NVIDIA RTX 3090, 3080, 3070 Specs, Cooler, Price, & Release Date
video description
Date: 2020-09-02
Related videos
Comments and reviews: 10
Speak
I have been saying this from when the rumours started about the XSX chip only being around 360mm 2, if AMD can make a 56CU chip with all the CPU stuff, IO stuff and all the other crap a full SOC needs, if they can make it that small (and still have very decent clocks when you compare it to Nvidia's cards) then AMD could easily make a 120CU chip if they wanted and still be smaller than Nvidia's 3080/3090 chip!
We know TSMC's 7nm+ has amazing yields and allows for very large, very complex dies, it's just slow and costly at the moment but that doesn't matter for the low volume 1500+ market.
AMD could make:
120CU > 3090, on 7nm+, could come with HBM, 1499- 1999
100CU > 3080Ti (close to 3090), on 7nm+, could come with HBM if needed 1199
80CU > 3080, >10GB GDDR6x, on the cheaper N7P node, 699
60CU > 3070, >8GB GDDR6x, on the cheaper N7P node, 499
50CU > 3060, 8GB GDDR6, on the cheaper N7P node, 349 (XSX sits slightly above this for 499 but that's a obviously a full system price)
40CU > RX580 199 (I'm not sure on this but there needs to be a card that properly beats the RX580 for value this generation)
I am not some fanboy saying this is what will DEFINITELY HAPPEN! I am just saying given what we know about the XSX chip and how much of a node advantage TSMC has over Samsung's 8nm this would be possible if AMD chose to go all out in the GPU space.
I am actually really impressed with Nvidia's 3070 and 3080, they look great, priced amazingly competitively and unless AMD does do something like the above I will almost certainly pick up one of Nvidia's cards because some of the new features do look pretty interesting, although maybe I'll buy a console depending on their price and games. Lets see how it all unfolds.
reply
I have been saying this from when the rumours started about the XSX chip only being around 360mm 2, if AMD can make a 56CU chip with all the CPU stuff, IO stuff and all the other crap a full SOC needs, if they can make it that small (and still have very decent clocks when you compare it to Nvidia's cards) then AMD could easily make a 120CU chip if they wanted and still be smaller than Nvidia's 3080/3090 chip!
We know TSMC's 7nm+ has amazing yields and allows for very large, very complex dies, it's just slow and costly at the moment but that doesn't matter for the low volume 1500+ market.
AMD could make:
120CU > 3090, on 7nm+, could come with HBM, 1499- 1999
100CU > 3080Ti (close to 3090), on 7nm+, could come with HBM if needed 1199
80CU > 3080, >10GB GDDR6x, on the cheaper N7P node, 699
60CU > 3070, >8GB GDDR6x, on the cheaper N7P node, 499
50CU > 3060, 8GB GDDR6, on the cheaper N7P node, 349 (XSX sits slightly above this for 499 but that's a obviously a full system price)
40CU > RX580 199 (I'm not sure on this but there needs to be a card that properly beats the RX580 for value this generation)
I am not some fanboy saying this is what will DEFINITELY HAPPEN! I am just saying given what we know about the XSX chip and how much of a node advantage TSMC has over Samsung's 8nm this would be possible if AMD chose to go all out in the GPU space.
I am actually really impressed with Nvidia's 3070 and 3080, they look great, priced amazingly competitively and unless AMD does do something like the above I will almost certainly pick up one of Nvidia's cards because some of the new features do look pretty interesting, although maybe I'll buy a console depending on their price and games. Lets see how it all unfolds.
reply
RandomlyDrumming
Well, as someone who has a close acquaintance working at Radeon Technologies, from the little inside info I have (if you can call it that), I can say that many people will be very surprised with what top RDNA2 can do...Why do you think Nvidia all of a sudden decided to so drastically drop the prices gen-to-gen, basically screwing over people who bought 2080Ti in the last few months or weeks? Nvidia actually knows MORE about RDNA2 than most tech tubers with their leaks. One of the reasons AMD is still keeping quiet about their tech is because of the consoles and agreements they signed with MS and Sony.
So, what my acquaintance told me a few hours ago when I quickly pinged him and asked him what does he make of Nvidia's presentation is (and I quote): Dude, we expected this and it's nothing new to us. If mng decides to stay on course with current plans regarding SKUs and pricing we're heading for another Ryzen moment only with GPUs ;) .
TL;DR - I'm definitely gonna wait for AMD to show their cards.
reply
Well, as someone who has a close acquaintance working at Radeon Technologies, from the little inside info I have (if you can call it that), I can say that many people will be very surprised with what top RDNA2 can do...Why do you think Nvidia all of a sudden decided to so drastically drop the prices gen-to-gen, basically screwing over people who bought 2080Ti in the last few months or weeks? Nvidia actually knows MORE about RDNA2 than most tech tubers with their leaks. One of the reasons AMD is still keeping quiet about their tech is because of the consoles and agreements they signed with MS and Sony.
So, what my acquaintance told me a few hours ago when I quickly pinged him and asked him what does he make of Nvidia's presentation is (and I quote): Dude, we expected this and it's nothing new to us. If mng decides to stay on course with current plans regarding SKUs and pricing we're heading for another Ryzen moment only with GPUs ;) .
TL;DR - I'm definitely gonna wait for AMD to show their cards.
reply
LBXZero
I am noticing a problem with the specs for the RTX 3000 Series. Everything that I am aware of, the CUDA cores in the RTX 3000 series are suppose to have 2 FP32 Units instead of the traditional 1. Looking at the 20 TFLOP number for the RTX3070, either that 2nd FP32 Unit is not being utilized, the CUDA cores are still 1 FP32 Unit, or Nvidia is exaggerating how many CUDA cores are in the RTX 3000 series to represent the FP32 performance. If they are 1 CUDA core with 2 FP32 Units, we can expect the RTX 3070 to perform at minimum equivalent to an RTX 2080 to the peak of 50% better than an RTX 2080ti depending on how being a single CUDA core with 2 FP32 Units affects the normal operation flow.
reply
I am noticing a problem with the specs for the RTX 3000 Series. Everything that I am aware of, the CUDA cores in the RTX 3000 series are suppose to have 2 FP32 Units instead of the traditional 1. Looking at the 20 TFLOP number for the RTX3070, either that 2nd FP32 Unit is not being utilized, the CUDA cores are still 1 FP32 Unit, or Nvidia is exaggerating how many CUDA cores are in the RTX 3000 series to represent the FP32 performance. If they are 1 CUDA core with 2 FP32 Units, we can expect the RTX 3070 to perform at minimum equivalent to an RTX 2080 to the peak of 50% better than an RTX 2080ti depending on how being a single CUDA core with 2 FP32 Units affects the normal operation flow.
reply
Sir
3 times quieter is typical marketing BS.
Quieter than what? What's the reference here?
Also loudness is subjective and thus hard to measure. A doubling of loudness has been found to be a level change between 6 and 10 dB, so it's rather fuzzy.
If you want to measure you can assume that the mean sounds 1/3 as loud as ... . Then you can do the math and say that the level is -16 dB or the sound pressure is roughly 10 dB lower than whatever their reference is. That you could measure, but it's all meaningless without reference.
Unless my math is wrong, wouldn't be the first time as the various forms in which sound can be measured are very confusing.
reply
3 times quieter is typical marketing BS.
Quieter than what? What's the reference here?
Also loudness is subjective and thus hard to measure. A doubling of loudness has been found to be a level change between 6 and 10 dB, so it's rather fuzzy.
If you want to measure you can assume that the mean sounds 1/3 as loud as ... . Then you can do the math and say that the level is -16 dB or the sound pressure is roughly 10 dB lower than whatever their reference is. That you could measure, but it's all meaningless without reference.
Unless my math is wrong, wouldn't be the first time as the various forms in which sound can be measured are very confusing.
reply
Marco
About the cooler: i may be misunderstanding the design, but doesn't the front fan just blow out hot air directly through the pcie slots, actually removing it form the case without interference?
I mean, compared to a standard open cooler card, where all air from the GPU is just blown around and into the case, you have some of the heat directly removed. I agree that a direct stream of hot air in front of the PCU cooler isn't good, but total hot air volume generated into the case should be less.
I can be totally worng though. I'm just doing guesswork.
reply
About the cooler: i may be misunderstanding the design, but doesn't the front fan just blow out hot air directly through the pcie slots, actually removing it form the case without interference?
I mean, compared to a standard open cooler card, where all air from the GPU is just blown around and into the case, you have some of the heat directly removed. I agree that a direct stream of hot air in front of the PCU cooler isn't good, but total hot air volume generated into the case should be less.
I can be totally worng though. I'm just doing guesswork.
reply
Dan
The GPU location on the motherboard historically makes it extremely in-efficient to cool with direct heatsink and fans (aka air cooler) on both CPU and GPU, even with a normal backplate the hot air from the GPU either gets stuck at the bottom of the case (bad) or goes up to the exhaust fan which is normally located right behind the CPU. So it seems logical that it will be the same hot air blowing up, but it will just blow up faster. All this unless the case has a side exhaust, which most cases these days don't have.
reply
The GPU location on the motherboard historically makes it extremely in-efficient to cool with direct heatsink and fans (aka air cooler) on both CPU and GPU, even with a normal backplate the hot air from the GPU either gets stuck at the bottom of the case (bad) or goes up to the exhaust fan which is normally located right behind the CPU. So it seems logical that it will be the same hot air blowing up, but it will just blow up faster. All this unless the case has a side exhaust, which most cases these days don't have.
reply
Sparky
RTX 3070 only seems like a good deal because its being compared to the RTX 2000 series. As you recall, that series was universally panned for being overpriced gimmicky trash which barely performed better than the cheaper previous cards. Previous cards which are now 4 year old tech.
Ray tracing and DLSS don't mean anything when literally 21% of all games that support it today (2 years later) are Wolfenstein Youngblood, Shadow of the Tomb Raider, and Quake 2.
reply
RTX 3070 only seems like a good deal because its being compared to the RTX 2000 series. As you recall, that series was universally panned for being overpriced gimmicky trash which barely performed better than the cheaper previous cards. Previous cards which are now 4 year old tech.
Ray tracing and DLSS don't mean anything when literally 21% of all games that support it today (2 years later) are Wolfenstein Youngblood, Shadow of the Tomb Raider, and Quake 2.
reply
ChaeDoc
The greenscreen guy is a streamer called Atrioc. He's most famous for being friends with Ludwig but he's a great streamer in his own right.
Thought i'd mention that just incase it made a difference regarding the gamer streamer controlled background thing. And also because i wanted to make the 'most famous' joke.
reply
The greenscreen guy is a streamer called Atrioc. He's most famous for being friends with Ludwig but he's a great streamer in his own right.
Thought i'd mention that just incase it made a difference regarding the gamer streamer controlled background thing. And also because i wanted to make the 'most famous' joke.
reply
Jacob
Anyone think Im going to get considerable CPU bottlenecking on an intel i5 9600k with a 3080? With my water cooling its quite happy going up to 4.5-4.6 Ghz and Im wondering if the 3080 would be good future proofing or a waste of cash. Especially as im getting more into VR and 1440p gaming
reply
Anyone think Im going to get considerable CPU bottlenecking on an intel i5 9600k with a 3080? With my water cooling its quite happy going up to 4.5-4.6 Ghz and Im wondering if the 3080 would be good future proofing or a waste of cash. Especially as im getting more into VR and 1440p gaming
reply
P re
By looking pictures of refrence coolers it looks more like second fan should push air through cooler to bottom of ATX case, not in cpu cooler. At least if you look fan blades orientation, maybe renders are wrong why whould they spin that fan backwards.
reply
By looking pictures of refrence coolers it looks more like second fan should push air through cooler to bottom of ATX case, not in cpu cooler. At least if you look fan blades orientation, maybe renders are wrong why whould they spin that fan backwards.
reply
Add a review, comment
Other channel videos
