Time to Take Moore's Law is Dead Nvidia Article Seriously...

Wasn't Ampere delayed though? So why rush if you've already set a delay and are still ahead of the competition?

When was Ampere supposed to launch?

If nvidia didn't just plain and simple screwed up, then the only alternative is that they deliberately stifled supply, just to float that £699 number out there long enough until AMD shows up, then they'll drop the £899 cards. I could believe that, and I'm inclined to do so when they could have released these cards for £899 or more and sold exactly as many. I do find it hard to believe nvidia weren't prepared, given they set their own release date with no competition in sight and somehow didn't know the market demand.

I think the 3080 'regular' will be mostly mythical. It'll exist, and stick around, just to mark that £699 price point.

The only way to know how much nvidia screwed up is to know how many cards were available for sale on Sep 17th. I don't think anyone knows that except nvidia.
 
LOL! NVIDIA having AIBs cost down their BOMs to stay close to MSRP at launch causing crappily built cards? It is almost like I have heard this somewhere before. Hmmm.



That's exactly what I was thinking when I heard reports of failure and then seeing the Igor labs report.
 
Not sure Nvidia would ask an AIB to use substandard components. I could see them doing it to maximize their profit, maybe. If they did they are stupid.
 
Not sure Nvidia would ask an AIB to use substandard components. I could see them doing it to maximize their profit, maybe. If they did they are stupid.

I think the point is that the margins are lower than expected to meet certain price points overall with this launch. If an AIB wants to have a higher margin they are going to use substandard parts. The FE cards are going to use normal parts because Nvidia can eat the loss of profit but a company like Zotac can't for example. A company like ASUS is just going to charge whatever they want because they can command it.
 
I think the point is that the margins are lower than expected to meet certain price points overall with this launch. If an AIB wants to have a higher margin they are going to use substandard parts. The FE cards are going to use normal parts because Nvidia can eat the loss of profit but a company like Zotac can't for example. A company like ASUS is just going to charge whatever they want because they can command it.

$10 doesn't make a difference is my point. Making bad stuff with substandard parts is terrible business. Only substandard companies do that kind of thing.
 
$10 * 100,000 units adds up.

It does and you add that cost to your MSRP. If they just plan to cheap out for a little while to maintain some kind of small margin then they are stupid. Hell, we already read that Zotac made their Trinity models subpar. They admitted it. I suspect there might be some XFX like pushback coming.
 
It does and you add that cost to your MSRP. If they just plan to cheap out for a little while to maintain some kind of small margin then they are stupid. Hell, we already read that Zotac made their Trinity models subpar. They admitted it. I suspect there might be some XFX like pushback coming.

They want to keep the MSRP under the magical $699 price point. Otherwise people will just wait for the FE version. Of course they will be waiting for a long time for it because it will never be highly stocked.

Nvidia is screwing their AIB partners on the price in order to look better for the consumer.

Without a comprehensive look at where the failures are, I would guess that a lot of them are on the cheaper end models where this kind of practice is acceptable to meet margins. It's not going to be on their high-end models where MSRP is relative.
 
They want to keep the MSRP under the magical $699 price point. Otherwise people will just wait for the FE version. Of course they will be waiting for a long time for it because it will never be highly stocked.

Nvidia is screwing their AIB partners on the price in order to look better for the consumer.

Meh. We'll have to wait to see if that's true. The AIBs are going to make bank. The demand for these cards is obvious.
 
They want to keep the MSRP under the magical $699 price point. Otherwise people will just wait for the FE version. Of course they will be waiting for a long time for it because it will never be highly stocked.

Nvidia is screwing their AIB partners on the price in order to look better for the consumer.

Without a comprehensive look at where the failures are, I would guess that a lot of them are on the cheaper end models where this kind of practice is acceptable to meet margins. It's not going to be on their high-end models where MSRP is relative.
So bottom line is the AIBs will not be able to make cards at MSRP price, if they did they will not work or perform worst than the FE.
 
Meh. We'll have to wait to see if that's true. The AIBs are going to make bank. The demand for these cards is obvious.

All things being equal, people will buy what is slightly cheaper. Hence it is in their best interest to have a product that is on the razor's edge of price, reliability, and performance.
 
So bottom line is the AIBs will not be able to make cards at MSRP price, if they did they will not work or perform worst than the FE.

I think what you are seeing is AIBs trying to make as much margin as possible. Nvidia left them in a tough spot because they are willing to accept a significantly less margin to meet a specific price point. The average AIB does not want that low of a margin so they are improvising.
 
All things being equal, people will buy what is slightly cheaper. Hence it is in their best interest to have a product that is on the razor's edge of price, reliability, and performance.

Word of mouth is strong. Reliability matters. The AIBs cutting corners are ruining their business.
 
Word of mouth is strong. Reliability matters. The AIBs cutting corners are ruining their business.

Maybe, maybe not. They obviously didn't make cards expecting them to fail. But I would bet you will see a new revision addressing these issues that is slightly more costly to build.
 
Maybe, maybe not. They obviously didn't make cards expecting them to fail. But I would bet you will see a new revision addressing these issues that is slightly more costly to build.

Making a reliable model is the way to go. Who wants to be Yugo?
 
I think what you are seeing is AIBs trying to make as much margin as possible. Nvidia left them in a tough spot because they are willing to accept a significantly less margin to meet a specific price point. The average AIB does not want that low of a margin so they are improvising.
Hard to know, if they can survive with much lower margins or not, if their business requires 30% to keep the doors open, honor returns/repairs marketing, employees and everything else, MSRP maybe out of reach for them to make cards for.
 
Not sure Nvidia would ask an AIB to use substandard components. I could see them doing it to maximize their profit, maybe. If they did they are stupid.

Pretty sure nVidia would ask that, and likely also include a clause to prevent that from being leaked or face penalties, as well as clauses related to supply constraints.

Why do people forget the past so quickly, especially in the tech world? its why most of these companies get away with legendary levels of BS.
 
So, to sum it up, the cheap AIB cards will pretty much be 0 over clocking, knock 5-10% off of the performance numbers to achieve stability, and you have a card below the FE card. And within striking range of a 2080ti
 
Cheaping out on components is fine as long as they still meet minimum spec for reliability and performance. Not every SKU needs to be a champion overclocker but there’s no excuse for shipping broken hardware that can’t maintain stock speeds.
 
So, to sum it up, the cheap AIB cards will pretty much be 0 over clocking, knock 5-10% off of the performance numbers to achieve stability, and you have a card below the FE card. And within striking range of a 2080ti

A 2080Ti and a 3070 in theory.

Speaking of which, I saw a price list, and some of the high end 3070's are going to retail close to $650+. Once you double the memory, it's looking like the 3070 16GB is going to be more than a FE 3080. We'll see if Big Navi helps keep the cost down or not.
 
Cheaping out on components is fine as long as they still meet minimum spec for reliability and performance. Not every SKU needs to be a champion overclocker but there’s no excuse for shipping broken hardware that can’t maintain stock speeds.

Sure, except the review numbers we saw, may represent the real world for a lot of the cheaper cards, once we get a new bios to drop their boost clocks. Lets hope we get some cards re-reviewed
 
Sure, except the review numbers we saw, may represent the real world for a lot of the cheaper cards, once we get a new bios to drop their boost clocks. Lets hope we get some cards re-reviewed

Exactly, in reality, nobody is going to "re-review" the cards, so the Day 1 reviews will stand as gospel.

Honestly, it's playing out closer to how MLID called it. He thought that the cooler would be inferior on the magical "$699 AIB" cards, but it's the capacitor configuration.

In fairness, it's hard to say if the capacitor situation was about keeping costs down or sheer availability. I haven't seen a breakdown of how much difference in price a 4+2 popscap/MLCC configuration is compared to a 5+1 or a 6+0.

Either way, if the ultimate solution is to limit the boost either by driver or firmware, then the Day 1 review isn't going to be an accurate representation of the performance of $699 cards.

IMO, AIBs are going to start to charge more for "properly functioning" cards.
 
I'm willing to bet any type of driver or firmware update to "fix" the issue will get a re-review by multiple sites.

I hope so. I would also hope their "Day 1" review also makes reference to the new updates.
 
I hope so. I would also hope their "Day 1" review also makes reference to the new updates.

They should definitely rerun the numbers but I don’t expect they will change much at all. In reviews most of the cards averaged boost clocks under 2Ghz and performance is mostly limited by bandwidth anyway. I would be surprised if capping clocks at 2Ghz will change stock performance in any measurable way.
 
They should definitely rerun the numbers but I don’t expect they will change much at all. In reviews most of the cards averaged boost clocks under 2Ghz and performance is mostly limited by bandwidth anyway. I would be surprised if capping clocks at 2Ghz will change stock performance in any measurable way.
Guess you would have to go back and re-run every benchmark if you were a real tech journalist to find out. Otherwise, just saying, "No biggy," and moving on is what will likely happen.

You can write it off, but the Day 1 reviews with higher clocks will stand "forever" and be used for buying purposes for the life of the GPU.

MLID was correct again, and you can try to explain it away as no big deal, but NVIDIA will certainly not be seeing it that way.
 
Guess you would have to go back and re-run every benchmark if you were a real tech journalist to find out. Otherwise, just saying, "No biggy," and moving on is what will likely happen.

You can write it off, but the Day 1 reviews with higher clocks will stand "forever" and be used for buying purposes for the life of the GPU.

MLID was correct again, and you can try to explain it away as no big deal, but NVIDIA will certainly not be seeing it that way.

All of that only matters if performance is actually different right? Not sure what part of my post you were responding to.
 
All of that only matters if performance is actually different right? Not sure what part of my post you were responding to.
So clocks are changing but performance is not? Or they are changing clocks, but no big deal? If clocks did not matter, why clock those to failure to begin with?
 
So clocks are changing but performance is not? Or they are changing clocks, but no big deal? If clocks did not matter, why clock those to failure to begin with?

We don’t know if the cap is a big deal until the numbers are run again on updated firmware/drivers. It also depends whether they just limit the top bins or lower clocks for all voltage bins. Much higher chance performance suffers if it’s the latter.
 
We don’t know if the cap is a big deal until the numbers are run again on updated firmware/drivers. It also depends whether they just limit the top bins or lower clocks for all voltage bins. Much higher chance performance suffers if it’s the latter.
Cool with me. NVIDIA got exactly what they wanted out of this. Shining Day 1 reviews which will stand for the entire product cycle. And I think that MLID said that was specifically part of their agenda.
 
Cool with me. NVIDIA got exactly what they wanted out of this. Shining Day 1 reviews which will stand for the entire product cycle. And I think that MLID said that was specifically part of their agenda.

Guess NV is smart?
 
Cool with me. NVIDIA got exactly what they wanted out of this. Shining Day 1 reviews which will stand for the entire product cycle. And I think that MLID said that was specifically part of their agenda.

I don't think those Day 1 reviews will last for the entire product cycle if some major/minor issue comes along that changes the results...most people reading those reviews are pretty knowledgeable as far as hardware/tech which is why the tech community started going crazy with CTD news shortly afterwards...the same people reading Day 1 reviews will follow the latest news on the cards
 
I don't think those Day 1 reviews will last for the entire product cycle if some major/minor issue comes along that changes the results...most people reading those reviews are pretty knowledgeable as far as hardware/tech which is why the tech community started going crazy with CTD news shortly afterwards...the same people reading Day 1 reviews will follow the latest news on the cards
Yeah, you probably know more about the impact of those than I do. What do I know....
 
Back
Top