generation CPUs and GPUs that consume low power than their counterparts run quite hotter than them. For example the ryzen 7800X3D runs 15 - 20 percent hotter than its i9 counterpart
Well it's different in GPUs and CPUs. In GPUs it depends on the cooling of each card, here the 3070 I have has a massive heatsink so it runs very cold, while the rtx 4060 is an entry level gigabyte so it has a crappy heatsink and thus runs hotter.
For the cpus it depends on a lot of things but the Ryzen thing you are saying is because the Ryzen 7000 has a really thick IHS so it conducts less heat than other cpus meaning that even with a good cooler it will be a little hotter than other generations.
it's power hungry because of crappy Samsung Foundry Node insteat TSMC @@teflonstestbench
Mine broke 2 days ago now I am getting a 2080 super 😭
In newer games rtx 4060 = 2080 Super
In older games Rtx 4060 = 2080.
Because newer games generally run better on newer architecture.
Loved the benchmarks and the progress bar for the still images.
As a constructive critique though, the white text on white background of the test results was really hard to figure out 😄
Thank you so much, and yeah, the end results I usually edit them before adding the stock footage behind and I didn't realized it was white. I usually don't bother about that because all the other text in the video has some black outline to prevent this, will try to fix it for the next one, thanks :)
Good to see my old 2080super still doing well i undervolted my card runs at 180w now lost 1 fps on average and runs way cooler 👍
Ive been thinking of upgrading my 2080 Super but a couple hundred for barley 10+ fps on some games isnt worth it. I get power consumption but people dont build racecars worrying about gas mileage
Thinking of buying a gaming pc with these
Specifications:
- Intel i7 9700k
- Cooler Master AIO Cooler
- 16GB HyperX Fury RGB
- Asus TUF Z390
- Asus RTX 2080 Super
- Cooler Master 650w
- Huntkey Closed Case
Thoughts?
Sounds good, but depending on the price you are paying for it maybe there are better options or you could save on some parts.
I mean, surely it is power efficient. But is it worth it? I'd say no, just wait for 5000-series release.
The 4000 series nvidia cards only rely on smaller process manufacturing (4nm) and dlss 3. If only they use 5nm or even 6nm, it wouldn't have been cut down this far. The die sizes, bandwidth, cores, bus width, all of those get reduced within 4000 series....truly 1 step forward 2 steps backward, at least for gaming pc consumer, Since they only care about their enterprise market
what I like about your tests is the comprehensive information
Thank you :) I try to make them useful
I feel like the 3070 still has way better value. I've already seen the 3070 matching the same price as the 4060 on ebay (from the UK) and you're also getting more performance. Yeah the 4060 might be more power efficient and have DLSS 3 but the 3070 does better overall on 1080p and 1440p too. I don't really play games that demand high VRAM and seeing the 3070 prices drop alot shows that I can finally make the upgrade.
The thing about the RTX 3070 value is that in the second hand market it can be a lot cheaper than a new 4060. It's true the DLSS frame gen is a selling point but I even think for raw ray tracing the RTX 3070 is better, tho I will test that soon.
considering the 4060 is a 4050 in everything but name this is a good test.
Yup, Nvidia pushed the naming up on this whole generation, so that's why we got a lot of super and ti versions with the RTX 4070 as a result xD
a true 4060 shouldve been like the 4060 ti
Metro Exodus is so well optimized it even runs well on ultra setting with ray tracing + DLSS on my RTX 2060 laptop
Me hace bien saber que mi rtx 2080 súper aún no está obsoleta
4050- i mean 60 looks like the Prius of nvidia gpus rn.
Not a whole lot of power, but not whole lot of consumption either. Its just good enough to get the job done.
Looks like a 2080 super is still able to leave it in the dust though. Thats not a good thing.
Yup, Nvidia could have done a perfect entry level GPU if they priced it and named it like a 4050. It would be perfect for new entry level builds but at that price tag it's a lot more attractive to go second hand
My 3070 is a gigabyte oc version the power consumption was 260w at 100% gaming usage! I undervolt it to 960mv the power usage went down to 190w and only lost 2fps but the temperature went down from 69c to 61c at max usage , overall 3070 is a very strong graphics card but sadly the vram holding it in some games specifically if you use ray tracing because the vram is not enough the power consumption gonna cut to half and the fps gonna drop so much
Sadly yeah, the VRAM is a huge problem with it, it's so tight and can not even be enough for some games at 1080p with rt and maximum settings, even if the core has enough power for it. Tho for the power consumption, the default is 220w, but on my EVGA I could go up to 270w so as these cards are really power hungry with some oc you can use that headroom. But yeah, probably undervolting is the best option for these cards.
@teflonstestbench it's possible to add more 8gb vram to it but sadly I don't have a professional guy that can do that thing in my city I mean it's going to only cost 30$ so getting 16gb will solve everything from what I saw in youtube some guys did it ,
Before I did the undervolt I tried overclocking and I reached 2050mhz on core and 7500 on memory
But after undervolt it's 1960mhz on core without overclocking the memory ,
In games there is no difference at all
At max 2 fps so I don't mind keeping the card cool with undervolt
@iikon- the adding more ram thingy does sound good, but in practice it means soldering new memory chips and modifying the BIOS, so even if it's only 30$ for the memory chips you need someone to do all that so it might be more expensive, also depending on the board and bios it might not work correctly so it's not a risk free mod to do
@teflonstestbench from what I saw the 3070 don't need a bios modification because the 3070 have a 16gb vram ready something like this I didn't until it correctly but they said when adding the 8gb to it the nvidia normal bios will recognize it as 16gb card already
I mean it depends, I think there was a RTX 3070 16gb version in the works so that's why it works out of the box, but you can't assume that is the case with every single card out there as different vendors might have edited the bios and there are different versions of it too, so I wouldn't just assume it will work out of the box for all, tho maybe it does.
I just went from a RTX 2060 to a RTX 4060. I have only had it a couple of days. Hope I did good with this. Any thoughts about this upgrade?
It's a big upgrade, you should see a big jump in performance so it's really nice. Hope you enjoy it :)
@@teflonstestbench Thank you
4060 efficiency were insane
I mean.. At least the 4060 doesn't draw that much power...
Mod that 8 GB 2080 Super with 16 or 24 GB VRAM and it works like a 4070 Ti Super , sometimes even better or close to 4080 Super , Only if you are playing in 4k Though
interesting that i dont see any comments on the image quality. plane to see that the 4060 has what it takes in the sharp imagine, lighting ,shadows and reflections. cant wait to see the image fidelity of the 5090. Or even apples new GPU when it is available.
3070TI... shouldve been used these are all ad104 but 3070 is cut down
Can you add For Honor to the test list too? A game that uses cpu more just to see the performance of cpu intensive games. Thanks!
I'll look into it, it could be interesting for CPU testing, thanks for the recommendation :)
My 150w rtx 3070ti mobile is the same GPU as the desktop 3070. Usually holds 1700 to 1740mhz in all games and the vram holds a +1500mhz OC solidly at 70 to 75 degrees. It should be within 1-3fps of the desktop 3070. So the 4060, while efficient, it isn't crazily more efficient as one might suspect based on the comparison here. Of course, if I repaste it with PTM7950 (slightly worse than liquid metal) I could prolly push the clocks even more. Point is, 100 to 150mhz more ain't gonna make much of a difference.
Nvidia couldn't even do the bare minimum and give it a 160bit bus with 10gb of vram. At least then it could've matched the 3060ti. And could use ray tracing due to the extra vram.
The thing is that the RTX 4060 is a card with the specs of a low cost but with the price of a mid range one. It should have been an RTX 4050. That's why while it is efficient the performance is bad and gets bottlenecked in a lot of points, while older cards with a llmore solid configuration can be tunned or down clocked or downvolted to be a lot more efficient and still faster. Specially the Ampere ones scale really bad with high clocks so if you run the 3070 with the clocks you are saying or even a bit less I bet you can get to even less than 150w of power while not loosing much performance.
And funnily enough as you mention the ray tracing I tried this cards in my latest video with ray tracing and the RTX 4060 is horrible, at the levels of the RTX 2080 super in ray tracing, that card is literally not designed to do ray tracing and get choked hard. I'm guessing by the memory or the lack of rt cores, as the architecture should be way better.
@@teflonstestbench Well, mine is a laptop with a 150w 3070ti. Shares the same specs as the desktop 3070 as they're the same GPU. Usually laptop gpu's are better binned than desktop GPU's. So for the desktop 3070, something like 1740mhz cap + UV would lead it to use around 150 to 180w or so.
Yeah, the 4060 is literally using an AD107 die. The xx7 die is reserved for the xx50 class GPU's. Thats why the 4060 is so poor in performance and specs. Realistically the 4060ti should've been a 192bit 12gb GPU and called the 4060. Likely would've performed ~10% slower than a rtx 3080. And the current 4070 super should've been a 256bit 16gb gpu. Would've likely performed around a desktop 3090 or 3090ti even.
@@teflonstestbench My personal theory is nvidia and AMD were hoping to get gddr7 for lovelace and rx7000. It would've covered for the narrow bus widths of these gpu's. Like how the 1650 got a decent 10% boost over the gddr5 version due to higher bandwidth. That was despite slightly lower clockspeeds. GDDR7 would've also let nvidia and amd use 2 or 3gb vram chips, so 4060 would've likely gotten more vram.
Of course, gddr7 won't fix the fact the 4060 is a AD107 die. A die usually reserved for xx50 class GPU's. The issue is the 3060 wasn't exactly that much faster than the 2060. And the 4060 is even worse uplift than the 3060 was. Thats 2 generations in a row the xx60 class hasn't seen much perf uplift. So the rtx 5060 needs to be an above average uplift in perf to compensate. Otherwise we're screwed.
I don't think gddr7 would have been the case there, I think Nvidia moved all the naming up to squeeze more money, as if you calm it a 60 series you can price it like one. Also about gddr7 i think it's far still from being massively produced, not because it may not ready but because of production problems. So I wouldn't be surprised even if we don't see it on the next generation or only on some cards.
I have the 3070 aorus master beautiful card but that thing uses a lot of power, I think I've reached 310W at 111% power limit with some overclocking, managed to get 2025mhz core clock speed and 8500mhz memory speed with temps maxing at 75 I think, default it reaches a good 280-290W but it's an absolute beast, I had the choice between a 2080ti and the 3070, I would've picked the 2080ti because it has more ram but it didn't look as good and was apparently more power hungry than the 3070.
I haven't had a 2080 ti yet, but the ampere architecture seems to be a lot more power hungry than the previous RTX 2000 one, so yeah, once you go past the sweet spot in core clock going further uses a ton of power, specially on the high end of the series that have big chips. I haven't messed around overclocking it much but it seems the best call is to undervolt the card as it seems the performance is really close if not the same and you can almost halve the power consumption as 310w for that is brutal xD
@teflonstestbench I have to agree, there was an improvement in performance and it wasn't that bad but the power draw is insane, I should've undervolted but rn I decided to just lower the power limit to 60% because I got a new cpu and I want to play it safe until I get a new psu, even tho it's at 60% power limit, it barely lost any performance and now draws between 190 to 210w which is impressive
I don't think it's a good idea to lower the power that much while having an overclock, maybe you should under lock and undervolt by curve and that might give you more stability and better performance with 60% the power. Tho that being said the regular tdp is 220w so you now basically have something really close to the regular 3070 funnily enough xD
@teflonstestbench already removed the oc when I set it to 60% to avoid those issues, it runs really well and the performance loss was like 5 fps from 65 to 60 in the benchmarks I ran with no stability issues, only issue I've seen is nvidia doesn't want to monitor the card anymore so it won't show me fps or other stats from it which is kinda weird but doesn't matter, will definitely look into an undervolt soon
3070 beast
should have used a 3060ti it is just slightly faster than 2080s
Will do next, I have one around and I also want to test these cards against the RX 6700 xt too
what does real gpu power usage mean?
power draw in watts(amount of electricity the part uses)
Is the amount of power the card is using measured externally, the other power value is the power consumption the card reports, but in some cases it is not very accurate.
Very nice white letters on a white backdrop.
4060 with dlc support or not ?
What do you mean?
why on earth are you using FSR on three NVIDIA gpus in games that support DLSS 😂
2 reasons:
The cyberpunk preset activates FSR by default, even in Nvidia cards, so to not change the preset I just use what the game uses.
Also all my other tests on other videos use FSR on that game for that same reason, so to make all the results comparable I just use the same settings for all the cards, even if it doesn't make the most sense on some cases.
3070 great 👍
I think 4060 is best (watt performance)
In terms of efficiency it is, but the RTX 3070 is faster
@@teflonstestbenchwho care about 20 fps difference 😏
It's not only that, but you can get the rtx 3070 for cheaper in the second hand market
@@teflonstestbenchmining:
@@teflonstestbenchyou are right but getting the 4060 for brand new and having no risk associated with if the card works and if it doesn’t you at least have a support network willing to help you get a new one or refund
How 4060 better 2080 super?🤔🤔🤔 All okay?
What do you mean?
Nah. This is fake. RTX 2080 better than 4060ti in 4k. 4060 is shit only for 1080p. It has only 128 bus bit. Trash.
4060 is scam..............
Huge cpu bottleneck
A low CPU percentage is the opposite of a bottleneck.
To be honest the 2080 was not a good buy when it came out
You mean the RTX 2080 non super? Or the super?
@@teflonstestbench yes
@@teflonstestbench After the generation before this one, this generation was very disappointing
If you mean the RTX 3000 series then yeah, some good products in performance and price but miners ruined that at the time and the VRAM of some cards was a joke
@@teflonstestbench I meant the gtx 1000 series which were very nice for their price, unlike the rtx 2000 series
Wow the 4060 is mediocre, it doesn't even surpass a 2080 2 generations ago 🫨