Nope. It's been the 104 die. Ampere changed that and they used a heavily cut down 102 die instead of the 103 because of TTM pressures from porting Ampere's big dies to SS8 from SS7 and because even SS8's yield was mediocre.
It doesn't matter what they call the dies. It's the relative position in the stack. In recent prior history, the xx80 GPU was almost always made on the 102 die, which was typically the top consumer die.
That's a poor bet going forward. PC ports typically have higher more demanding visuals which use more VRAM, you're also assuming DirestStorage will deliver. $600 is a lot of money to bet that it all works out. I wouldn't make that bet. That's just planned obsolesce on nvidia's part.
nv's naming is meaningless at this point. the only thing it conveys is how much they think the uninformed shoppers will overpay.It doesn't matter what they call the dies. It's the relative position in the stack. In recent prior history, the xx80 GPU was almost always made on the 102 die, which was typically the top consumer die. But if you go back prior to Pascal, the top die didn't use the 102 designation. Instead it was GM-200 for Maxwell or GK-100/GK-100 (or GK-110/GF-110 for the second generation) for Kepler/Fermi.
The only other time an xx70 GPU was on the third die was Turing, and that is probably one of the worst regarded generations of NVidia cards in long time. And when you look back to Fermi and Kepler when NVidia really started to solidify their modern naming scheme the x70 GPU was on the top die, in some cases having close to 90% of the computational power as the top of the line GPU being sold. Now it's barely 1/3 of the top-end GPU and when NVidia releases a 4090 Ti it might drop below that.
This is an xx70 card in name only. It's a pale imitation of what xx70 once meant and the worst xx70 card relative to the top card ever released. Though it's been something that's been diluted over time, I think they've ruined a proud heritage by calling this thing a 4070. They may as well rename it from Lovelace to Loveless.
Nope. It's been the 104 die. Ampere changed that and they used a heavily cut down 102 die instead of the 103 because of TTM pressures from porting Ampere's big dies to SS8 from SS7 and because even SS8's yield was mediocre.
100% agree. You can just look at the memory bus and tell it's not a traditional 70 series part too.It doesn't matter what they call the dies. It's the relative position in the stack. In recent prior history, the xx80 GPU was almost always made on the 102 die, which was typically the top consumer die. But if you go back prior to Pascal, the top die didn't use the 102 designation. Instead it was GM-200 for Maxwell or GK-100/GK-100 (or GK-110/GF-110 for the second generation) for Kepler/Fermi.
The only other time an xx70 GPU was on the third die was Turing, and that is probably one of the worst regarded generations of NVidia cards in long time. And when you look back to Fermi and Kepler when NVidia really started to solidify their modern naming scheme the x70 GPU was on the top die, in some cases having close to 90% of the computational power as the top of the line GPU being sold. Now it's barely 1/3 of the top-end GPU and when NVidia releases a 4090 Ti it might drop below that.
This is an xx70 card in name only. It's a pale imitation of what xx70 once meant and the worst xx70 card relative to the top card ever released. Though it's been something that's been diluted over time, I think they've ruined a proud heritage by calling this thing a 4070. They may as well rename it from Lovelace to Loveless.
I'd be comparing the 4k results since the 4090 is often cpu bottlenecked at 1440p. That would put 4070 at 51% of 4090's performance.Names are just names, nobody cares. Price and performance matter.
If 4070 is intended to be 2560*1440 card, I got 59% of the performance of the 4090 for 37% price, based on techpowerup average numbers.
I cannot complain.
BTW compared to RX7000 series this card behaves normally, it has acceptable low power draw in idle, multi monitor and video playback. RX7000 seems to be broken in this area.
I wouldn't be surprised. I had an EVGA Hydrocopper with 2 years of warranty left up for a couple months for $750 CAD (~$550 USD) in an admittedly small market and didn't sell it. That's a bit of a niche card, but the market seems really soft for used GPUs right now.The 4070 is going to be able to play lots of games well right now, for sure. I think we’d all prefer if the core of the 4070ti was strapped to a 256 bit bus with 16GB of ram for $600 but obviously it ain’t happening.
What I think will be curious to watch is the used 3080 market. It should get rocked. A used 10GB used for some purpose - and likely an EVGA to boot - for about $500 or a brand new 4070 for $600 plus tax. Seems like an easy decision to me.
Now, if the 3080 drops to more like $400 and we get 12GB 3080 and 3080 Tis for more like sub $500, well my value meter needle might move. That’s what I will be watching, even though I have little need to move on it right now.
People are so thrilled with them that they do not even have time to write the reviews.All of these cards are in stock at Newegg. I don't even see any reviews on them except for literally like 2 people that bought one. I see youtubers saying stuff like "At $500 this card would have shook the world to it's core!" I say the stupid thing needs to be $299 because it's a 4060.
IDK 60 series can be a real mixed bag. Like 760 was a slowed down 670. 960 was a 760 with lower power draw. 1060 6GB was a 980 with more VRAM. 2060 was a 1080 with less VRAM. 3060 12GB was a 2070 with more VRAM.All of these cards are in stock at Newegg. I don't even see any reviews on them except for literally like 2 people that bought one. I see youtubers saying stuff like "At $500 this card would have shook the world to it's core!" I say the stupid thing needs to be $299 because it's a 4060.
4070 in stock everywhere, multiple models. Seems like we are over the price limit. 4070Ti needs to go back to ~380$ like gtx1070, and rtx4070 even lower than that, or production must be halted So which one will it be?
4070 in stock everywhere, multiple models. Seems like we are over the price limit. 4070Ti needs to go back to ~380$ like gtx1070, and rtx4070 even lower than that, or production must be halted So which one will it be?
Is the story of TSMC refusing to cancel Nvidia wafer orders still true, or has it changed? What you write might be a bit more complex than you imagine.LOL!
Hate to burst your bubble, But no significant price cuts are coming anytime soon.
This is early in the cycle, so it will be pretty easy to match production to sales.
But if you wait a year or two, you might get one for $100 off. I'm sure the two year wait to save $100 will be totally worth it.
Meh a $10 box I'm going to throw away anyways is a waste. I'd rather they just charged me $10 less. Glad Corsair finally stopped giving those stupid Crown Royal looking bags with their RMx power supplies.I gotta say that the packaging is like 1000x improvement, unboxing the 4070 RTX feels like opening a mysterious artifact or relic. Between the super rugged external box and the crazy black Nvidia time capsule, the packaging has to add AT LEAST $10 to BOM. So at least we're getting our moneys worth there.
Congrats on the new card. Had it been available when I did my new build earlier this year It may have been a hard choice vs. the 6900xt i bought for $600.Names are just names, nobody cares. Price and performance matter.
If 4070 is intended to be 2560*1440 card, I got 59% of the performance of the 4090 for 37% price, based on techpowerup average numbers and local lowest prices.
I cannot complain.
BTW compared to RX7000 series this card behaves normally, it has acceptable low power draw in idle, multi monitor and video playback. RX7000 seems to be broken in this area.
I considered this, but discarded the idea. My reasoning was that, first, there is the issue with poll trolls that intentionally sabotage results. Not out of malice or to further some personal agenda. Strictly for the LULZ. Next, it relies on our often imperfect memories. It could have asked but been dismissed so offhandedly, that they made no memory of it.
I would certainly be interested in more response. I know @GodisanAtheist mention his memory is of Nvidia being surveyed more than AMD in his builds too. That'd make 2 of us so far that lean toward there being an issue with how the survey is conducted.
On topic: The interwebz exploded with content for the 4070 today. Pedro does nothing but gameplay, so I enjoyed his video.
No more frilly packaging or useless bags.Meh a $10 box I'm going to throw away anyways is a waste. I'd rather they just charged me $10 less. Glad Corsair finally stopped giving those stupid Crown Royal looking bags with their RMx power supplies.
Congrats on the new card. Had it been available when I did my new build earlier this year It may have been a hard choice vs. the 6900xt i bought for $600.
That has to be about the most flawed value argument I've ever seen though. Price comparing nearly any card to a 4090 in this way is bound to make that card look like a good value. The 4090 has no peers and is priced in a tier all it's own to reflect that.
Is the story of TSMC refusing to cancel Nvidia wafer orders still true, or has it changed? What you write might be a bit more complex than you imagine.