AMD claims that "Big Navi" will be their "halo product"

I am still skeptical about 50% perf/watt gains on the same 7nm node, but if true would be very impressive.

Also interesting to see the difference between Nvidia and AMD. Nvidia with its vastly superior budget is able to use one architecture for both gaming/HPC, etc, and dominate easily in both markets. Whereas the smaller budget of AMD requires two architectures.


I will say it is about time though that APUs finally move on from Vega. APUs will get very large perf/watt gains from that which only makes them even better than Intel which they already are anyways for mobile.
 
I am still skeptical about 50% perf/watt gains on the same 7nm node, but if true would be very impressive.

Also interesting to see the difference between Nvidia and AMD. Nvidia with its vastly superior budget is able to use one architecture for both gaming/HPC, etc, and dominate easily in both markets. Whereas the smaller budget of AMD requires two architectures.


I will say it is about time though that APUs finally move on from Vega. APUs will get very large perf/watt gains from that which only makes them even better than Intel which they already are anyways for mobile.

Nvidia also does this in some regards. The Ampere A100 features none of the "gaming stuff" (RT cores etc etc) that gaming cards will feature. Nvidia will be using the same architecture elsewhere, but the optimisation and orientation will be very different.

I agree with RDNA 2 on APUs. It will be a big change, especially when it comes to memory bandwidth efficiency. RDNA 1 gets a lot more from its memory bandwidth than Vega ever did. That said, Vega in the Ryzen 4000 series APUs back big perf/watt improvements that weren't seen elsewhere and weren't due to 7nm.
 
I am still skeptical about 50% perf/watt gains on the same 7nm node, but if true would be very impressive.

Also interesting to see the difference between Nvidia and AMD. Nvidia with its vastly superior budget is able to use one architecture for both gaming/HPC, etc, and dominate easily in both markets. Whereas the smaller budget of AMD requires two architectures.


I will say it is about time though that APUs finally move on from Vega. APUs will get very large perf/watt gains from that which only makes them even better than Intel which they already are anyways for mobile.
Big part of that is fixing power related bugs in rdna1...for a 7nm card, the 5700xt eats A LOT of power, a tad more than even 12nm gpus from nvidia, which is very very bad. So the 50% improvement is just fixing a big problem and advertising it as a big improvement. Just like they advertised 50% higher IPC for Zen vs Buldozer. The biggest leap ever in IPC. Sure, when you go from a piece of crap to something half decent, the improvement is huge, but it doesn't mean you did some engineering marvel to make it happen. You just fixed your own crap.
 
Big part of that is fixing power related bugs in rdna1...for a 7nm card, the 5700xt eats A LOT of power, a tad more than even 12nm gpus from nvidia, which is very very bad. So the 50% improvement is just fixing a big problem and advertising it as a big improvement. Just like they advertised 50% higher IPC for Zen vs Buldozer. The biggest leap ever in IPC. Sure, when you go from a piece of crap to something half decent, the improvement is huge, but it doesn't mean you did some engineering marvel to make it happen. You just fixed your own crap.
There's no bugs(As you can tell from how well these archs and recent TSMC nodes always work in power restricted environments like APUs), they just clock them closer to their limits on the desktop models, power/heat increases at an x^2 rate with voltage. They couldn't match NVidia on IPC because of their limited budget, so they just squeezed everything they practically could out of the silicon they could produce on their then miniscule budgets. Of course, now the money has been flowing in from Ryzen for a few years, following architectures have been having healthier R&D budgets.
 
Last edited:
Excite.

Typo. It's the Playstation 5 that's coming out, not 4.

There's no bugs(As you can tell from how well these archs and recent TSMC nodes always work in power restricted environments like APUs), they just clock them closer to their limits on the desktop models, power/heat increases at an x^2 rate with voltage. They couldn't match NVidia on IPC because of their limited budget, so they just squeezed everything they practically could out of the silicon they could produce on their then miniscule budgets. Of course, now the money has been flowing in from Ryzen for a few years, following architectures have been having healthier R&D budgets.

Yeah, this has been something AMD has had to do a few times now. In order to stay competitive, they've had to crank the clock speeds past the point of ideal efficiency. If they've remedied that with RDNA2, or improved it at least, that's excellent. I mean, that's partly why WattMan exists, and why undervolting their GPUs is so profitable.
 
I don't like AMD's marketing department. They end up saying all sorts of hype things and then engineers can't deliver. And then they spend time trying to match their big mouths instead of optimizing things that actually increase performance. Remember that boost graph AMD engineer drew before launch? Or drivers for example.

Other companies also say absolute rubbish but at least their products deliver advertised performance.

Saying the "Halo" product is a bit too much. Halo products are balls to the wall, cutting edge products that defy any reason. Rampage Extreme, RTX Titan. At those realms, there is no common sense, prices lose all meaning. Other reports say that Big Navi will be up to 2080 Ti performance. That is not a Halo product... They could have called 5700XT or Radeon 7 a halo product because it was the fastest thing they made.

If they had something that would compete with 3080 Ti or Titan (another version) I wouldn't mind them calling it a Halo product. But Nvidia has so powerful GPUs above mere mortal market that they can just pull one and rename it.

A good thing is that they will aim a bit higher. Having only one green card that beats your product is much more than having 4. Also, by having stronger AMD cards near the top there will be competition, and prices will be more tolerable. So where most of us dwell there will be more choices. That is good.

I understand that they are hyped and that they want to show how they have powerful cards, but don't oversell it. It is a question will you beat Nvidia's 2-year-old 2080 Ti and that is not even the most powerful GPU they have, and you are talking about Halo products. Be smart and focus on what you have better than the competition. Why would someone who buys cards in performance range between 2060 and 2080 choose your card over the green team's? That is where your strength is.
 
If RDNA2 has 50% IPC gains then technically an RDNA2 version of the 5700XT would surpass a 2080Ti.

If they're calling this a halo product presumably its 64CUs or more, so a 3080Ti is definitely within their bite.

NVidia does have larger GPUs but they can't realistically use a GA100 chip for consumer graphics, it doesn't have the interfaces or consumer facing units. The 2080Ti/Titan used a die size not too far off GV100, so I think it's safe to assume NVidias top tier gaming cards are now operating at the same limits of the process node as their top tier compute cards anyway

I do admit I much prefer AMDs marketing though, they're the only of the large companies that you can tell have engineers inputting more directly to their marketing, often they do let the numbers speak for themselves when they've got something good.
 
Last edited:
If RDNA2 has 50% IPC gains then technically an RDNA2 version of the 5700XT would surpass a 2080Ti.

If they're calling this a halo product presumably its 64CUs or more, so a 3080Ti is definitely within their bite.

NVidia does have larger GPUs but they can't realistically use a GA100 chip for consumer graphics, it doesn't have the interfaces or consumer facing units. The 2080Ti/Titan used a die size not too far off GV100, so I think it's safe to assume NVidias top tier gaming cards are now operating at the same limits of the process node as their top tier compute cards anyway

I do admit I much prefer AMDs marketing though, they're the only of the large companies that you can tell have engineers inputting more directly to their marketing, often they do let the numbers speak for themselves when they've got something good.

Nvidia is going 7 nm too. Imagine what 2080 Ti will do with just a die shrink. And they have optimized the hell out of that architecture for new cards. It is not like AMD's fight with Intel where they are competing with many years old 14 nm architecture.

So yea I don't think they have a chance for a Halo product. But I may be wrong.
 
A halo product is just any product that could have a halo effect, even if they underperform to a 3080Ti, if they're close but undercut the price significantly it could still be the stronger halo product, kinda like with the insane halo effect they got from the original Ryzen 8-cores.
 
A halo product is just any product that could have a halo effect, even if they underperform to a 3080Ti, if they're close but undercut the price significantly it could still be the stronger halo product, kinda like with the insane halo effect they got from the original Ryzen 8-cores.

They have hyped all previous GPUs.

After years of renaming mediocre products:
New Vega is coming. Hype everyone... Meh...
New Radeon 7 is coming. Hype everyone... Meh...
New Navi is coming. Hype everyone... 5700X is out... Meh...
New Big Navi Halo product is coming. Hype everyone...

See where I am going?
 
They have hyped all previous GPUs.

After years of renaming mediocre products:
New Vega is coming. Hype everyone... Meh...
New Radeon 7 is coming. Hype everyone... Meh...
New Navi is coming. Hype everyone... 5700X is out... Meh...
New Big Navi Halo product is coming. Hype everyone...

See where I am going?
Not really, Navi was super successful, since the 5700 launch AMD has been almost matching NVidia's sales figures at many etailers and dominated their target price brackets.

Radeon 7 was a weird one sure, but you can't really say it was hyped given they literally first mentioned it on the day of its launch, and throughout the whole launch stressed it wasn't really a consumer orientated card and was meant for a few niche markets & workloads, and in those specific workloads it was genuinely very competitive with the 2080 and 2080Ti for less money.

Vega was a genuinely competitive product that just got price gouged by cryptocurrency miners for its first 6 months.

If anything I think it's just some forum users than do the hyping.
 
Last edited:
People that chose 5xxx cards over Nvidia are selling them within a month and buying green equivalents due to driver issues. That has never happened before. Even people on this forum did some advanced tweaking to make them stable. Even though it is not all of them swapping and tweaking 5xxx cards are far from good a product.

So let us put all things on the table.

Gaming performance/price that is in favor of AMD. Good
Streaming on AMD can be done but Nvidia just rules there.
Productivity... Are we even talking about this?
Divers and stability on AMD are a dumpster fire.

So if you go with AMD card you get great value for your money in gaming performance, you shouldn't buy it if you want to stream, you can't get value from it if you do production work, a driver may run stable sometimes but it is not guaranteed. And how is that a better product?
 
People that chose 5xxx cards over Nvidia are selling them within a month and buying green equivalents due to driver issues. That has never happened before. Even people on this forum did some advanced tweaking to make them stable. Even though it is not all of them swapping and tweaking 5xxx cards are far from good a product.

So let us put all things on the table.

Gaming performance/price that is in favor of AMD. Good
Streaming on AMD can be done but Nvidia just rules there.
Productivity... Are we even talking about this?
Divers and stability on AMD are a dumpster fire.

So if you go with AMD card you get great value for your money in gaming performance, you shouldn't buy it if you want to stream, you can't get value from it if you do production work, a driver may run stable sometimes but it is not guaranteed. And how is that a better product?
Sure, if you spend enough time digging around on forums you'll find vocal people with issues, but I prefer actual facts over anecdotes, and the hard data says their RMA rates are in the range of 1-3% depending on SKU, IE pretty much exactly the same as NVidia's cards.

Honestly mate, on this one just look at the data. Navi seems to be AMD's most successful launch since the HD7970. They have gone from being clearly behind to matching NVidia's sales and RMA rates, because they're delivering more performance for less money.

Before Navi independent testing indicated that AMD's drivers were more stable than NVidia's, and now the arch is mature it seems like that may be the case again, is it not worth avoiding listening to people on forums when on the topic of tech, forums are just an amplification of emotional anecdote and the regurgitation of 5 year old tropes?

If you personally don't want the product, fair enough, but can you blame them for hyping a product that has completely changed the GPU market in terms of price/perf?(Which to most people in the defining factor of technological progression)
 
Last edited:
I've had more driver issues/crashes on Nvidia cards than on AMD ones ... other peoples experience is different, but Nvidia ain't perfect either.
 
I've had more driver issues/crashes on Nvidia cards than on AMD ones ... other peoples experience is different, but Nvidia ain't perfect either.

Yeah have had issues with my 2070 super where the fans go to 100% and the rig crashes.

People just expect all AMD GPU drivers to be bad. It's always been that way because for their first 8 years they were terrible.
 
Sure, if you spend enough time digging around on forums you'll find vocal people with issues, but I prefer actual facts over anecdotes, and the hard data says their RMA rates are in the range of 1-3% depending on SKU, IE pretty much exactly the same as NVidia's cards.
What are you using as a source for RMA numbers? Ones I've used (one local site and hardware.fr) have stopped reporting since 2017 or so.
 
Its a big gamble on AMD's part, this architecture is the only GPU that they are going to have on the market. PS5, XBOX4, Consumer GPU's, Server and workstation GPU's, APU's and more than likely going into the mobile (phone) chips that they are building with Samsung. However this performs this is all they've got atm so they'll need to hype it and just pray that it is as well received by the public as they hope. If its not you can guarantee they'll be working balls to the wall to get the most out of it through driver optimizations and bios updates if they can. With everything they have put into it and the inclusion in the next gen consoles expected to compete with PC gaming I have high hopes for RDNA2. As many fps as a 3080ti, probably not. cost as much as a 3080ti, Doubtful
 
Back
Top