Maybe the joke's been made a hundred times, but give yourself a pat on the back for performance improvements being "in the cards," that got a solid guffaw from me
Anandtech doesn't do reviews, they turned into the news only and spam channel with "Deals" on hardware. Doesn't matter if its shitty hardware, its a deal and they make money.
I doubt they even have people to review stuff anymore. Given the current owners are more "technical" types vs hobbyists like old days.
You're going to have to head over to Tom's Hardware. Years ago, the hierarchy was reversed. Anand had such a great site. We'd get the best reviews that were fairly independent. My self 20 years ago would be wondering how TF did this all change. It's too bad that AT didn't survive until today. I miss it being a site that did reviews. It's closer to cnet these days. Lame.
Those cards will either come down in price or be replaced by higher performing variants once the 30 series stock is gone and the 40 series mainstream cards are launched. If NVIDIA made the 12 GB 4080 cost $600 who is going to buy the 3080 for any more than $400? At $400 NVIDIA is probably looking at as loss for the 3080.
"If NVIDIA made the 12 GB 4080 cost $600 who is going to buy the 3080 for any more than $400?"
The problem is the 12 GB 4080 is actually a 4070 gen (see specs) with a new price tag, going from $499 to $899. Hopefully they won't succeed with this scam.
Everyone was expecting MSRPs to go up, but those 4080 prices are delusional. 3080 10GB MSRP was $700, and now they want $900 for the 12GB 4080? In this economy? Post Merge?
To be fair, they were greedy fvckers with the RTX-30 series as well. But at least they had the performance to back it up, especially when it came to compute/cryptomining.
So if their x2 Performance uplift in rasterization is true, then their higher MSRP can be justified. But that's a BIG "IF" knowing we're talking about Nvidia here and their history of exaggerating claims and cheating benchmarks.
I think the best advice is to buy a pre-order asap if you're in the market and have the money. You remember what happened last time. But if money is tight, or you are patient, then wait for a couple months to see the rest of the lineup (RTX 4070, 4060) and also see AMD's upcoming RDNA-3 (RX-6000 series) of cards. Meanwhile, the thrifty gamers out there will be upgrading from their used GTX 1070 / RX5700 / RTX 2070Ti cards, and grabbing some second hand cards (RTX 3070Ti / RX 6800) from the first group listed above.
This 4080 is most likely the "origin" 4070Ti, they charge the same price for 1-tier-lower card because of the current market condition. Seriously 192-bit & 7680 CUDA cores for RTX 4080 lol
They are trying to not sell their old cards for a loss. If they price the new cards lower then no one is going to buy the old cards except for a price that wouldn't turn a profit for them.
Yes, it's probably exactly the same idea and purpose. It would have been a big deal for Intel if they had been able to get the Arc cards out in 2020 like they originally had promised.
As expected, Nvidia is using RT and DLSS to make their claims. Not only this, but they are: - using frame interpolation in the calulation to double the FPS, and - using performance DLSS mode...
Nvidia is doing that to mislead their customers. So, in Cyberpunk, we know that a 4090 is 2 times faster without frame interpolation, and RT on, and DLSS performance...
-This means raster is probably around 50%... Nvidia did exactly the same thing with Ampere.
NVIDIA is using RT and DLSS to make some of their claims. How is that misleading? How are RT and DLSS "fake"? They even have specific claims about where it what times faster. I'm not sure why you're claiming it's 50% faster in rasterization-only even under a specific game. I mean I'm sure it is if you choose the right settings. And if you choose other settings it's probably 100% faster. It has over twice the TFLOPS and, although we don't have a lot of information on the architecture yet, there's no reason to doubt that the computational efficiency should have gone down from Ampere to Ada
Before we go wild about the shader count appearing like it's a step down from the 3xxx series, let's be clear that we don't know what makes up a shader unit for these chips. The 3xxx series effectively double counted their units, so maybe the 4xxx series is moving back to a more realistic view of shader units. If that's the case, than the 4080 will have more full shader units than the 3080. We really don't have enough information at this point, but it wouldn't be surprising with the increases in transistor counts.
Actually, you could argue none of those are 4080s. The 192 bit bus is usually reserved for the 60s card. The 4080 12 GB also has less than half the compute than the 4090 which is also usually reserved for the 60s card. Nvidia has gone and moved the tiers by several rungs and more than doubled the price.
You can’t just blame the increase purely on increased costs either. Other companies also produce chips and I have not seen such increase in prices from them.
Nothing compelling here about next gen gaming GPUs to justify the cost of entry. I can't see this being a good thing for the already shrunken desktop gaming market.
Agree with PeachNCream, nothing compelling to justify the cost. Not in the market for a 4090 or a 4080 at those prices with power consumption through the roof. In today's world Nvidia's best release for ages was the original RTX-3070 FE (well for 2k gaming) for performance vs power vs price. Sadly unavailable nowadays.
performance increase the nvidia way. instead of incresing the real perfomance like doing double the work per second. they change the work. now interpolated images are increasing the framerate. yeah that saves a lot of computing power.. but it´s not doubling the compute performance. i can´t say my PS5 is doing twice the work when i let my TV doing motion interpolation. i get that the topic is more complicated in teh case of the RTX 4000 series...but the nvidia marketing is a shity as always....
RTX 4090 is where Nvidia is aiming at. They want the top paying consumers to buy that card and present it as best performance for value vs Ampere.
Important points are when Ampere launched Jensen know that thing was going to miners which is why he got special Micron made GDDR6X memory and also beta tested the waters with insane price markups for 100% price increase on a 3080 for just 15% max perf and VRAM. He succeeded since AMD did not get their marketshare increased in anyway when FE was selling at MSRP at BestBuy in America and EVGA got their queue system for MSRP. So now post all that Covid he wants to normalize this BS pricing.
Now Ada is normalizing the high prices. RTX4080 AD103 is not even full die, it is also cut die just like RTX4090 AD102. So you can expect Ti versions either next year or later. All buyers are going to get shafted soon just like 3090 Buyers when Nvidia haphazardly designed that GPU to have awful MSVDD rail power delivery and flaky PCB design which were destroyed by NewWorld add the extreme power excursions, only to be fixed by 3090Ti PCB which is the best of Ampere generation proper pricing at $1000 now.
So Nvidia thinks they can charge more now since $1000 is a price cut over $2000 of 3090Ti and they can bank on $1500 MSRP of 3090. Ultimate greedy pile of cancer move. But it doesn't stop there. USD vs World Currency is in shambles everywhere the prices of common imported goods have risen hard. In EU the price increase is 25% more than American. RIP
RTX4080 12GB AD104 is mega gimp rip off lol, it's not even as fast as 3090Ti in Raster but also a 192Bit bus with crippled VRAM and defective die on top for $900, AIB markup is $100-250 on all cards so this gen is going to be the worst of all. I really really hope people are not dumb to waste cash and give money to this trash company and look at alternatives of either buying Ampere / RDNA2 / RDNA3 or skip entirely.
Finally the DLSS3.0, Awful technology they are faking the damn frames how can anyone not see this BS ? They are adding fake frames guys !! It's much worse than upscaling nonsense they started, now that FSR2.0 essentially destroyed Nvidia they launch even bigger scam rofl. ALSO bonus is Ampere gets shafted by so called OFA, if you see the slide deck they showed, you can clearly see OFA present in Ampere but CEO Jensen is lying to face that it wont work on Ampere cards. Mega castration, not only proprietary trash DLSS but also gated behind new gen, so expect 4000 getting EOLed the moment 5000 launches.
Also showing that pile of rubble Cybertrash ? Portal lol. WTF Nvidia, your trash DLSS2.x did not improve a bit on the new 40 series because it was a SHAM as FSR2.0 even caught up now that you cannot show off how the new ADA Tensor and RT cores can perform better you simply add new DLSS lol, if this was legit innovation we could have seen Ada Lovelace RTX4090 performance on METRO EXODUS with existing DLSS not new crap gated behind a paywall and more integration, this is junk move. Awful trash that is like Hairworks garbage. So they show off pathetic awful trash game like Cyberpunk and 21FPS lol vs Fake Frame Interpolation at 100FPS !!
Guys if you have 1 cell you should avoid this scam and move away from this cancer company. I'm an owner of RTX3090 FE and saying that.
Behold, the overpriced GPU and the software demos for the %1. I mean who here thinks outside a few specific companies is using this software on a daily basis. %100 most of this is a "proof of concept" but they just want developers to know what they can do. lol
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
48 Comments
Back to Article
popej - Tuesday, September 20, 2022 - link
The title looks 2 years old ;)Gavin Bonshor - Tuesday, September 20, 2022 - link
Thanks for spotting that, I've edited it now :)Unashamed_unoriginal_username_x86 - Tuesday, September 20, 2022 - link
Maybe the joke's been made a hundred times, but give yourself a pat on the back for performance improvements being "in the cards," that got a solid guffaw from meshabby - Tuesday, September 20, 2022 - link
Speaking of jokes... when are gpu reviews coming back to AnandTech?Thunder 57 - Tuesday, September 20, 2022 - link
That's a good one. Anandtech may go the way of the dodo before that happens.herozeros - Tuesday, September 20, 2022 - link
My curious bits are still waiting for a M1 Ultra deep dive . . . lot has changed hereDifferentFrom - Tuesday, September 20, 2022 - link
I'm still waiting for the Polaris deep dive.Kangal - Wednesday, September 21, 2022 - link
I'm just waiting for any dive, could even be a poolpowerarmour - Wednesday, September 21, 2022 - link
Probably when Intel are competitive in 2028imaheadcase - Saturday, September 24, 2022 - link
Anandtech doesn't do reviews, they turned into the news only and spam channel with "Deals" on hardware. Doesn't matter if its shitty hardware, its a deal and they make money.I doubt they even have people to review stuff anymore. Given the current owners are more "technical" types vs hobbyists like old days.
bigboxes - Saturday, September 24, 2022 - link
You're going to have to head over to Tom's Hardware. Years ago, the hierarchy was reversed. Anand had such a great site. We'd get the best reviews that were fairly independent. My self 20 years ago would be wondering how TF did this all change. It's too bad that AT didn't survive until today. I miss it being a site that did reviews. It's closer to cnet these days. Lame.Jeff72 - Tuesday, September 20, 2022 - link
GeForce RTX 4080, 16GB and 12GB GDDR6 versions, starting at $899not
GeForce RTX 3080, 16GB and 12GB GDDR6 versions, starting at $899
Chaitanya - Tuesday, September 20, 2022 - link
"4080 12GB" 7680 CUDA cores and 4080 16GB 9728 CUDA cores, so its nice price hike from $500 of 3070 to mere $900 this time around.Yojimbo - Wednesday, September 21, 2022 - link
Those cards will either come down in price or be replaced by higher performing variants once the 30 series stock is gone and the 40 series mainstream cards are launched. If NVIDIA made the 12 GB 4080 cost $600 who is going to buy the 3080 for any more than $400? At $400 NVIDIA is probably looking at as loss for the 3080.gescom - Thursday, September 22, 2022 - link
"If NVIDIA made the 12 GB 4080 cost $600 who is going to buy the 3080 for any more than $400?"The problem is the 12 GB 4080 is actually a 4070 gen (see specs) with a new price tag, going from $499 to $899. Hopefully they won't succeed with this scam.
Rudde - Tuesday, September 20, 2022 - link
Completely new frames that skip the CPU and the rendering pipeline? How much is made by the game and how much is just guessed by AI at that point?Mr Perfect - Tuesday, September 20, 2022 - link
Everyone was expecting MSRPs to go up, but those 4080 prices are delusional. 3080 10GB MSRP was $700, and now they want $900 for the 12GB 4080? In this economy? Post Merge?Gothmoth - Tuesday, September 20, 2022 - link
they are greedy fukheads..... no wonder most of my friends are playing on consoles now.Kangal - Wednesday, September 21, 2022 - link
To be fair, they were greedy fvckers with the RTX-30 series as well. But at least they had the performance to back it up, especially when it came to compute/cryptomining.So if their x2 Performance uplift in rasterization is true, then their higher MSRP can be justified. But that's a BIG "IF" knowing we're talking about Nvidia here and their history of exaggerating claims and cheating benchmarks.
I think the best advice is to buy a pre-order asap if you're in the market and have the money. You remember what happened last time. But if money is tight, or you are patient, then wait for a couple months to see the rest of the lineup (RTX 4070, 4060) and also see AMD's upcoming RDNA-3 (RX-6000 series) of cards. Meanwhile, the thrifty gamers out there will be upgrading from their used GTX 1070 / RX5700 / RTX 2070Ti cards, and grabbing some second hand cards (RTX 3070Ti / RX 6800) from the first group listed above.
Mr Perfect - Tuesday, September 20, 2022 - link
And I suspect the "RTX 3060 now starting at $329 (again?)" is in reference to the rumored 8GB 3060. https://videocardz.com/newz/geforce-rtx-3060-8gb-a...Getting back to MSRP by slashing the memory size, bus, and bandwidth would be a bold move, if that's what they're doing.
flatwhite - Tuesday, September 20, 2022 - link
This 4080 is most likely the "origin" 4070Ti, they charge the same price for 1-tier-lower card because of the current market condition. Seriously 192-bit & 7680 CUDA cores for RTX 4080 lolshabby - Tuesday, September 20, 2022 - link
Lol that is definitely a lower tiered card, good luck selling that junk Nvidia.DougMcC - Tuesday, September 20, 2022 - link
I'm sure they'll sell all they can make for a quarter until RDNA3 competition forces a repricing.Yojimbo - Wednesday, September 21, 2022 - link
They are trying to not sell their old cards for a loss. If they price the new cards lower then no one is going to buy the old cards except for a price that wouldn't turn a profit for them.Gothmoth - Tuesday, September 20, 2022 - link
lol... 900$ for a 12 GB card in 2022. nvidia has a screw loose....Chaitanya - Tuesday, September 20, 2022 - link
So nGreedia is pulling same stunt they did with 1060 few years back but now they are selling 70 series as 80 series.xol - Tuesday, September 20, 2022 - link
Shader Execution Reordering sounds like Nvidia's equivalent in Intel Arcs Thread Sorting Unit and probably is.Yojimbo - Wednesday, September 21, 2022 - link
Yes, it's probably exactly the same idea and purpose. It would have been a big deal for Intel if they had been able to get the Arc cards out in 2020 like they originally had promised.eva02langley - Tuesday, September 20, 2022 - link
As expected, Nvidia is using RT and DLSS to make their claims. Not only this, but they are:- using frame interpolation in the calulation to double the FPS, and
- using performance DLSS mode...
Nvidia is doing that to mislead their customers. So, in Cyberpunk, we know that a 4090 is 2 times faster without frame interpolation, and RT on, and DLSS performance...
-This means raster is probably around 50%... Nvidia did exactly the same thing with Ampere.
Yojimbo - Wednesday, September 21, 2022 - link
NVIDIA is using RT and DLSS to make some of their claims. How is that misleading? How are RT and DLSS "fake"? They even have specific claims about where it what times faster. I'm not sure why you're claiming it's 50% faster in rasterization-only even under a specific game. I mean I'm sure it is if you choose the right settings. And if you choose other settings it's probably 100% faster. It has over twice the TFLOPS and, although we don't have a lot of information on the architecture yet, there's no reason to doubt that the computational efficiency should have gone down from Ampere to Adaquorm - Tuesday, September 20, 2022 - link
Any announcements on the professional end (availability dates) Didn't have time to watch. Looks like they will call it RTX 6000 again?Ryan Smith - Tuesday, September 20, 2022 - link
Yeah, it's RTX 6000 again. I'll have something up later (after 4090/4080).WaltC - Tuesday, September 20, 2022 - link
I liked the "relative performance" bar charts...;) Not very interesting from where I'm sitting.Dizoja86 - Tuesday, September 20, 2022 - link
Before we go wild about the shader count appearing like it's a step down from the 3xxx series, let's be clear that we don't know what makes up a shader unit for these chips. The 3xxx series effectively double counted their units, so maybe the 4xxx series is moving back to a more realistic view of shader units. If that's the case, than the 4080 will have more full shader units than the 3080. We really don't have enough information at this point, but it wouldn't be surprising with the increases in transistor counts.regsEx - Tuesday, September 20, 2022 - link
No SM structure slides this time?boozed - Tuesday, September 20, 2022 - link
One of those 4080s is a 4080 and the other one isn't.citan x - Tuesday, September 20, 2022 - link
Actually, you could argue none of those are 4080s. The 192 bit bus is usually reserved for the 60s card. The 4080 12 GB also has less than half the compute than the 4090 which is also usually reserved for the 60s card. Nvidia has gone and moved the tiers by several rungs and more than doubled the price.You can’t just blame the increase purely on increased costs either. Other companies also produce chips and I have not seen such increase in prices from them.
boozed - Thursday, September 22, 2022 - link
I think you're rightPeachNCream - Wednesday, September 21, 2022 - link
Nothing compelling here about next gen gaming GPUs to justify the cost of entry. I can't see this being a good thing for the already shrunken desktop gaming market.EasterEEL - Wednesday, September 21, 2022 - link
Agree with PeachNCream, nothing compelling to justify the cost. Not in the market for a 4090 or a 4080 at those prices with power consumption through the roof. In today's world Nvidia's best release for ages was the original RTX-3070 FE (well for 2k gaming) for performance vs power vs price. Sadly unavailable nowadays.Still dislike that middle connection though.
Gothmoth - Wednesday, September 21, 2022 - link
performance increase the nvidia way.instead of incresing the real perfomance like doing double the work per second.
they change the work. now interpolated images are increasing the framerate. yeah that saves a lot of computing power.. but it´s not doubling the compute performance. i can´t say my PS5 is doing twice the work when i let my TV doing motion interpolation. i get that the topic is more complicated in teh case of the RTX 4000 series...but the nvidia marketing is a shity as always....
Yojimbo - Wednesday, September 21, 2022 - link
Yeah, stupid tricks to get the best experience, like has been driving graphics forward for the last 30 years.MTEK - Wednesday, September 21, 2022 - link
DisplayPort 1.4? *cries*Yojimbo - Wednesday, September 21, 2022 - link
Nvidia is going to run out of letters to append before GX.Silver5urfer - Wednesday, September 21, 2022 - link
RTX 4090 is where Nvidia is aiming at. They want the top paying consumers to buy that card and present it as best performance for value vs Ampere.Important points are when Ampere launched Jensen know that thing was going to miners which is why he got special Micron made GDDR6X memory and also beta tested the waters with insane price markups for 100% price increase on a 3080 for just 15% max perf and VRAM. He succeeded since AMD did not get their marketshare increased in anyway when FE was selling at MSRP at BestBuy in America and EVGA got their queue system for MSRP. So now post all that Covid he wants to normalize this BS pricing.
Now Ada is normalizing the high prices. RTX4080 AD103 is not even full die, it is also cut die just like RTX4090 AD102. So you can expect Ti versions either next year or later. All buyers are going to get shafted soon just like 3090 Buyers when Nvidia haphazardly designed that GPU to have awful MSVDD rail power delivery and flaky PCB design which were destroyed by NewWorld add the extreme power excursions, only to be fixed by 3090Ti PCB which is the best of Ampere generation proper pricing at $1000 now.
So Nvidia thinks they can charge more now since $1000 is a price cut over $2000 of 3090Ti and they can bank on $1500 MSRP of 3090. Ultimate greedy pile of cancer move. But it doesn't stop there. USD vs World Currency is in shambles everywhere the prices of common imported goods have risen hard. In EU the price increase is 25% more than American. RIP
RTX4080 12GB AD104 is mega gimp rip off lol, it's not even as fast as 3090Ti in Raster but also a 192Bit bus with crippled VRAM and defective die on top for $900, AIB markup is $100-250 on all cards so this gen is going to be the worst of all. I really really hope people are not dumb to waste cash and give money to this trash company and look at alternatives of either buying Ampere / RDNA2 / RDNA3 or skip entirely.
Finally the DLSS3.0, Awful technology they are faking the damn frames how can anyone not see this BS ? They are adding fake frames guys !! It's much worse than upscaling nonsense they started, now that FSR2.0 essentially destroyed Nvidia they launch even bigger scam rofl. ALSO bonus is Ampere gets shafted by so called OFA, if you see the slide deck they showed, you can clearly see OFA present in Ampere but CEO Jensen is lying to face that it wont work on Ampere cards. Mega castration, not only proprietary trash DLSS but also gated behind new gen, so expect 4000 getting EOLed the moment 5000 launches.
Also showing that pile of rubble Cybertrash ? Portal lol. WTF Nvidia, your trash DLSS2.x did not improve a bit on the new 40 series because it was a SHAM as FSR2.0 even caught up now that you cannot show off how the new ADA Tensor and RT cores can perform better you simply add new DLSS lol, if this was legit innovation we could have seen Ada Lovelace RTX4090 performance on METRO EXODUS with existing DLSS not new crap gated behind a paywall and more integration, this is junk move. Awful trash that is like Hairworks garbage. So they show off pathetic awful trash game like Cyberpunk and 21FPS lol vs Fake Frame Interpolation at 100FPS !!
Guys if you have 1 cell you should avoid this scam and move away from this cancer company. I'm an owner of RTX3090 FE and saying that.
elforeign - Thursday, September 22, 2022 - link
Roids bro?Harry_Wild - Thursday, September 22, 2022 - link
Everyone call the NVidia CEO by his first name: Jensen! LOL!imaheadcase - Saturday, September 24, 2022 - link
Behold, the overpriced GPU and the software demos for the %1. I mean who here thinks outside a few specific companies is using this software on a daily basis. %100 most of this is a "proof of concept" but they just want developers to know what they can do. lol