When the Gpu's prices will go down ?

[andybird123]

Sitting on stock doesn't negatively affect margins - stock counts as an asset.

If you've incurred the cost of creating stock but it's not selling then your costs are higher and your revenue and profit is lower decreasing your effective margin for that period
A company's margin is the percentage difference between costs and revenue (profit), but to have that ratio you actually have to have the revenue - stock isn't revenue
If they end up having to slash prices in order to shift stock then yes it very much does affect margins on those products

 

[Zarax]

Or.. the best experience is toward the latter phase of the phones lifecycle where all the bugs have been ironed out and the device is on offer.

Can we apply that to GPUs though? Let's say an almost EOL 6700XT (best price/performance right now IMHO).
Would it be a good investment VS Navi 32?

 

[Troezar]

It's becoming same as phones. Spend $250 every 2-3 years and you'll likely have a better experience over time than a $1000 phone.

Indeed, I usually go £400 (OnePlus 7T) currently and run it until it dies. I'd be happy to pay less and may well do that next time.

 

[lsg1r]

The 3090 was never good value, so 71% better for $100 more doesn't make the 4090 good value.

unless you used it for content creation, it was a LOT cheaper than a 'quadro' or whatever nvidia's renamed them to these days lol, the 4090 is arguably even better in that regard....

 

[Troezar]

unless you used it for content creation, it was a LOT cheaper than a 'quadro' or whatever nvidia's renamed them to these days lol, the 4090 is arguably even better in that regard....

That's where some of the "arguments" come from. One product but with owners having different uses, for some it's "cheap", to others it's questionable value because their needs are not the same.

 

[Zarax]

Indeed, I usually go £400 (OnePlus 7T) currently and run it until it dies. I'd be happy to pay less and may well do that next time.

My last 3 phones:

- Ulefone Armor 6e: 3 years for 200€
- Poco X3 Pro: 3 years for 200€
- Ulefone Armor 12s: Should be another 2-3 years for 250€

 

[lsg1r]

That's where some of the "arguments" come from. One product but with owners having different uses, for some it's "cheap", to others it's questionable value because their needs are not the same.

It's also where sales numbers and the likes can be 'skewed' because it will fall in the gaming sales figures but the largest portion of buyers is more likely to be smaller size content creation businesses saving themselves thousands....

 

[gpuerrilla]

Can we apply that to GPUs though? Let's say an almost EOL 6700XT (best price/performance right now IMHO).
Would it be a good investment VS Navi 32?

Your example answers it as there are no Navi32 available so unless they are coming out very soon, why not? As nvidia/AMD went high end first this gen its scuppered the usual release patterns. Either way if your going to spend every 2-3 years, Jensen will have no qualms milking your cow solution. The issue so far has been highlighted that the lower in the stack the worse improvements seem to be.

 

[Joxeon]

The 3090 was never good value, so 71% better for $100 more doesn't make the 4090 good value. In terms of performance per dollar progression, the 3080 is the most relevant high end card for comparison. I only checked one review which had it at 20.8%, but even at 17.6% the 4070 Ti still better performance per dollar at 4K with an extra 2GB VRAM, better power efficiency and AV1 encoding.

The 3090 was 12% faster than a 3080 for 115% more money. The 4090 is 67% faster than a 4070ti for 100% more money and remember the halo card isn't supposed to have good value scaling compared to the mid range yet the 4090 actually looks very good up against a 4070ti.

 

[Eddie99]

The 3090 was 12% faster than a 3080 for 115% more money. The 4090 is 67% faster than a 4070ti for 100% more money and remember the halo card isn't supposed to have good value scaling compared to the mid range yet the 4090 actually looks very good up against a 4070ti.

Cheapest 4090 you can actually buy is £1700, cheapest 4070 Ti is £800. How exactly does the 4090 look very good against a 4070 if it's 67% more performance for 112% more money? It's substantially worse performance per dollar.

 

[kevman]

Plus 3080 was not available for a large chunk of the 30 series lifespan. Many including me bought a 3090 as it was not possible to pick an 80 series card in that first year or 2

 

[gpuerrilla]

Lets be honest most of the lifetime of a 3080 they were selling for between 800 and 1200. You had to stick with FE drops for any value, there were some early AIB lucky buyers but then there was queue-gate.

 

[lsg1r]

Cheapest 4090 you can actually buy is £1700, cheapest 4070 Ti is £800. How exactly does the 4090 look very good against a 4070 if it's 67% more performance for 112% more money? It's substantially worse performance per dollar.

You know as well as the rest of us that the higher up the stack the performance per pound decreases... it's not a simple straight line. There's more silicon being used, probably lower yield too, there's more ram and the cooler and other hardware is suitable for a higher power usage etc... it all adds cost and it isn't a linear cost. It's also not just about gaming performance for some of us, there are things I could do with a 4090 that a 4070ti wouldn't be able to do due to the fact it has less memory, the software would literally go, sorry no can do not enough vram...

We shouldn't even be arguing that the 4070ti SHOULD be better value than the 4090 in a current line up, it's not in the same tier of cards for starters, still doesn't stop the 4070ti being overpriced relative to previous generations...

At the same time no one seems to be taking into account the potential life span of the 4090 versus the 4070ti, we've already got games which are pushing the limits of 10-12GB vram cards (arguably due to poor optimisation/conversion from console ports) so it's not unrealistic to potentially see 16GB vram being a 'recommended' spec within the next few years.

 

[gpuerrilla]

I wouldn't talk about vram @lsg1r we dont need more threads locked.

 

[tommybhoy]

Plus 3080 was not available for a large chunk of the 30 series lifespan. Many including me bought a 3090 as it was not possible to pick an 80 series card in that first year or 2

Lets be honest most of the lifetime of a 3080 they were selling for between 800 and 1200. You had to stick with FE drops for any value, there were some early AIB lucky buyers but then there was queue-gate.

I checked mid Feb this year, there were four 3080FE drops since 17th Dec 2021, yes four, 80FE was only on sale something like 13 times since launch, or been on general sale for less than an hour since launch.

 

[Purgatory]

You know as well as the rest of us that the higher up the stack the performance per pound decreases... it's not a simple straight line. There's more silicon being used, probably lower yield too, there's more ram and the cooler and other hardware is suitable for a higher power usage etc... it all adds cost and it isn't a linear cost. It's also not just about gaming performance for some of us, there are things I could do with a 4090 that a 4070ti wouldn't be able to do due to the fact it has less memory, the software would literally go, sorry no can do not enough vram...

We shouldn't even be arguing that the 4070ti SHOULD be better value than the 4090 in a current line up, it's not in the same tier of cards for starters, still doesn't stop the 4070ti being overpriced relative to previous generations...

At the same time no one seems to be taking into account the potential life span of the 4090 versus the 4070ti, we've already got games which are pushing the limits of 10-12GB vram cards (arguably due to poor optimisation/conversion from console ports) so it's not unrealistic to potentially see 16GB vram being a 'recommended' spec within the next few years.

The 4090 is not as great as people think either, especially for people using them for work or creative apps. The problem that many seem to forget is the 4090 is an actual downgrade from a 3090/ti as I keep repeating there is no NVLINK and that means no dual card support for applications that require NVLINK or more vram than 24GB, with NVLINK I get 48GB and double the cuda cores pooled for my apps.

4090 sadly is not an upgrade path for me and people in my situation and only update is the Ampere A-series that have NVLINK and more VRAM than 24GB, even ADA A-series has only 48GB max sized card and no NVLINK..

Nvidia has made a real mess for many of us using them for pro work and they deliberately did that after seeing how the 3090s sold to pro users in NVLINK setups and why they tried to ban blower 3090s too because they don't want them used in servers or small setup workstations as they want you to buy A-series (Quadro) cards..

Lets hope NVLINK comes back to 50xx series but sadly I doubt it as only cards with it now are Pro gpus for the server side and even their workstation cards this generation have it removed and they are trying to sell the Ampere A-series to people that need NVLINK and more VRAM than 48GB.

4090 is nothing more than a 4080 really now and really what it should be sold as and the 4090ti if one comes out as the 4080ti.. but sadly that's not how it played out this generation and Nvidia is really damaging its product line to gamers and pro users, they want to change the whole product stack and rewrite their whole product catalogue and of course the consumer is the one losing out.

 

[lsg1r]

snip

NVLink is being or even has been phased out, even on 'quadro' (sorry can't get used to calling them A series) and this is coming from nvidia. Something to do with the bandwidth of pcie5 removing the need for nvlink and it can all be done via pcie instead, it's supposedly there in ada.

Last I checked you could share ram without nvlink, not personally had a need for that so not fully up to speed.

 

[Purgatory]

Lets be honest most of the lifetime of a 3080 they were selling for between 800 and 1200. You had to stick with FE drops for any value, there were some early AIB lucky buyers but then there was queue-gate.

Lucky to even get one as £1200.. I was after a 3080 for HTPC back then and never could buy an FE when they showed up and only options were very overpriced AIB 3080s and that was never going to happen when 3090s at one point were only a little more to buy than the AIB 3080s.. and even then you had to be lucky to get a 3090 at a "sensible" price compared to the 3080 AIBs.

 

[Purgatory]

NVLink is being or even has been phased out, even on 'quadro' (sorry can't get used to calling them A series) and this is coming from nvidia. Something to do with the bandwidth of pcie5 removing the need for nvlink and it can all be done via pcie instead, it's supposedly there in ada.

Last I checked you could share ram without nvlink, not personally had a need for that so not fully up to speed.

It's not being phased out, it's being only added to server gpus and very high-end GPUS this generation and being sold as a high-end feature only now. PCIe5 can't do what NVLINK does so far and there is no silicon on ADA for any form of NVLINK functions and can't share VRAM or pool CUDA cores.

That's why they have the Ampere A-series for that still being made and sold for them features, check Nvidia's pro line up and site regarding all this.

In short NVLINK has now become a luxury feature for their top of the line products and guessing when they phase out Ampere A-series they will have another gpu to take that role on once 48GB VRAM becomes more the norm on pro cards and by then 96GB cards with NVLINK will be about I'm guessing and hoping they may bring out 48GB 5090/6090 by then with NVLINK again.

 

[t2na]

Lucky to even get one as £1200.. I was after a 3080 for HTPC back then and never could buy an FE when they showed up and only options were very overpriced AIB 3080s and that was never going to happen when 3090s at one point were only a little more to buy than the AIB 3080s.. and even then you had to be lucky to get a 3090 at a "sensible" price compared to the 3080 AIBs.

I managed to get a 3070 FE relatively easily but they were a little easier to grab as most people were so focussed on the 3080's during the drops - and then the 3060Ti's. The 3080 FE's would sell so fast you couldn't even refresh the page.