$700-800 Ideal Price for GeForce RTX 4080: TechPowerUp Poll Surveying 11,000 Respondents

[Godrilla]

The last poll I recollect with Ampere eventually got us the pricing we still see today. The only way we will see these prices if AMD can supply the demand and Nvidia will have no choice but to lower prices.

 

[the54thvoid]

It's almost arrogant what Nvidia did in terms of the pricing for the 4080. Given there are 4080 AIB cards that reach to the 4090FE price, it's an absolute joke that anybody would buy one (at $1500-$1600), although Nvidia only need to restrict 4090 supply to drive people to buy the worse option at the higher price point. It is truly staggering.

It's fine to spend your money on what you want but for the perf diff between the two card, I figure people willing to spend 1300 bucks would probably be just as happy jumping for the 4090.

Then again, as has been mentioned, the 4080 is the xx80 model. It's not the xx90, xx90ti, or even the xx80ti. It's the enthiusiast card, NOT the halo product. I'm happy to see so many sitting on the shelves on the digital storefronts. But... one last thought - is Nvidia really just holding off to clear the 3000 series because they are disappearing?

 

[trog100]

i spent £1100 quid on a 2080ti.. before that i had two 1070 cards in SLI mode and before that a pair of 980ti cards in sli mode.. before that it was a pair of 970 cards in sli mode..

high end or near high end has never been cheap to me for a long time.. nvidia got rid of sli on mid range cards for a reason.. so that they double the price for a high end card as opposed to buying two cheaper cards in sli mode..

trog

 

[defaultluser]

For this stupid 1200$ price they sell current stocks 4080 in 2-4year

yeah- that or else we get a price cut 6 months from now

 

[the54thvoid]

nvidia got rid of sli on mid range cards for a reason.. so that they double the price for a high end card as opposed to buying two cheaper cards in sli mode..

trog

What are you smoking? SLI disappeared because it was a driver-tastic nightmare and relied too heavily on developers. Nvidia didn't drop SLI to charge more for single cards - that's the worst apologist excuse I've ever heard. Are you trolling?

 

[Space Lynx]

I voted $900 mainly due to inflation correction, i'd say between 800 and 900 is fair.

I agree with this.

 

[evernessince]

If you factor in inflation, $800 seems more like it. GTX780 released in 2013 was around that price (in terms of buying power) as well.

Until you look at the 4080's die size, which is significantly smaller than the 780 or 3080 for that matter. $800 is being generous to Nvidia. The fact that they are asking $1,200 is frankly disgusting.

The poll poses a trick question - because it does not ask users what to them is a realistic price, but what would they pay for an RTX 4080. To that? 400 USD is a perfectly viable answer, even if NVIDIA actually loses money on selling one.

GPUs have a lot of markup, particularly at the high end. What's "realistic" varies a lot due to that. Anything above the BOM is possible. Often when I see people use the world realistic, it's to impose their own idea of realistic on others. People have more power to change reality than they think.

I just gave in and bought an Asus tuf 3070 v2 for $585. Mainly due to the fact my youngest daughter needed a new gpu. I gave her my old 2070 which worked great for 1440p @144hz gaming on it. I was trying to hold out for the 4070 but this card dropped below 600 bucks and I jumped on it. I will not pay the increased prices for the 4xxx cards. These prices are crazy for GPU's. The 4080 should be back to $699 due to ethereum becoming POS.

Holy smokes that price is a ripoff. You could have gotten a 6900 XT for $20 more or a 6800 XT for $520 (AsRock model). More raster performance in either case and equal RT performance with the 6900 XT. Last gen AMD cards have been dropping in price much faster than Nvidia.

This is part of the reason Nvidia can charge whatever it wants, people buy them regardless. Nvidia's pricing of the 4000 series isn't dumb from a business perspective because this will clearly continue.

We'll see how stubborn nVIDIA will insist on it's current pricing when 7900XTX/XT launches. If AMD is at least somewhat competitive I doubt the price will not be lowered - unless nVIDIA and their shareholders are fine with cards sitting on shelves.

I really don't think AMD's pricing is that good either, especially the $900 7900 XT. AMD just did a demonstration on how much cheaper chiplets make their chips and yet they jacked up the price of their 2nd best GPU and priced the much smaller 7900 XTX the same as the 6900 XT. AMD isn't going to start a price war with Nvidia, they are joining right in.

If Nvidia cards are sitting on shelves it's because people literally can't pay. As evidenced time and time again, people will buy Nvidia even when they are getting significantly less.

i spent £1100 quid on a 2080ti.. before that i had two 1070 cards in SLI mode and before that a pair of 980ti cards in sli mode.. before that it was a pair of 970 cards in sli mode..

high end or near high end has never been cheap to me for a long time.. nvidia got rid of sli on mid range cards for a reason.. so that they double the price for a high end card as opposed to buying two cheaper cards in sli mode..

trog

Most people don't have an expensive flagship card. The problem lies in the fact that price hikes have been pushed down the entire stack, often in dramatic fashion like the 4080.

 

[Mindweaver]

Holy smokes that price is a ripoff. You could have gotten a 6900 XT for $20 more or a 6800 XT for $520 (AsRock model). More raster performance in either case and equal RT performance with the 6900 XT. Last gen AMD cards have been dropping in price much faster than Nvidia.

This is part of the reason Nvidia can charge whatever it wants, people buy them regardless. Nvidia's pricing of the 4000 series isn't dumb from a business perspective because this will clearly continue.

Yeah but my decision was mainly due to the better performance in VR the 3070 has over the 6800/6900. I did look at those cards as well. Also, I forgot to add I had a 60 dollar instant rebate that made the price $525.84. It's more than I wanted to pay but I don't think I did bad. The price is back up to 650. Clearly Nvidia's pricing isn't going to continue because prices are coming down, albeit not fast enough. lol

 

[Vayra86]

The poll poses a trick question - because it does not ask users what to them is a realistic price, but what would they pay for an RTX 4080. To that? 400 USD is a perfectly viable answer, even if NVIDIA actually loses money on selling one.

Interpretation I think, or semantics - a realistic price is one the market 'will bear'. After all, if people say its unrealistic, people consider it a bad deal.

This is an example of how marketing can have such a huge impact on perception. If Nvidia was able to sell past x80 models at $500-$800 and still make well over 50% margins, then the 4080 must include more silicon features to require an almost doubling of price to maintain margins.

If this is the case then the 4080 is something different and not directly comparable to past generations. But Nvidia doesn’t change the model name (Geforce) or model numbering (4080).

Thats a somewhat cowardly move by Nvidia’s marketing department showing very little initiative to sell something new. AMD is also at fault for not changing its GPU naming with the introduction of RDNA (although they did change their CPUs to Ryzen/Epyc from Athlon/Opteron/FX).

Let that be a lesson to you all. If you change something so significantly that it costs more, you need to change the name and market it differently. Otherwise past product comparisons will be brutal.

I don't understand what you're saying I think. Nvidia isn't selling something new, right? It's just an x80 that has an impressive performance gap to the next card up the stack. In that sense, its not even a great x80. Nvidia's marketing department would have balls of steel if they tried to sell this as something it's not; they have experience with that, and even just a small lie about a number on a spec sheet (4GB) was enough for an outrage.

Hell, they even did try to sell an x80 as something it's not, the 12GB version. That didn't go quite so well for them either, and that's exactly the same thing: interpretation of that marketing backfired massively - the specs didn't align with the model number.

 

[Fleurious]

Unless that’s Canadian dollars it’s still too much.

 

[mab1376]

I remember with the ATI Radeon 9800 XT came out and thought "who would ever spend that much on a video card alone?"

 

[evernessince]

Yeah but my decision was mainly due to the better performance in VR the 3070 has over the 6800/6900. I did look at those cards as well. Also, I forgot to add I had a 60 dollar instant rebate that made the price $525.84. It's more than I wanted but I don't think I did bad. The price is back up to 650. Clearly Nvidia's pricing isn't going to continue because prices are coming down, albeit not fast enough. lol

Well the thing is the 3070 has worse performance in VR as well:

VR Wars: the RTX 3070 vs. the RTX 2080 Ti – FCAT VR Performance benchmarked

VR Wars: Ampere vs Turing – the RTX 3070 vs. the RTX 2080 Ti – Performance benchmarked Using FCAT VR & the Vive Pro This ... Read more

babeltechreviews.com

VR Wars: The Red Devil RTX 6900 XT versus the RTX 3090 Founders Edition (Part 2)

VR Wars: The Red Devil RX 6900 XT vs. the RTX 3090 FE – 15 Games Performance benchmarked using the Vive Pro This review presents ... Read more

babeltechreviews.com

The 6900 XT wins by a large margin. Aside from the poor performance you are getting with the 3070, it's small VRAM size is going to be a restricting factor given VR tends to consume more of it. I really don't see a positive here, you are forcing yourself into another upgrade in short order.

 

[ARF]

Fun fact - If you were to apply "GPU's should cost +60% more because they are +60% faster than last gen" consistently from nVidia's early RIVA TNT2's through to today, then today's average GPU would cost $3m. Likewise, RAM also used to cost £100/MB at once point. Using same "nVidia pricing logic", 2x 16GB sticks of DDR4 'should' cost £3.2m. We should also be thankful that nVidia doesn't make storage devices or this chart would be a flat horizontal line that never dipped below $1m per GB, at which point modern 1TB SSD's would cost $1bn each...

Back in the real world, the whole point of progress is "tech gets better at same price", not "price ends up completely divorced from reality because 'tech improved as expected'".

This is correct but AMD is also to blame because of lack of initiative to put the price back to their normals:

1. We are in a deep recession post-pandemics and energy transition.
2. Consumers are getting poorer by the day, so no more money for not vital purchases.
3. PC market in all segments is in an astonishing downward spiralling.

Radeon prices today - fluctuating like nothing happens and the market is in a normal situation which is extremely weird and unnatural.
Euro:

Wrong prices:
Radeon RX 6400 - 133.99
Radeon RX 6500 XT - 170.00
Radeon RX 6600 - 279.00
Radeon RX 6600 XT - 345.49
Radeon RX 6650 XT - 329.00
Radeon RX 6700 XT - 399.99
Radeon RX 6750 XT - 482.86
Radeon RX 6800 - 549.00
Radeon RX 6800 XT - 662.92
Radeon RX 6900 XT - 749.00
Radeon RX 6950 XT - 849.00

Correct prices:
Radeon RX 6400 - 89.99
Radeon RX 6500 XT - 119.00
Radeon RX 6600 - 189.00
Radeon RX 6600 XT - 225.49
Radeon RX 6650 XT - 229.00
Radeon RX 6700 XT - 299.99
Radeon RX 6750 XT - 362.86
Radeon RX 6800 - 419.00
Radeon RX 6800 XT - 479.92
Radeon RX 6900 XT - 559.00
Radeon RX 6950 XT - 629.00

 

[Vayra86]

What are you smoking? SLI disappeared because it was a driver-tastic nightmare and relied too heavily on developers. Nvidia didn't drop SLI to charge more for single cards - that's the worst apologist excuse I've ever heard. Are you trolling?

Is that so, and a fact? I would say SLI disappeared for many more reasons... bit of a double edged blade imho. The driver nightmare is a fact, of course; money went into that too. For developers, the same thing applies in that sense. (Now they got RT in return ) On the technical side, we've seen a battle over latency and high refresh is a market demand just the same as perfect frame pacing. SLI doesn't fit in here. And what about Gsync...

But at the same time, the absence of SLI in a world where API's can do mGPU makes not quite a lot of sense, except when you use a perspective of maximizing profit; after all:
- the performance cap is now fixed to a single card as opposed to multiple, AND
- you're working to reduce the pressure on memory capacity to get performance (since Turing-), AND
- you consider that the overall performance level of cards is enough to make people actually move down the stack for decent gaming

Seems like a perfect world to keep buying two cards with better perf/$ for great performance. We've seen Nvidia move the way it does the past three to four generations. They've adjusted their strategy, their product stack, their time to market between generations and their grasp on AIBs and selling their own FE's. All of this supports the idea that they certainly did apply some strategy and SLI removal fits in nicely with increased margins - do less to get more. Its also an absolute fact that to get maximum performance, you need to move higher up the stack, destroying perf/$.

 

[Deleted member 120557]

April 2017 - today
Ryzen 5 1600 - 220$
Ryzen 5 7600 - 229$
with inflation it should be 267$

May 2016 - today
Nvidia 1080 - 599$
Nvidia 4080 - 1199$
with inflation it should be 743$

It certainly has nothing to do with Moore's Law or inflation. Stop using inflation as an excuse

First off, I will say I think the 4080 is overpriced, but people are too kind to the 10 series. All the die sizes were tiny relative to their name. The 1080 was 314mm2. The 4080 is actually smaller than I thought at 378 but the wafer cost TSMC is charging for 5nm is 2-3x what 16nm was. Moore’s law is effectively dead because TSMC is charging what the market will bear.

Plus, the 1600 to 7600 is a bad example. Since AMD uses chiplets now the expensive CPU portion is only ~70mm2 compared to a much bigger 213 for the 1600 though the io die being 120 makes it much closer in total. AMD is still taking a margin cut over their 1600 most likely.

 

[ARF]

First off, I will say I think the 4080 is overpriced, but people are too kind to the 10 series. All the die sizes were tiny relative to their name. The 1080 was 314mm2. The 4080 is actually smaller than I thought at 378 but the wafer cost TSMC is charging for 5nm is 2-3x what 16nm was. Moore’s law is effectively dead because TSMC is charging what the market will bear.

Plus, the 1600 to 7600 is a bad example. Since AMD uses chiplets now the expensive CPU portion is only ~70mm2 compared to a much bigger 213 for the 1600 though the io die being 120 makes it much closer in total. AMD is still taking a margin cut over their 1600 most likely.

GTX 1080 started at 599$, how much of that is the cost of the chip itself? 20%? 30%?
RTX 4080 is a 20% larger chip at 100% higher cost per chip.

So, its contribution to the bill of materials in that asking 1200$+ is mere 20-30%, too.

 

[Mindweaver]

Well the thing is the 3070 has worse performance in VR as well:

VR Wars: the RTX 3070 vs. the RTX 2080 Ti – FCAT VR Performance benchmarked

VR Wars: Ampere vs Turing – the RTX 3070 vs. the RTX 2080 Ti – Performance benchmarked Using FCAT VR & the Vive Pro This ... Read more

babeltechreviews.com

VR Wars: The Red Devil RTX 6900 XT versus the RTX 3090 Founders Edition (Part 2)

VR Wars: The Red Devil RX 6900 XT vs. the RTX 3090 FE – 15 Games Performance benchmarked using the Vive Pro This review presents ... Read more

babeltechreviews.com

The 6900 XT wins by a large margin.

The 6900 XT was out of my budget. The biggest thing now is that it's done and I'm happy with my card. I only paid 26 bucks over MSRP that pretty much matches the performance of the 3070 Ti. Plus, the big picture here is that I used my wife's money so in reality I got a free 3070. Just to be clear you have made very valid points that I did look at before hand.. I just could not find a 6900 xt for $525.84. Let's not keep derailing the thread over my poor GPU decision making.. I do miss the good ol days when you were with your people and they were just excited you got a new piece of hardware.. lol Now kick me and send me a PM telling me about how E Cores are crap... hehehe j/k don't send me that pm.. I'm happy with my mediocre gaming pc..

 

[Vayra86]

First off, I will say I think the 4080 is overpriced, but people are too kind to the 10 series. All the die sizes were tiny relative to their name. The 1080 was 314mm2. The 4080 is actually smaller than I thought at 378 but the wafer cost TSMC is charging for 5nm is 2-3x what 16nm was. Moore’s law is effectively dead because TSMC is charging what the market will bear.

Plus, the 1600 to 7600 is a bad example. Since AMD uses chiplets now the expensive CPU portion is only ~70mm2 compared to a much bigger 213 for the 1600 though the io die being 120 makes it much closer in total. AMD is still taking a margin cut over their 1600 most likely.

Absolutely the die was smaller! But that's entirely the beauty of Pascal. It did so much more with so little. That's a shrink The Way It's Meant to be Played. Part of that is also that Nvidia had been stuck on 28nm for só long.

Today, a shrink enables an immediate maxing out of the silicon and then it is still not enough, so we need retarded power targets.

 

[ARF]

Absolutely! But that's entirely the beauty of Pascal. It did so much more with so little. That's a shrink The Way It's Meant to be Played. Part of that is also that Nvidia had been stuck on 28nm for só long.

Today, a shrink enables an immediate maxing out of the silicon and then it is still not enough, so we need retarded power targets.

The retarded power targets are an "innovation" to maximise the profits with as little as possible physical input aka materials

 

[Vayra86]

The retarded power targets are an "innovation" to maximise the profits with as little as possible physical input aka materials

No they are not, the 4090 is 608 sq/mm just about like the 28nm Titan, heck its even larger.

Although, as little as possible materials, sure, if you look at the 12VHPWR adapter lol

 

[ARF]

No they are not, the 4090 is 608 sq/mm just about like the 28nm Titan, heck its even larger.

8N Turing were larger - 754 sq. mm second tier chip, 545 sq. mm third tier chip, 445 sq. mm forth tier chip, 284 sq. mm fifth tier chip.

608 sq. mm for nvidia is a relatively small chip.

 

[gffermari]

The question is fine and it didn't need any explanation.
Obviously the creator asked about what the reasonable price of a 4080 would be, not actually meant how much would you pay for that.
No one cares how much you would pay for a gpu.

Below 500 and over 1200 answers are ridiculous.

 

[ARF]

Below 500 and over 1200 answers are ridiculous.

If so, why did I buy the mighty Radeon HD 4890 top of the line GPU for only $195 brand new back in summer 2009?

Look, if nvidia says " you pay as much as we wish", you answer: "F**k you, nvidia!" and buy something else for a normal asking price

 

[Vayra86]

8N Turing were larger - 754 sq. mm second tier chip, 545 sq. mm third tier chip, 445 sq. mm forth tier chip, 284 sq. mm fifth tier chip.

608 sq. mm for nvidia is a relatively small chip.

Right, I suppose your definition of relative is different from mine.

 

[gffermari]

If so, why did I buy the mighty Radeon HD 4890 top of the line GPU for only $195 brand new back in summer 2009?

None of the last 4-5-6 gens of nvidia gpus have been less than 500$.
AMDs gpu prices are irrelevant.