• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

NVIDIA ‘Ampere’ 8nm Graphics Cards

I really think the 3060Ti is too close to the 3070 for comfort (within 10%), especially when overclocked (within 5%). If I had recently bought a 3070, especially for scalped prices, then I would be seriously annoyed. It even has the same VRAM.

The only card that is a clear jump ahead is the RTX3080, which is in turn only a small amount behind the 3090.

The AMD's get absolutely crushed in RT performance in general, especially with NVDLSS. If AMDLSS isn't released soon then even the 6900XT is going to look bad vs lower Ampere cards when RT is enabled.
Completely agree. Luckily I can return my 3070 for a fiver.
It was going to be a stop gap until next year when I upgrade my monitor. The 3060ti is only about 2 or 3 fps behind the 3070 at 1080p but if I can grab a FE it is £200 cheaper
 
Completely agree. Luckily I can return my 3070 for a fiver.
It was going to be a stop gap until next year when I upgrade my monitor. The 3060ti is only about 2 or 3 fps behind the 3070 at 1080p but if I can grab a FE it is £200 cheaper

The FEs have been best value since this launch started. Nvidia (and AMD to a lesser extent) does not need the AIBs anymore (assuming they could ramp up their own manufacturing). Best value for us as consumers is the FEs, by a country mile.
 
The FEs have been best value since this launch started. Nvidia (and AMD to a lesser extent) does not need the AIBs anymore (assuming they could ramp up their own manufacturing). Best value for us as consumers is the FEs, by a country mile.
Of course Nvidia need AIB's, they can't produce enough cards alone.
 
Completely agree. Luckily I can return my 3070 for a fiver.

you got your 3070 for a fiver ? Bargain. :D

The FEs have been best value since this launch started. Nvidia (and AMD to a lesser extent) does not need the AIBs anymore (assuming they could ramp up their own manufacturing). Best value for us as consumers is the FEs, by a country mile.

of course Nvidia need the AIB's where do you think the AIB's buy their base components/board from ? AMD ? :D Probably make more money off that than selling FE's alone.
 
you got your 3070 for a fiver ? Bargain. :D



of course Nvidia need the AIB's where do you think the AIB's buy their base components/board from ? AMD ? :D Probably make more money off that than selling FE's alone.

Lets be clear, they don’t NEED their AIBs, however, they WANT to keep them around as they absolutely use them to make more profit. The ‘own brand’ cards are sold at a loss, that much is clear. It enables Nvidia and AMD to get an awesome card out in the wild for a lower RRP than is really sustainable. They make up for it with a higher average selling price via AIBs.

They then sells their cores to their AIB partners for a higher profit than they could achieve building their own cards and selling them at RRP.

They effectively squeeze their AIB partners at the low end and force them to sell more expensive cards to actually make a reasonable margin.

Most people can’t buy a card for RRP, therefore the RRP is fake (IMO of course). It’s all about creating an image of an RRP that doesn’t actually exist.

TLDR: They could absolutely bin off their partners tomorrow and use a contract manufacturer like Foxconn to built every single card. But to do that they would have to increase the RRP from what it is as it isn’t sustainable to sell those cards at that price in volume. It’s all business and clever marketing and it works.
 
Of course Nvidia need AIB's, they can't produce enough cards alone.

of course Nvidia need the AIB's where do you think the AIB's buy their base components/board from ? AMD ? :D Probably make more money off that than selling FE's alone.

If they wanted to, they could produce their own products completely. Facebook did it with the Quest 2, Microsoft and Sony do it with the consoles. Why couldn't nvidia manufacture their own chips and just use them in their own products as a spin off business? Manufacturing capacity can be bought, that isn't the issue.


Lets be clear, they don’t NEED their AIBs, however, they WANT to keep them around as they absolutely use them to make more profit. The ‘own brand’ cards are sold at a loss, that much is clear. It enables Nvidia and AMD to get an awesome card out in the wild for a lower RRP than is really sustainable. They make up for it with a higher average selling price via AIBs.

We don't know that the FE/Ref cards are sold at a loss because we don't know the cost to build. Aside from the GPU & ram chips themselves, all the other components are run of the mill electronics. We dont therefore know what the rrp's would be if nvidia/amd chose the foxconn business model as you suggest in your TLDR.

The reason, in my view, that AIBs are more expensive is that they are marking the chips up twice. Nvidia sell the chips at a profit to the AIBs, who then have to add on a 2nd mark up to make a profit for themselves. If nvidia/amd made them themselves, this 2nd mark up wouldn't exist. There would be a cost to the volume manufacturing, which might be a lot less than the AIBs cost, because nvidia would be making more of the same sku in their own/contracted factories.

The quality of the FEs this time, from nvidia especially, indicate perhaps their foray into the AIBs market share.
 
If they wanted to, they could produce their own products completely. Facebook did it with the Quest 2, Microsoft and Sony do it with the consoles. Why couldn't nvidia manufacture their own chips and just use them in their own products as a spin off business? Manufacturing capacity can be bought, that isn't the issue.
Maybe I should have qualified my statement with: "without radically changing their business model".

If you were Nvidia then why would you not want to just sell the chips and avoid all of the rest of the manufacturing and logistical hassle?
 
We don't know that the FE/Ref cards are sold at a loss because we don't know the cost to build. Aside from the GPU & ram chips themselves, all the other components are run of the mill electronics. We dont therefore know what the rrp's would be if nvidia/amd chose the foxconn business model as you suggest in your TLDR.

The reason, in my view, that AIBs are more expensive is that they are marking the chips up twice. Nvidia sell the chips at a profit to the AIBs, who then have to add on a 2nd mark up to make a profit for themselves. If nvidia/amd made them themselves, this 2nd mark up wouldn't exist. There would be a cost to the volume manufacturing, which might be a lot less than the AIBs cost, because nvidia would be making more of the same sku in their own/contracted factories.

The quality of the FEs this time, from nvidia especially, indicate perhaps their foray into the AIBs market share.

At a loss may be going too far but it certainly isnt far from the truth. What is clear they are certainly taking a hit on both the cooler and PCB design on these FE cards to achieve the RRP. The cooler is obviously more expensive than you get on a basic AIB card and the PCB design is far more complex and uses more expensive components.

It's widely reported now that an AIB just can't make a card at RRP (certainly not using a reference design) and make a profit at the same time. That's using a cheaper cooler and board design than what is used on the FE cards. The FE card is a loss leader assuming an arms length cost for the parts they sell to AIB partners, you can deduce that Nvidia makes less money from selling a FE than they would just selling the core to an AIB. The FE card costs more to produce and they sell it for less, therefore Nvidia couldn't cut out their AIB partners and sell their ampare cards for RRP and still make the same profit. AIB's only make a profit because people pay £200+ more for a 'STIX OC' for a marginal gain in performance and RGB.

Maybe I should have qualified my statement with: "without radically changing their business model".

If you were Nvidia then why would you not want to just sell the chips and avoid all of the rest of the manufacturing hassle?

I've already explained it, its about creating the illusion of an RRP which isnt actually available to most people. Just look at reviews, they don't compare value or price/performance at the average selling price. They all use the RRP which you could argue isn't real, the average selling price will be way higher (ignoring the gouging by retailers). Hardly anything has an average selling price above RRP, normally things are 'discounted' below it.

It's not a hassle to manufacture it, you use a contract manufacturer like Foxconn and they do all the hard work. They already do that for the FE.

To be clear, the AMD are doing the exact same thing with their 'limited' RRP cards.
 
Wasn't the cooler meant to be $150 according to that leak in the summer. Makes you think MSI and Gigabyte selling 3080s for £850 to £950 on a certain website. Making a lot of profit?
 
We don't know that the FE/Ref cards are sold at a loss because we don't know the cost to build. Aside from the GPU & ram chips themselves, all the other components are run of the mill electronics. We dont therefore know what the rrp's would be if nvidia/amd chose the foxconn business model as you suggest in your TLDR.

Some stuff like the VRAM we can get the price at retail for the modules but what nVidia, etc. actually pay is another matter.
 
Some stuff like the VRAM we can get the price at retail for the modules but what nVidia, etc. actually pay is another matter.

There's no way they pay the same prices as we pay - the end consumer. Factor in bulk discounts and mates rates, strip clubs for execs etc. ;)
 
Wasn't the cooler meant to be $150 according to that leak in the summer.

Who knows. I don't really believe that. If volume manufacturing, the costs would come down.


If you think about it, GPUs are a really strange product. Its the only product I can think of where the main component (the chip) overshadows the product itself (the GPU package).

We all know the XR2 chip was used in the Quest 2, and that's a great selling point, but the product as a whole is still the focus. The product is far more than just the chip.

We all know the RDNA2 architecture was used in the consoles, and again that's a selling point to those in the know, but the product as a whole still takes focus and a lot of people buying it don't care what the chip is.

We don't fuss about what chips are used in high end TVs, or in phones. We don't (generally) fuss about what engine is used in a car, as long as it meets the criteria we set for reliability, power etc.



And because the focus of the entire device is really just the chip, then who cares who makes the pcb (as long as it works properly). What are the AIBs really adding here? Are they adding value to the product? I'd argue not a lot. All they do is package the device into an interface. Anyone can do that, including nvidia's/AMD's own contracted factories.
 
Back
Top Bottom