Soldato
- Joined
- 14 Aug 2009
- Posts
- 3,106
True, but (from a consumer side) that's irrelevant.
-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
Poll: Do you think AMD will be able to compete with Nvidia again during the next few years?
- Thread starter CAT-THE-FIFTH
- Start date
AMD have no money to develop more graphics processors - this is why we see basically the same Polaris 10 in RX 480 (2016)/580 (2017)/590 (2018) 3 years as rebrands.
While nvidia develops different processors for basically every card - TU102, TU104, GP102, TU106, GP104, GP106.
6 different dies!
Soldato
- Joined
- 14 Aug 2009
- Posts
- 3,106
Consumer mentality that put AMD where is today, is not about how many dies you have to market, but how those dies perform in terms of price/performance and user experience. You could have 1 or 2 Polaris, plus Vega 56/64 (like it was HD7850/7870 and HD7950/7970), as long as they offer the performance and perceived quality of use desired by consumers. Even 3 dies would be enough if at the price range they're sold offer something solid.
Well, Intel and nVIDIA actions may have taken away from AMD profits, I'll give him that, but is not the only thing that has hurt AMD sales.
The same mentality that killed 3DfX and forced the best manufacturer Matrox to stop releasing any new graphics products.
Must be something wrong with the consumer.
What killed 3DFX was inability to execute (something AMD is going great guns on at the moment), and driving all their OEMs to Nvidia by being greedy and trying to enter the retail market.
Spending huge amounts of money on a halo product that has to be subsidized by their professional products (as Nvidia have had to do) isn't worth it when there is plenty of other low hanging fruit to grab first. There will be AMD high end products, but right now AMD are still building the technology and the revenue to make that happen. Chiplets, interposers, HBM, etc, these are all technologies that will come together to make a jump forwards from the monolithic graphic chips that have come to the end of their lives. This will be a parallel to what AMD have done to bring their CPU products back into contention.
And despite the naysayers and complaints, Vega is actually a great card for most people, only let down by it's launch price and mining inflation. People who bought them recently at sensible prices are pretty happy with them, compared to the insanely priced Nvidia cards that don't offer much more for more than double the price.
Associate
- Joined
- 29 Jun 2016
- Posts
- 541
I'm not convinced. Lisa Su has injected a hell of a lot of cash into R&D, and we all know they're trying a "Zen" approach in the GPU side of the business.
R&D increased by 25% in 2018 over 2017:
https://www.overclock3d.net/news/cp...s_increased_by_25_since_this_time_last_year/1
R&D increased every year by a good amount since Lisa Su came in in late 2014 (2015 budget probably set before she got into post):
https://www.statista.com/statistics/267873/amds-expenditure-on-research-and-development-since-2001/
Further, AMD paid off a lot of debts a few years back and are making profits for the first time in donkeys years. Lisa Su has said she will keep pushing R&D as a high priority.
She's got a PHD in semi-conductors and knows exactly what it takes to make successful products.
Also note that throwing money at the problem isn't a solution. You need capability and that takes years to build up, hence the gradual but noticeable increase in R&D funding year on year.
Just wondering, Lisa Su hasn't been CEO for that long at AMD in fairness, isn't Ryzen the first CPU that was developed from the ground up under her?
I'm assuming Polaris and Vega from the ground up were not, they seemed to be in development before her.
Associate
- Joined
- 29 Jun 2016
- Posts
- 541
I think Zen started in like 2012.
There is no doubt she has had a positive impact on the business though. AMD was in a shocking place in 2015...
I think, but please correct me if I'm wrong, that RTG was run quite separately from the CPU business until recently. Raja going and Lisa taking over I think was a good move to try and get RTG functioning, where it really does look like there was inconsistent strategy and/or leadership. I mean the whole Vega vs. Navi thing was quite eye opening, with Raja getting upset that his Driver Developers were taken off him.
I expect Lisa will have stomped that out quick smart, driven a consistent and well thought out strategy, and we'll see the benefits in Navi. The late-phase polish to a product makes a huge difference in my experience of developing electronics and FPGAs.
Associate
- Joined
- 29 Jun 2016
- Posts
- 541
Actually, reading the Zen Wiki, here is Lisa Su in action:
"Zen was originally planned for 2017 following the ARM64-based K12 sister core, but on AMD's 2015 Financial Analyst Day it was revealed that K12 was delayed in favor of the Zen design, to allow it to enter the market within the 2016 timeframe, with the release of the first Zen-based processors expected for October 2016."
That will have been a huge write-off, and not an easy decision to take or push through upper management.
"Zen was originally planned for 2017 following the ARM64-based K12 sister core, but on AMD's 2015 Financial Analyst Day it was revealed that K12 was delayed in favor of the Zen design, to allow it to enter the market within the 2016 timeframe, with the release of the first Zen-based processors expected for October 2016."
That will have been a huge write-off, and not an easy decision to take or push through upper management.
Don't know is my answer now. I really really hope so though. I need AMD competing and pushing Intel so I can get better Intel chips and although nvidia annihilate AMD at the top end in graphics, it would be nice if they released something to compete.
Soldato
Long ass video, but he has some interesting ideas that are at least plausible - whether true is yet to be seen.
Soldato
Look, if "mainstream" Navi is in par of GTX1080Ti performance at mid range prices ( sub £300), is a huge jump in price/perf for the segment.
And being "mid range" is nothing to be shy about tbh. As a GTX1080Ti is perfect card to run 2560x1440 & 3440x1440 144hz.
So, from the video:
Navi 12 ~~ RX Vega 56
Navi 10 ~~ GTX 1080 Ti
Navi 20 > RTX 2080 Ti
This is a good plan. Only to pull them forward for launches as soon as possible.
Where are all these chips going to be made? Intel are up against the wall not able to keep up and they have heaps of fabs. From what I can see we are looking at all the chiplets coming from TSMC and the IO chip coming from glofo.
Sure the chiplet idea should hopefully give great yields, but still - can TSMC produce enough for AMD to actually take significant share from Intel and Nvidia even if the chips are better?
Sure the chiplet idea should hopefully give great yields, but still - can TSMC produce enough for AMD to actually take significant share from Intel and Nvidia even if the chips are better?
I think 1080 Ti is a bit too much to ask for. I'm personally expecting 1080 non-Ti perf for ~£300.
Just don't see AMD being able to get that much extra perf from GCN on 7nm, to see a mid-range chip hitting 1080 Ti levels.
Also isn't the chiplet thing just for Zen 2 at this time?
GlobalFoundries, Samsung, TSMC.
I have zero idea what performance Navi will be. I am not expecting anything special. I was let down on both Fury and Vega where I thought performance would be better. This time I genuinely don’t know how good or bad Navi will be.
I hope it is very good and can at least compete with Nvidia on performance minus the tensor and rt cores. They at least have that going for them now. They should be able to use the space on the die to match nvidia on performance and provide a much lower price giving people with zero interest in RT and DLSS a cheaper option.
I hope it is very good and can at least compete with Nvidia on performance minus the tensor and rt cores. They at least have that going for them now. They should be able to use the space on the die to match nvidia on performance and provide a much lower price giving people with zero interest in RT and DLSS a cheaper option.
Radeon RX Vega 64 was released in August 2017. This year, AMD skipped to launch a Radeon replacement - if they don't do it till August 2019, Radeon RX Vega 64 will be already a 2-year-old.
Question, if Radeon RX Vega 64's pixel and texture fillrates are respectively 104.3 Gpixels/s and 417.3 GTexels/s, while nvidia RTX 2080 Ti's pixel and texture fillrates are respectively 136 Gpixels/s and 420.2 GTexels/s, why is the nvidia RTX 2080 Ti 100% faster than RX Vega 64?
Question, if Radeon RX Vega 64's pixel and texture fillrates are respectively 104.3 Gpixels/s and 417.3 GTexels/s, while nvidia RTX 2080 Ti's pixel and texture fillrates are respectively 136 Gpixels/s and 420.2 GTexels/s, why is the nvidia RTX 2080 Ti 100% faster than RX Vega 64?
because fill rates do not make a GPU. Vega has plenty of bottlenecks to not be able to use other advantages. nVidia on the other hand worked out lean mean gaming arch, without any compute stuff, then added RTX circuitry with RTX and called it a day. AMD has to design single arch for both server market and gaming market since, they were a bit broke few years back, so they could not make two archs simultaneously. I'm hoping now, while they are in the black AMD will have two designs, one for servers, another for gaming, and if they do, they might kick some major arse