whilst that can work for luxury items i am not sure what peoples boss would say if they were told, "sorry boss i am not coming in today because i am not using the train and i cant fill up with fuel!"

there are people in power who can maybe force the issue sometimes, like bringing the entire sector of a country to its knees by holding employees jobs to ransom until they get money for a certain wall to be built.... but not many of us have that power.

 

No, because your boss would be at home too

 

Don't fill up? I think there's something to be said for the French approach

 

del

 

Is it just me who see posts like this and is way more interested in it than the posts I CAN see

 

Lol it was probably about the 1080 ti anyway

 

But arguably not the money. Besides, time alone means nothing. Intel have had more than enough time to get their 10nm sorted out, but they still haven't. Sometimes you just hit a massive hurdle that takes forever to work out.

Well yeah, of course it's a design choice because Radeon 7 isn't a new product. It's a repurposed MI50 to serve as a short term PR stunt. Nvidia put out a stupid price for the 2080, AMD saw that the MI50 with some clock tweaks could get about the same raster performance so reclaimed some defective dies or overstock to have a quick slice of that overpriced pie and to keep the "first to 7nm" train rolling.

The "design choice" comes with Navi. The "design choice" comes with Arcturus.

If AMD still can't get it right next year with Arcturus then complaints are valid, because they will have had both time and money to "get it right".

 

The card was 1 year old already Ofc it was cheap as it was surpassed by a newer model......

 

AMD Radeon VII is not a proper chip. Is failed (cut down) GPU that cannot be used on MI60 or MI50 products.
And this failure appears AFTER the HBM is strapped on the substrate, interposers etc. So it cannot be reversed back and remove HBM etc.

So for AMD the question is do they throw $500+ into the trash bin, or salvage it by strapping a heatsink and sell it recouping as much money as possible?
(16GB HBM2 alone costs more than $300 to AMD)

There is no mistake, or missed opportunity, or price fixing. Just simple logic of losing a lot of money or trying to lose as much less as possible or better break even on re purposing a product

 

Not really, there was no new generation out only the Ultra. At best the modern equivalent of 8800GTX vs Ultra would be the an x80ti vs the equivalent titan card. But that's not really a fair comparison either as the GTX had the same amount of memory and that OC2 version I had was the same clock speeds as the Ultra anyway. There's a reason people still talk about the 8800GTX and not the Ultra.

 

I am thinking about the 8800gt here, that was a really good card for the money back then.

 

8800GT was the better card, not the Ultra.

 

Which was not out until December (I purchased the GTX in September) and slightly worse than the GTX ( https://www.anandtech.com/show/2365/9 ). When I paid £350 for the GTX the only better card was the Ultra which was basically the Titan of its day: Slightly better and significantly more expensive.

 

My BFG 8800 GTX was £329.99 ex VAT in December '06 from OcUK.

 

Hehehee I do still have the t-shirt came with

 

Yeah, the Vega 2 card is a pure compute card, it was made purely for compute, it's also a pipecleaner 7nm product and a shrink. It has very expensive memory doubled and a 1TB/s bandwidth which increases the cost again. It's designed for the professional market full stop. It happens to be faster for gaming so AMD said why not make it available. It's not optimised or efficient for gaming.

It's also retaining full FP64 performance like previous Pro cards except at a cheaper price. It offers I think 6.4TF of FP64 performance, Titan V is 7TF in the pci-e version, similar 'home' compute cards cost 3-4 times as much with about the same performance. This card was never meant to be an optimised or cheap gaming card. It's a straight compute card designed for that market and having more expensive memory for that reason. Because of the shrink it happens to be a bit faster for gaming and frankly they've made it available pretty damn cheap considering thy can sell pretty much every one as a expensive Professional card at a far higher price. They are basically giving people an option to get a higher performance card maybe not far off cost because hell, they are making it and why not.

If you think all that means AMD is price fixing with nvidia, because a product that has reasons to be expensive and is being sold with likely very low margins to give AMD users another option then frankly you're crazy.

If this was like previous home compute cards it would cost $1000, at $700 the difference is it probably (yet to be seen) won't have pro driver support but maybe it will, which actually makes this one of the cheapest ever cheap compute cards. IF it doesn't have pro driver compatibility then it's just a cheaper option for those doing compute that write their own software and don't need pro support for specific packages and there are a lot of people who fill that area. For those again who happen to want a faster card and it works for them, great. But this isn't years of optimisation to bring the cheapest gaming card, this is a product aimed and having an increased cost to produce for a different market that happened to end up faster for gaming. It's not amazing price performance, but it was never aiming or put a single second into trying to be that during development.

It's hilarious because 5-6 months ago when AMD were talking about Vega 7nm as a pro only card all about compute people were saying along the lines of "but it will be faster due to faster clocks, why not release a gaming version".... so AMD did, released a gaming version of a card potentially close to cost and people still go nuts over it.

I'm not buying one, Navi will offer far higher price/performance. I'm expecting similar performance at almost half the price from the first Navi and maybe 70-80% higher performance from a big Navi later on when a bigger die becomes viable. Those will offer huge price/performance gains that we want from new generations.

If someone wants 25-35% higher performance and doesn't want Nvidia then they have an option, better than not having the option.

 

Vega was built from the ground up to use HBM memory. Changing the memory configuration of Vega to use GDDR instead of HBM would have needed a complete redesign as well as completely new drivers. It would have cost AMD a fortune.

It's only a stop gap card to get rid of chips that didn't make the cut for the Instinct cards. This is AMD creating opportunity out of nothing. They had a chance to release a card with minimum effort and investment and they took it.

 

Just like the RX 590, although I'd argue they missed the opportunity for a full RX 600 refresh with 15% uptick across the range.

 

Argument: The Radeon VII wasn't built as a gaming card. It's a data centre card that has been re-branded as a gaming card.
And AMD are not actually bothered much about gaming sales.

https://www.youtube.com/watch?v=bXTSv50UBq4&t=0s

 

Same thing could be said for the original Vega, compute card first.