Potato Head
Newcomer
Throughout their history - from their very beginning - AMD has only done well when their competitors falter. Pentium 4, 14nm+++++++, etc. I don’t see Nvidia faltering anytime soon.
Value of Hardware Unboxed benchmarking
- Thread starter Phantom88
- Start date
Throughout their history - from their very beginning - AMD has only done well when their competitors falter. Pentium 4, 14nm+++++++, etc. I don’t see Nvidia faltering anytime soon.
AMD benefited a lot with Zen 1+2 in terms of Intel entering its big stumble, but they've been really dang solid with Zen 3+4 all on their own. Zen 5 is perhaps not that good in many respects for consumers, though 9800X3D does show it's being held back and that there's a good amount of improvement to come from this sort of new super wide architecture base.
Also, while Nvidia is likely never gonna stumble as hard as Intel did with 10nm and the resulting domino of problems that created, it is looking like Blackwell is gonna be their most disappointing new generation since Fermi. So AMD does have an opportunity here to get back in the conversation with RDNA4 if they can stop tripping over themselves.
They're calling it 4090Ti but I wonder what they think NVIDIA could've done differently. 575W and basically hitting the reticle limit. I guess he's comparing to some other IHV's superior architecture... oh wait I just remembered NVIDIA was already the leader in perf/w and perf/mm².
It's not just HUB, I'm tiring of the constant negativity from reviewers. Seems clear that if there was a better path forward someone would be taking it.
I think some reviewers understand it, just not all of them. Some reviewers, just like some participants on this forum, would've rather seen a doubling of performance for a sub-$1000 pricetag -- you know, to go back to the good ol' days when an video card could be overclocked by 33%, a CPU could be overclocked by 50%, and it only cost $249 USD plus tax. Hell, I want all of that too, but I'm also keenly aware of where Moore's law has left us today. We're closing in on a hundred billion transistors, 32GB of GDDR7 ram, enough power to run every light in my house four times over but wedged into a dual slot cooler, and even with an almost linear increase in performance over the 4090 relative to core count, I get that the pricing isn't going to be what we want. Especially since, in every review, the only competition is NVIDIA themselves.
It looks like the 5090 is only going to serve people with exceptionally high rez devices (4k +) and at absurd framerates, or those who are going to do some AI stuff (because all thet extra INT4 math capacity will absolutely crush AI workloads.) For those who don't need these things, the 5080 is probably gonna be a great bet, or hell maybe even the 5070Ti...
It looks like the 5090 is only going to serve people with exceptionally high rez devices (4k +) and at absurd framerates, or those who are going to do some AI stuff (because all thet extra INT4 math capacity will absolutely crush AI workloads.) For those who don't need these things, the 5080 is probably gonna be a great bet, or hell maybe even the 5070Ti...
Sometimes I wonder if the older tech reviewers are kind of mad because performance scaling slowing rapidly kind of kills off their business model. If performance stops scaling and gpu vendors start to push more and more towards software features and ai stuff, then reviews that just do performance/watt are going to get pretty boring if it's 10% every time.
DegustatoR
Legend
It's the same thing as with 40 series launch again - reviews are being written against some dream of how a good GPU would perform and how much it would cost. This dream may be based on previous experience but it's still just a dream which makes all comparisons to it pointless by default.
Now if some vendor would actually make a considerably better GPU in either price or performance then that'd be a valid point for criticism. As it is though it's just whining that things aren't as good as they used to be 20 years ago - while things are in fact many times better, it's the improvements which aren't as good anymore. But this is a typical scenario of any technical evolution. Big steps early on always turn into smaller improvements which are often invisible to a person outside of the industry.
Now if some vendor would actually make a considerably better GPU in either price or performance then that'd be a valid point for criticism. As it is though it's just whining that things aren't as good as they used to be 20 years ago - while things are in fact many times better, it's the improvements which aren't as good anymore. But this is a typical scenario of any technical evolution. Big steps early on always turn into smaller improvements which are often invisible to a person outside of the industry.
I think there were legitimate complaints on pricing of the 40 series. 4070Ti (12GB) for $800 and $1200 4080 were not reasonable IMO. But the 50 series is quite a lot better on that front. The 5080 costs about $950 in 2022 dollars.
Again if NVIDIA were so out of line I think AMD would have a field day. But AMD consistently prices just a bit below NV. Unless they truly hate marketshare I think there is more to it than NVIDIA greed.
Also if the cards get scalped, it's hard to argue that they are too expensive. It pains me to say since I love a good value GPU.
Rich from Digital Foundry made basically the exact same comment by the way, that this feels more like a 40 series Titan part than any new generation.
As for people being negative, they have every reason to be. The thing most people want is BETTER VALUE. A meaningful improvement in performance per dollar. And Nvidia have all the power to offer that, but they wont. People are rightly tired of the greed, no matter how much this forum bizarrely keeps trying to defend it. Nvidia could offer a 5080 at $1000 that's actually a further cut down GB202 die, but nope. Nvidia could offer 5080 specs(which are really just upper midrange) for more like $700, but nope. This would represent a good leap that people would be pleased about. But it would impinge on Nvidia's ridiculous, earth swelling profits, so nope.
There's also just a disappointing lack of architectural gains here with performance efficiency or power efficiency. Nvidia and AMD have both made improvements here with processors even on the same node(not that 3nm wasn't available), so unless you believe that Lovelace is literally the peak architectural efficiency possible from a graphics processor, I dont know what to say. It's simply disappointing. It just is, and I dont know why on earth you guys just cant recognize that. It's far more absurd to suggest that reviewers be positive about it when it's not really delivering anything good for consumers.
I need to hear an argument as to why AMD hasn't swept in with much better value cards. They have every reason to do so.
Subtlesnake
Regular
No company is going to voluntarily reduce their margins unless forced to by market forces.
Yah, I'm in the same camp. Their competitors aren't doing it either, so what's the reason?
They don't know how 5070 and 5070 Ti perform and don't want the PR catastrophe of cutting prices couple weeks after launch nor cut their profits unnecessarily by pricing them too low.
As side product they get time to optimize software and more FSR 4 titles. Who knows, maybe they'll use the time to enable multiframe generation built into FSR frame gen since based on code it's been built for it since beginning
No, I'm saying why aren't they offering 5080 performance at $700 as the poster above suggested? 7900 xtx launched well above that. I still don't think it's down that low.
Subtlesnake
Regular
I doubt they can reach 5080 performance with Navi 48. They may offer close to 5070 Ti performance (in raster) for $500 or $600.
If it's cheaper than the 5070, performs like a 5070Ti, has 16GB, and FSR4 is good (and well supported) they could move some cards IMO. It seems like a tall order but they have to do something big.
Because when the decisions about this gen were made they had no idea what kind of chips NVIDIA would produce and chose to focus on mainsteream
Because AMD want to ride the coattails of Nvidia's rising margins, too? Because AMD can also be greedy? It's really not complicated guys.
Do I need to bring up Intel here, because you seem to be completely forgetting them?
Well that's not actually some universal truth at all, but regardless, it changes nothing about consumers being entirely right to complain.
Something being better for a company's financials doesn't mean that consumers should just swallow it and be ok with it. Do you realize how utterly absurd that line of thinking is?
You're basically saying, "So long as it helps make a company more money, you have no right to complain as a consumer". I cant stress enough how utterly bizarre this forum is in pushing this line of thinking constantly when it comes to defending Nvidia. Absolute bonkers.
AMD prices would be lower if Nvidia's prices were lower.
Some people just understand how the world works and that screaming at the sky will change nothing. It is an unassailable truth that publicly held companies have only one obligation - profit maximization for shareholders (within the law). Pretending otherwise is just living in la la land.
Well he’s doing it primarily to rile up the crowd and draw visits to his content. He definitely understands it and explicitly says that Nvidia couldn’t do much while staying on on 4nm without an architecture overhaul.
In general Steve focuses on consumer outcomes and not the “why” behind them which is a valid approach to take. It certainly plays well with his fans.
DegustatoR
Legend
That in itself is disingenuous as well though.
We have no indication that "an architectural overhaul" would lead to anything being better. There are no better architectures on the market, on the contrary everyone is chasing Nvidia's h/w.
And while moving to a more advanced node would allow for some power savings it is highly likely that the same complexity increase would in fact cost even more to the end buyer making perf/price on 5090 even worse.
Similar threads
