The other area, where higher memory bandwidth can make a significant difference, is VRAM size limitation. When the game fills all the GPU memory, it may start to spill the data to the system RAM, this process usually causes stutters and sometimes even freezes. If you take two GPUs from the same company with the same VRAM size, then the one which has higher memory bandwidth will create less significant issues, as it can swap data in and out of it's memory faster. You may saw comparisons of 3070 and 3070 Ti, for example. And sometimes, when a game uses all it's 8 GB buffer, Ti version provides smoother experience. This is because of it's higher VRAM bandwidth. At the same time, it is not correct to compare two GPUs from the different companies, even though they may have the same VRAM size. The reason for this is that Nvidia GPUs possess more effective algorithms of data compression, thus they use less amount of VRAM than AMD GPUs at the same conditions. Although the difference is not that big, it is significant enough. This is one of the reasons why in recent years AMD usually release GPUs with more VRAM than Nvidia at the same performance tier.
Good result. Ada is very power efficient architecture, so there is no need to increase voltage. What about power consumption? Has it increased? And have you increased power limit above 100%?
my 4070 nn ti inno3dx3 push +1800=12300 = 24600mhz .i see a video 4070 ti inno3dx3 push +2300 =12800= 25600mhz (effective) with tunedit (inno3d program) because afterburner can`t do over 2k overclock
My rtx 4070 super inno3d twin oc pushes +1800 mhz on vram, reaching 12300mhz, sadly it's power limited so i have to undervolt it to 0.950 mV at 2715 core clock
Nevertheless, good result. Though I think even 975 mV at ~2800 MHz would be enough to fit in 220W. At least my 4070 Ti can do it in most cases. 4070 Super is indeed power limited, unlike 4070 Ti. More expensive versions have extended power limits, but they cost almost as much as the Ti. Nvidia is a very calculating company.
There is no need in this for Nvidia GPUs, especially 4000 series, which are very power efficient cards and never reach their power limits in games. But for AMD cards you'd better increase power limit, otherwise memory overclock could cause performance decrease in some cases.
Yep. At 1440p/Ultra it is perfectly fine. In some cases you will have to use DLSS to get 60+ fps experience (Cyberpunk, Dying Light 2). But even at native 1440p games usually consume about 8-10GB at this res (including RE4 Remake and infamous TLoU P1). Problems can happen in some games at native 4K maxxed out, but you won't be able to run AAA modern games at 60+ fps with this GPU anyway. And with DLSS/FSR even 4K gaming is possible. So I do not think it is relevant, unless you are keen to play at 4K/30fps. The only real issue for the users of 1440p screens, in my opinion, might be frame generation usability in future games. FG consumes additional VRAM (usually it is around 1GB). If you will want to play future game maxxed out with enabled FG, 12GB of VRAM might become a deal breaker. But this is just speculation and will depend on how well optimized future games are.
@@okiqs My congratulations! I would say, it is worth overclocking VRAM on majority of GPUs. You can gain slight performance boost without any significant increase in power draw or temperature. I always use MSI Afterburner for this purpose, so can not comment on ASUS software. I'd recommend to perform undervolt as well. I am using the folowing settings: 915 mV, 2595 MHz. This way my own sample runs just 3-5% slower than stock, but memory overclock compensates it with a margin. The good thing about undervolt - temperature decrease is quite significant (-10°C) and power draw reduced down to 170-180W.
It depends on what you are seeking. If you are OK with it's power consumption and temperatures, then it definitely worth to overclock this GPU. It will bring about ~10% performance improvement without any meaningful increase in power draw or temperatures. First try to detect the highest stable VRAM frequency for your sample, then overclock a core. Usually the highest possible core frequency can be achieved at reasonable voltage value (not the maximum). If you want to reduce power consumption and temperatures of a GPU, then you'd better overclock VRAM at first, then try to choose the optimal ratio of core frequency and core voltage. For 4070 Ti the most optimal core voltage is in the range of 900 mV to 950 mV. Therefore you could either preserve the same level of performance and achieve much lower power consumption or increase GPU performance a little bit (by 2-5%) and significantly reduce power draw. Anyway, I'd definitely recommend to fine tune your card, since stock GPU settings are usually less optimal than custom ones.
@@starstreamgamer3704why dont just use dlss, its good especially with higher res, I am waiting for my 120hz TV pairing with 4070ti which handles everything fine even in native 4k 60hz. BTW do you have coil whine on your gpu?
@@Tpecep I always use DLSS (native or modded) at 1440p or 4K. It is an amazing feature. But still I wouldn't mind 5-10% performance gain for free. Coil whine is a problem for most modern GPUs. And my 4070 Ti is not an exception. But undervolt helped me to reduce it to the minimum level (almost inaudible). Though some people say that replacing power supply may help in this regard as well.
I will make a similar comparison at 4K. I bet the difference will increase to 7-8%. If you're lucky and won a silicon lottery, you will be able to overclock VRAM to 25000 MHz (600 GB/s). Then you will get 8-9% more fps at 4K.
You can't make up for not having enough ram, when your out, your out. An oc might get you a few fps here and there regardless of being out or not. Also the left side with no OC your not over the vram limit, show both over the limit has one OC and one not. Then also show both under the limit and one OC and one not.
Watch some 3070 and 3070 Ti comparisons. For example, Daniel Owen made a great one. In his video he shows how Ti can provide much higher 1% low fps (less stutters) in certain circumstances due to higher memory bandwidth, which helps Ti to move data in and out of it's memory more quickly, thus Ti uses it's 8 GB buffer more effectively then vanilla 3070.
@@starstreamgamer3704 yes that's true, but having higher mem bandwidth in general will almost always help whether your or out of vram or not, but it's not going to save a card from the effects of running out of vram and dipping into system ram. People have used this excuse for Nvidia in the past, their cards using faster ram will make up for not having a lot of vram and that's nonsense. Yea maybe you will get a few less stutters, but its not optimal regardless.
You are very wrong. Faster memory makes easier to load and unload assets which can make difference And you can see he bas no vram bottlenecks in most of cases because there are plenty of other things in pipeline that makes card bottleneck, not only gpu overall and memory load
@@Tpecep it doesn't matter how fast it's doing it, it's still goin to stutter or load things in late, whether the stutter is smaller or the loads it in less late, still doesn't matter, you don't want to play a game that way. I am very right. These days it's even effecting ray tracing performance, this is why amd cards are being Nvidia cards with ray tracing on, when the Nvidia card runs out of vram. So yes you need enough vram full stop, faster ram isn't goin to solve that issue.
My congratulations. Great result. I have not seen any better settings config. Just make sure it is stable. "Control" and "Alan Wake 2" are very VRAM sensitive games - the first one gives artifacts (black screen) if memory is not stable, the second one just crushes. To test the stability of cores, just launch any modern game at native 4K with max settings. If it does not crash in 10-20 minutes, the config is stable.
@@starstreamgamer3704 I've ran 3D Mark Speed Way stress test for 20 minutes and passed with 99.7% stability and various games such as RDR2 with no issues I'll give Alan Wake 2 a shot but I believe this is stable :) I've only tested at 1440p as I don't own a 4k monitor though. oh and if you want to know which model I have I have the ASUS TUF 4070 Ti non OC version. Update - I ran Alan Wake 2 for 15 minutes and haven't noticed any artifacts or stability issues!
@@RavenaLux Good for you. Enjoy your gaming experience! Even if you do not have a 4K screen, you can still take advantage of higher resolution. Enable (DL)DSR 2.25x (through Nvidia Control Panel, 3D settings tab). This will let you GPU render games at 4K (2160p) and downscale it to the native res of your screen. This feature vastly improves image quality. You can usually change rendering resolution in game settings. And even big modern games can still take advantage of it - just use DLSS. 4K + DLSS Quality looks much better than native 1440p. Even 4K + DLSS Balanced still has an edge over 1440p. While 4K + DLSS Performance looks very similar to 1440p, but performs quite a bit better. Plus, there is also 1.78x factor, which is in between 4K and 1440p. Do not forget about "DSR smoothness" parameter - for (DL)DSR the most optimal value, in my experience, is ~65%.
@@RavenaLux Good for you. Enjoy your gaming experience! Even if you do not have a 4K screen, you can still take advantage of higher resolution. Enable (DL)DSR 2.25x (through Nvidia Control Panel, 3D settings tab). This will let you GPU render games at 4K (2160p) and downscale it to the native res of your screen. This feature vastly improves image quality. You can usually change rendering resolution in game settings. And even big modern games can still take advantage of it - just use DLSS. 4K + DLSS Quality looks much better than native 1440p. Even 4K + DLSS Balanced still has an edge over 1440p. While 4K + DLSS Performance looks very similar to 1440p, but performs quite a bit better. Plus, there is also 1.78x factor, which is in between 4K and 1440p. Do not forget about "DSR smoothness" parameter - for (DL)DSR the most optimal value, in my experience, is ~65%.
@@starstreamgamer3704 Thanks for the tip! but since it's stable at 1440p which is what I game on only I'll just leave it at that, I'm quite happy with this 😁
After you test it is easy to calculate that 4070 "non Ti" will be around 4-5% faster than 3080 10GB or almost exactly the same as 3080 12GB. For 600$ it looks like fair deal.
Maybe. We will know it soon enough. The relative performance level will probably depend on resolution. 3080/12 has enormous memory bandwidth (900+ GB/s) and in some games at 4K at almost matches 4070 Ti. But at 1440p it will probably be close to 4070. There are rumors that there will be two versions of 4070. One with 8-pin power connector and 200W TDP and the other one with 12VHPWR connector and enhanced TDP. I bet the "enhanced" version will cost more than $600 (excluding maybe just "Founders edition").
I think so, I paid £585 for a 2070 3 years ago. I have to say that I'm a little disappointed with the 40×× series, and I think the 4070, 4070Ti & 4080 could all do with a plus 30 FPS performance bump.
@@oktc68 I agree. The only card which justifies it's price in comparison with the previous generation is 4090. 4080, 4070 Ti and 4070 should either be more powerful or less expensive. +10-15% more performance (and more VRAM - 20GB for 4080 and 16GB for 4070/4070 Ti) or it should have been $999 for 4080, $699 for 4070 Ti and $499 for 4070.
I got a 4070 to replace a 3070 It’s fantastic at 1080p and 1440p…. At 4k it’s ok. As stated above, on average it’s the same or slightly worse than a 3080 in some games, in other games the 4070 can beat a 3090 even at 4k, just depends on the game. Power hungry games and console conversions seem to fair worse at higher resolutions but unless you compare side by side you’d probabley never know that the 4070 was slightly inferior to a 3080 at 4k. The reason I got it was because it *uses half the power* of my OC Aorus 3070 which could easily pull over 250W and did most of the time. The 4070 hovers around 110-150w only hitting 200w in the most demanding games and stress tests. The 4070 undervolted like a champ! This means my room is cooler, PC quieter and I subsequently play more games and even got a gaming mouse 😀
I wanted to make such a comparison, but this card is not fast enough to run modern games at native 4K/Max. And if you use DLSS set to "Quality", this means games are rendered at 1440p and then upscaled to 4K via tensor cores. So the results will be the same as shown in this video. For older games it might be more useful though. I tested one game at 4K (Horizon Zero Dawn) and found out that VRAM OC brings 8% more frames at native 4K. But since there is no DLAA in this game and modded DLAA (via DLSS Tweaks) performs significantly worse than TAA (and do not achieve locked 60 fps), I do not see any sense in not using DLSS Quality. So testing HZD at native 4K is kinda irrelevant.
They do. But only 4090 can truly take advantage of such a high memory bandwidth at native 4K res. Other cards simply can not run modern AAA titles at native 4K anyway, even 4080 struggles a bit. And at native 1440p res (or 4K with DLSS) there is not much use in 700+ GB/s memory bandwidth anyway. So 3080 12GB/Ti and 3090 (Ti) do not really benefit much from (almost) doubled bandwidth. Only in some older titles 3090 Ti can provide 4K/60+fps and outperform 4070 Ti. While in newer games at 1440p/60+fps usually 4070 Ti is roughly on par with 3090 Ti, despite having twice as narrow memory bus.
@@starstreamgamer3704I found the 4070 Ti to be a great match for 3440x1440. My memory can comfortably hit 600 GBps and +180 core giving a noticeable boost in most games. 4k is a big jump from 3440 in gpu requirements, I basically get 4090 4K frames at 3440 which is shocking considering the 4090 costed 2.3x my card (more now due to China 4090 ban)
@@starstreamgamer3704 try cp2077 path tracing MAX GRAPHICS on 4070ti 4k balanced + NVFG ON vs 3090ti path tracing max FSR3FG MOD ON and see whats happen.. poor guy 500gb/s. The real 500gb/s issues only you see with high bandwidth demand.. rt/pt/dlss/fg on.. see the "magic" NVFG ON 20fps vs NVFG OFF 30fps.. 500GB/S IS A 3060TI DEMAND. But see, nvidia abort 4070ti and release 4070tiS 688GB/S LMAO
Good for you. Looks like you won the silicon lottery in terms of VRAM overclock. You might even got better result if Afterburner would not have any limits. What is the performance improvement percentage-wise in your case?
It is! Look at it's tech specifications on any website (TechPowerUp, for example). Memory config: Memory Size - 12 GB Memory Type - GDDR6X Memory Bus - 192 bit Bandwidth - 504.2 GB/s
Why not? There is no significant increase in either power consumption or temperatures or fans speed. So overclocking memory (VRAM) is literally free performance for very low effort. Overclocking GPU is a different story though and depends a lot on the type and size of a GPU cooler, room temperatures, PC case quality, etc. I'd rather undevoted a GPU and overclocked memory. It is a perfect combination of performance/temperatures/noise ratio.
Super test, thanks!
The other area, where higher memory bandwidth can make a significant difference, is VRAM size limitation. When the game fills all the GPU memory, it may start to spill the data to the system RAM, this process usually causes stutters and sometimes even freezes. If you take two GPUs from the same company with the same VRAM size, then the one which has higher memory bandwidth will create less significant issues, as it can swap data in and out of it's memory faster.
You may saw comparisons of 3070 and 3070 Ti, for example. And sometimes, when a game uses all it's 8 GB buffer, Ti version provides smoother experience. This is because of it's higher VRAM bandwidth.
At the same time, it is not correct to compare two GPUs from the different companies, even though they may have the same VRAM size. The reason for this is that Nvidia GPUs possess more effective algorithms of data compression, thus they use less amount of VRAM than AMD GPUs at the same conditions. Although the difference is not that big, it is significant enough. This is one of the reasons why in recent years AMD usually release GPUs with more VRAM than Nvidia at the same performance tier.
I wanted to know how much FPS boost would VRAM overclocking give and got the answer!
Thanks for the video!
I am glad I could help.
I overclocked the memory to 23,400MHz and the core to 3,075MHz without adjusting the voltage. In some games, performance can be improved by 10%。
Good result. Ada is very power efficient architecture, so there is no need to increase voltage.
What about power consumption? Has it increased? And have you increased power limit above 100%?
my 4070 nn ti inno3dx3 push +1800=12300 = 24600mhz .i see a video 4070 ti inno3dx3 push +2300 =12800= 25600mhz (effective) with tunedit (inno3d program) because afterburner can`t do over 2k overclock
My rtx 4070 super inno3d twin oc pushes +1800 mhz on vram, reaching 12300mhz, sadly it's power limited so i have to undervolt it to 0.950 mV at 2715 core clock
Nevertheless, good result. Though I think even 975 mV at ~2800 MHz would be enough to fit in 220W. At least my 4070 Ti can do it in most cases.
4070 Super is indeed power limited, unlike 4070 Ti. More expensive versions have extended power limits, but they cost almost as much as the Ti. Nvidia is a very calculating company.
do you need to adjust the power limit for vram overclocking?
There is no need in this for Nvidia GPUs, especially 4000 series, which are very power efficient cards and never reach their power limits in games.
But for AMD cards you'd better increase power limit, otherwise memory overclock could cause performance decrease in some cases.
is 12gb enough for 4070 ti at 1440p with RT ?
Yep. At 1440p/Ultra it is perfectly fine. In some cases you will have to use DLSS to get 60+ fps experience (Cyberpunk, Dying Light 2). But even at native 1440p games usually consume about 8-10GB at this res (including RE4 Remake and infamous TLoU P1).
Problems can happen in some games at native 4K maxxed out, but you won't be able to run AAA modern games at 60+ fps with this GPU anyway. And with DLSS/FSR even 4K gaming is possible. So I do not think it is relevant, unless you are keen to play at 4K/30fps.
The only real issue for the users of 1440p screens, in my opinion, might be frame generation usability in future games. FG consumes additional VRAM (usually it is around 1GB). If you will want to play future game maxxed out with enabled FG, 12GB of VRAM might become a deal breaker. But this is just speculation and will depend on how well optimized future games are.
@@starstreamgamer3704 thx, i got Asus 4070 Ti TUF OC today. Do you recommend overclocking VRAM with Asus gpu tweak 3?
@@okiqs My congratulations!
I would say, it is worth overclocking VRAM on majority of GPUs. You can gain slight performance boost without any significant increase in power draw or temperature.
I always use MSI Afterburner for this purpose, so can not comment on ASUS software.
I'd recommend to perform undervolt as well. I am using the folowing settings: 915 mV, 2595 MHz. This way my own sample runs just 3-5% slower than stock, but memory overclock compensates it with a margin. The good thing about undervolt - temperature decrease is quite significant (-10°C) and power draw reduced down to 170-180W.
Recomiendas hacer overclock a la 4070 ti de stock ???
It depends on what you are seeking. If you are OK with it's power consumption and temperatures, then it definitely worth to overclock this GPU. It will bring about ~10% performance improvement without any meaningful increase in power draw or temperatures. First try to detect the highest stable VRAM frequency for your sample, then overclock a core. Usually the highest possible core frequency can be achieved at reasonable voltage value (not the maximum).
If you want to reduce power consumption and temperatures of a GPU, then you'd better overclock VRAM at first, then try to choose the optimal ratio of core frequency and core voltage. For 4070 Ti the most optimal core voltage is in the range of 900 mV to 950 mV. Therefore you could either preserve the same level of performance and achieve much lower power consumption or increase GPU performance a little bit (by 2-5%) and significantly reduce power draw.
Anyway, I'd definitely recommend to fine tune your card, since stock GPU settings are usually less optimal than custom ones.
@@starstreamgamer3704why dont just use dlss, its good especially with higher res, I am waiting for my 120hz TV pairing with 4070ti which handles everything fine even in native 4k 60hz. BTW do you have coil whine on your gpu?
@@Tpecep I always use DLSS (native or modded) at 1440p or 4K. It is an amazing feature. But still I wouldn't mind 5-10% performance gain for free.
Coil whine is a problem for most modern GPUs. And my 4070 Ti is not an exception. But undervolt helped me to reduce it to the minimum level (almost inaudible). Though some people say that replacing power supply may help in this regard as well.
OH MY GOD.. show it in 4k!
I will make a similar comparison at 4K. I bet the difference will increase to 7-8%.
If you're lucky and won a silicon lottery, you will be able to overclock VRAM to 25000 MHz (600 GB/s). Then you will get 8-9% more fps at 4K.
You can't make up for not having enough ram, when your out, your out. An oc might get you a few fps here and there regardless of being out or not. Also the left side with no OC your not over the vram limit, show both over the limit has one OC and one not. Then also show both under the limit and one OC and one not.
Watch some 3070 and 3070 Ti comparisons. For example, Daniel Owen made a great one. In his video he shows how Ti can provide much higher 1% low fps (less stutters) in certain circumstances due to higher memory bandwidth, which helps Ti to move data in and out of it's memory more quickly, thus Ti uses it's 8 GB buffer more effectively then vanilla 3070.
@@starstreamgamer3704 yes that's true, but having higher mem bandwidth in general will almost always help whether your or out of vram or not, but it's not going to save a card from the effects of running out of vram and dipping into system ram. People have used this excuse for Nvidia in the past, their cards using faster ram will make up for not having a lot of vram and that's nonsense. Yea maybe you will get a few less stutters, but its not optimal regardless.
You are very wrong. Faster memory makes easier to load and unload assets which can make difference
And you can see he bas no vram bottlenecks in most of cases because there are plenty of other things in pipeline that makes card bottleneck, not only gpu overall and memory load
@@Tpecep it doesn't matter how fast it's doing it, it's still goin to stutter or load things in late, whether the stutter is smaller or the loads it in less late, still doesn't matter, you don't want to play a game that way. I am very right. These days it's even effecting ray tracing performance, this is why amd cards are being Nvidia cards with ray tracing on, when the Nvidia card runs out of vram. So yes you need enough vram full stop, faster ram isn't goin to solve that issue.
@@anthonyrizzo9043 VRAM used in testing at 1440P Ultra
Cyberpunk: 6.25GB
Cyberpunk w/RT: 8.75GB
DOOM Eternal: 8.25GB
DOOM Eternal w/RT: 8.5GB
Hogwarts: 6.0GB
Hogwarts w/RT: 6.0GB
Project CARS 3: 4.2GB
Portal with RTX: 6.0GB
Red Dead 2: 6.0GB
Spider Man: 7.8GB
Spider Man w/RT: 8.25GB
Starfield: 5GB
Tiny Tina: 4.75GB
I have my 4070 Ti undervolted at 2700mhz @ 0.950Mv and overclocked my memory to 600GB/s stable is that good?
My congratulations. Great result. I have not seen any better settings config.
Just make sure it is stable. "Control" and "Alan Wake 2" are very VRAM sensitive games - the first one gives artifacts (black screen) if memory is not stable, the second one just crushes.
To test the stability of cores, just launch any modern game at native 4K with max settings. If it does not crash in 10-20 minutes, the config is stable.
@@starstreamgamer3704 I've ran 3D Mark Speed Way stress test for 20 minutes and passed with 99.7% stability and various games such as RDR2 with no issues I'll give Alan Wake 2 a shot but I believe this is stable :) I've only tested at 1440p as I don't own a 4k monitor though. oh and if you want to know which model I have I have the ASUS TUF 4070 Ti non OC version.
Update - I ran Alan Wake 2 for 15 minutes and haven't noticed any artifacts or stability issues!
@@RavenaLux Good for you. Enjoy your gaming experience!
Even if you do not have a 4K screen, you can still take advantage of higher resolution. Enable (DL)DSR 2.25x (through Nvidia Control Panel, 3D settings tab). This will let you GPU render games at 4K (2160p) and downscale it to the native res of your screen. This feature vastly improves image quality. You can usually change rendering resolution in game settings. And even big modern games can still take advantage of it - just use DLSS. 4K + DLSS Quality looks much better than native 1440p. Even 4K + DLSS Balanced still has an edge over 1440p. While 4K + DLSS Performance looks very similar to 1440p, but performs quite a bit better. Plus, there is also 1.78x factor, which is in between 4K and 1440p.
Do not forget about "DSR smoothness" parameter - for (DL)DSR the most optimal value, in my experience, is ~65%.
@@RavenaLux Good for you. Enjoy your gaming experience!
Even if you do not have a 4K screen, you can still take advantage of higher resolution. Enable (DL)DSR 2.25x (through Nvidia Control Panel, 3D settings tab). This will let you GPU render games at 4K (2160p) and downscale it to the native res of your screen. This feature vastly improves image quality. You can usually change rendering resolution in game settings. And even big modern games can still take advantage of it - just use DLSS. 4K + DLSS Quality looks much better than native 1440p. Even 4K + DLSS Balanced still has an edge over 1440p. While 4K + DLSS Performance looks very similar to 1440p, but performs quite a bit better. Plus, there is also 1.78x factor, which is in between 4K and 1440p.
Do not forget about "DSR smoothness" parameter - for (DL)DSR the most optimal value, in my experience, is ~65%.
@@starstreamgamer3704 Thanks for the tip! but since it's stable at 1440p which is what I game on only I'll just leave it at that, I'm quite happy with this 😁
After you test it is easy to calculate that 4070 "non Ti" will be around 4-5% faster than 3080 10GB or almost exactly the same as 3080 12GB. For 600$ it looks like fair deal.
Maybe. We will know it soon enough. The relative performance level will probably depend on resolution. 3080/12 has enormous memory bandwidth (900+ GB/s) and in some games at 4K at almost matches 4070 Ti. But at 1440p it will probably be close to 4070.
There are rumors that there will be two versions of 4070. One with 8-pin power connector and 200W TDP and the other one with 12VHPWR connector and enhanced TDP. I bet the "enhanced" version will cost more than $600 (excluding maybe just "Founders edition").
I think so, I paid £585 for a 2070 3 years ago. I have to say that I'm a little disappointed with the 40×× series, and I think the 4070, 4070Ti & 4080 could all do with a plus 30 FPS performance bump.
@@oktc68 I agree. The only card which justifies it's price in comparison with the previous generation is 4090.
4080, 4070 Ti and 4070 should either be more powerful or less expensive. +10-15% more performance (and more VRAM - 20GB for 4080 and 16GB for 4070/4070 Ti) or it should have been $999 for 4080, $699 for 4070 Ti and $499 for 4070.
I got a 4070 to replace a 3070
It’s fantastic at 1080p and 1440p…. At 4k it’s ok. As stated above, on average it’s the same or slightly worse than a 3080 in some games, in other games the 4070 can beat a 3090 even at 4k, just depends on the game. Power hungry games and console conversions seem to fair worse at higher resolutions but unless you compare side by side you’d probabley never know that the 4070 was slightly inferior to a 3080 at 4k.
The reason I got it was because it *uses half the power* of my OC Aorus 3070 which could easily pull over 250W and did most of the time. The 4070 hovers around 110-150w only hitting 200w in the most demanding games and stress tests. The 4070 undervolted like a champ! This means my room is cooler, PC quieter and I subsequently play more games and even got a gaming mouse 😀
192bit is enough for 1440p. You must test it in 4k
I wanted to make such a comparison, but this card is not fast enough to run modern games at native 4K/Max. And if you use DLSS set to "Quality", this means games are rendered at 1440p and then upscaled to 4K via tensor cores. So the results will be the same as shown in this video.
For older games it might be more useful though. I tested one game at 4K (Horizon Zero Dawn) and found out that VRAM OC brings 8% more frames at native 4K. But since there is no DLAA in this game and modded DLAA (via DLSS Tweaks) performs significantly worse than TAA (and do not achieve locked 60 fps), I do not see any sense in not using DLSS Quality. So testing HZD at native 4K is kinda irrelevant.
@@starstreamgamer3704 light game turn on 4k max but modern games turn on in 4k not im max, just medium in nativ 4k
I mean that helps but 3090/ti and 4090 still have almost double the bandwidth
They do. But only 4090 can truly take advantage of such a high memory bandwidth at native 4K res. Other cards simply can not run modern AAA titles at native 4K anyway, even 4080 struggles a bit. And at native 1440p res (or 4K with DLSS) there is not much use in 700+ GB/s memory bandwidth anyway. So 3080 12GB/Ti and 3090 (Ti) do not really benefit much from (almost) doubled bandwidth. Only in some older titles 3090 Ti can provide 4K/60+fps and outperform 4070 Ti. While in newer games at 1440p/60+fps usually 4070 Ti is roughly on par with 3090 Ti, despite having twice as narrow memory bus.
@@starstreamgamer3704I found the 4070 Ti to be a great match for 3440x1440. My memory can comfortably hit 600 GBps and +180 core giving a noticeable boost in most games. 4k is a big jump from 3440 in gpu requirements, I basically get 4090 4K frames at 3440 which is shocking considering the 4090 costed 2.3x my card (more now due to China 4090 ban)
@@starstreamgamer3704 try cp2077 path tracing MAX GRAPHICS on 4070ti 4k balanced + NVFG ON vs 3090ti path tracing max FSR3FG MOD ON and see whats happen.. poor guy 500gb/s. The real 500gb/s issues only you see with high bandwidth demand.. rt/pt/dlss/fg on.. see the "magic" NVFG ON 20fps vs NVFG OFF 30fps.. 500GB/S IS A 3060TI DEMAND. But see, nvidia abort 4070ti and release 4070tiS 688GB/S LMAO
mine is stable at 25000 MHz (600 GB/s) and 3045mhz on core with 1.1mv
Good for you. Looks like you won the silicon lottery in terms of VRAM overclock. You might even got better result if Afterburner would not have any limits.
What is the performance improvement percentage-wise in your case?
@@starstreamgamer3704 10-11% at 4k, yes i just put the maximum on afterburner, dont let me go higher.
@@TheDesconection poor lier #2
overclock mama uns 4 fps, se for abaixo de 60 fps tlvz vale a pena mas dependendo a mamada de energia não compensa
wtf 4070 ti isn't 504 GB/s
It is! Look at it's tech specifications on any website (TechPowerUp, for example).
Memory config:
Memory Size - 12 GB
Memory Type - GDDR6X
Memory Bus - 192 bit
Bandwidth - 504.2 GB/s
overclocking for 3 or 4 fps doesn't worth it.
Why not? There is no significant increase in either power consumption or temperatures or fans speed. So overclocking memory (VRAM) is literally free performance for very low effort.
Overclocking GPU is a different story though and depends a lot on the type and size of a GPU cooler, room temperatures, PC case quality, etc.
I'd rather undevoted a GPU and overclocked memory. It is a perfect combination of performance/temperatures/noise ratio.