Howdy, Stranger!
It looks like you're new here. If you want to get involved, click one of these buttons!
Quick Links
OPINION: AMD Still Doesn't Get It - MMORPG.com
System
Member UncommonPosts: 12,599
OPINION: AMD Still Doesn't Get It - MMORPG.com
I have never seen a company so seemingly content, so thoroughly blasé with spectacular mediocrity as AMD. Let's dive in.
Comments
AMD, on the other hand, will have had roughly two years of data to draw upon when they inevitably release their next generation cards. These cards will undoubtedly contain more efficient and more powerful hardware."
See what I did there?
Also, in what universe would a company close to bankruptcy, as was AMD a few years back, decide to attack on two major fronts simultaneously?
Their CPU position vs Intel was abysmal, while their GPUs were very competitive (2xx generation) to mildly competitive (3xx and later). It was a no-brainer: maintain the okay-ish GPU position, while your CPU division regroups and attacks on their front.
Now that they have their CPU counter offensive well on the way and have recaptured large parts of the territory, their GPU division will work on catching up with Nvidia. What's so difficult to understand here?
I agree that AMD has engineering deficit, but until NVidia manages to launch their 7nm products their process node advantage more than offsets that deficit.
Huge dies let you get top end performance that you just can't do with small ones. But AMD ceding the market for $1200 video cards to Nvidia, and often the $700 card tier as well, isn't a huge chunk of the market.
One thing you're missing is the price tags. If you go to New Egg and want the cheapest new cards you can find, the list prices are:
Radeon RX 5600 XT: $280
GeForce RTX 2060: $320
Radeon RX 5700: $330
It's worth noting that the RX 5700 has 8 GB of memory and a lot more bandwidth than the other two cards there. Nvidia may be claiming that the RTX 2060 is $300, but New Egg apparently didn't get that memo. It's a fair question to wonder whether it's just taking prices a while to adjust or the $300 price tag is a marketing stunt and not a real price.
As for AMD's updated RX 5600 XT BIOS, yeah, that's a last minute stunt. Unless your card says that it comes with the new BIOS, I'd assume that the old BIOS is the performance you'll get.
Even so, the Radeon RX 5600 XT is a third bin down of a GPU chip. Since when has AMD or Nvidia ever launched a full GPU chip and a salvage part of it, and then later another, further crippled salvage part, and had the latter actually be a good deal? The Radeon HD 4830 from 2008 is the most recent example that comes to mind. Further crippled salvage parts tend to be low volume parts, and are often sold in only one market, as that's all it takes to get rid of your stock.
Tensor cores are a waste of die space in a consumer part. Putting them in top end compute cards like the Tesla V100 makes sense. In consumer hardware, is just a waste of die space that inflates production costs without offering any actual benefit to gamers. And however much people like to deny it, costs of production inevitably get factored into retail price tags. AMD made the right call by not having anything analogous. It's Nvidia that goofed on that round.
Nvidia does have a head start on real-time ray tracing. But realistically, it's going to be several years before that matters, and by the time it does, this generation's hardware isn't going to be able to handle it. The added experience might ultimately give Nvidia some benefit by the time it matters or it might not. AMD had four generations of hardware tessellation on the market before Nvidia's first, but that didn't end up being a huge, obvious advantage for AMD.
Additionally AMDs next generation GPUs will likely be on 7nm EUV process instead of current 7nm manufacturing process.
With NVidia, I can generally tell that x is more recent, therefore more powerful than y because of its naming scheme. A 1060 is more recent in the series 10xx than the 1040. And anything in series 10xx is more recent and more powerful than anything 9xx or below.
With AMD, I'm just like "fuck it!" *throws hands up in frustration.*
That's not how Nvidia's naming works and by introducing a 16xx and 20xx series at the same time they actually made the naming worse than with what AMD is doing.
the 70 series from a generation ago is still stronger than the 50 series of this generation.
970 is better than 1050
Many people make this mistake in thinking the new generation is better
edit- Proof http://gpuboss.com/gpus/GeForce-GTX-970-vs-GeForce-GTX-1050-Ti
note thats even comparing to the TI (I hadnt seen that) which is the "best" 1050 you can get.
the way it works is the initial number is the generation 9xxx or 10xxx and the secondary number determines if its budget or not (more or less).
Like I said, Voodoo 2 FTW
My next rig (or GPU upgrade- depending) I'll be seriously considering going AMD...I'm at a point where I have more than enough graphics power and most of the games I play dont really require that much... edit- Looking at the games I play on Steam, many are a decade old or older lol
Bang for the buck is what I'll be looking at next over raw powers and features.
Graphics cards are probably the easiest upgrade to research with the easiest to obtain info, benchmarks, and useful reviews aimed at people with my level of knowledge. It doesn't matter if its AMD or Nvidia - just the sweet spot of price and performance.
Maybe a bad analogy and maybe not but thats how I saw things- However, these days we've reached a point where even you can game with an 'off the shelf' computer fairly well and those small increases are not really worth the huge pricetag anymore...For me anyhow...Especially when most of the games I play dont require much power at all.
Clearly doing something right.
Over the years I've built PCs using probably just as many AMD CPUs and GPUs as intel and nVidia: whichever fits the bill when I'm ready to build. Brand loyalty in these things is for fools.
“Microtransactions? In a single player role-playing game? Are you nuts?”
― CD PROJEKT RED
For both vendors' consumer parts, the first digit (or occasionally two) is the generation, and the next digit is relative performance within a generation. Since AMD bought ATI, they've had the 2000, 3000, 4000, 5000, 6000, 7000, 200, 300, 400, Vega, 500, and now another 5000 series. Meanwhile, Nvidia has had an 8000, 9000, 200, 400, 500, 600, 700, 900, 1000, 2000, and now 1600 series. You could argue that Nvidia's numbering is a little better than AMD's, but it's not a huge gap.
If you want a numbering scheme that makes sense, try the Sony PlayStation line.
Most likely, Nvidia's "7 nm" GPUs will be on a 7 nm EUV node such as what TSMC is calling 7+. AMD is moving to that node as well, and it's a significantly better node than the 7 nm that they're on right now. It's also a major redesign as compared to plain 7 nm. TSMC is claiming that their "6 nm" node is the natural successor to 7 nm, while 7+ is totally different.
Furthermore, every new generation has a chance to change arbitrarily many things. At the end of 28 nm, AMD and Nvidia were basically tied on efficiency (comparing a Fury X to a GTX 980 Ti). Then they both did a die shrink, and suddenly Nvidia was way ahead.
Or to go back a little further, AMD's Tahiti and Fiji GPUs were both of the GCN architecture and both built on TSMC's 28 nm process node. Yet Fiji offered about double the energy efficiency of Tahiti. Apparently they found a lot of ways to make it more efficient that they hadn't by the time they launched the first chip. There might be a whole lot of things that they can fix about RDNA in subsequent chips, just as there were with GCN.
Or more dramatically, consider their previous VLIW architecture. The Radeon HD 2900 XT was a disaster, as it was hot, late, and slow. It was inferior to Nvidia's GeForce 8800 GTX in pretty much every way, in spite of being on a better process node. A little over a year later, AMD had fixed the architecture with the Radeon HD 4870, and suddenly had an enormous die size advantage over Nvidia that would last for about 3 1/2 years. AMD used that to cut prices so aggressively that Nvidia discontinued products before they had successors and completely withdrew from the market for consumer GPUs that cost over $100 for several months.
I'm not predicting how AMD and Nvidia will compare with their next generation of GPUs, likely on a 7 nm EUV node. I'm only saying that how they compare can flip around dramatically in a single generation and has in the past.
DLSS is basically a complete disaster unless you are into blurry images. So people buying Nvidia are paying for all that extra junk on the die that in reality, they will never use.
People that make the G-Sync monitors have to pay an extra premium to Nvidia to use the tech while AMD's FreeSync is just that free. Hence just buying a G-Sync monitor to use your Nvidia card with will cost you another $100-150. Far more monitors offer FreeSync. As to Nvidia's claim that G-Sync will work with FreeSync monitors, that is a very shaky claim.
I can get AMD cards at discounted prices, Nvidia, not a chance.
AMD has been sticking it to Nvidia lately on price and will continue to do so. If you want to go with Nvidia expect to pay a premium for the name.
Seems to me that AMD is doing great in the GPU market at present and I don't see Nvidia doing much about it, they still cannot compete on price which is the deciding factor in many buying decisions.
Millions..Billions of dollars later and after 30+ years we barely see any improvement in games or the PC's.We still have sub par games only attaining 30-50 frames on super expensive rigs,console games are predominantly 28-30 frames,poorly optimized games,poorly done Bios,poorly done drivers etc etc,almost like we are moving backwards but costs go up.
Like what is the sense of making pc's 10x better,gpu's 10x better but games only 10% better but run 50% worse,we never make headway.
Never forget 3 mile Island and never trust a government official or company spokesman.
The 5700xt is the best bang for the buck if you're looking for 1440p gaming. You can preach features like dlss and ray tracing all day long but at the moment they are doing almost nothing for gamers.
First off both features have to be implemented by the devs and few games support either of them.
Also unless you're playing at 1080p you can forget ray tracing and a decent frame rate because it's still way too demanding.
Nvidia has convinced gamers they would be missing out on something if they didn't have these two features. So gamers but into their marketing and still can't even use these features unless it's 1080p or they have a $1000 gpu and by the time ray tracing is optimized enough to use you will have bought a while new card. But don't worry Nvidia will convince you about some other tech you just have to have but can't use.
And you'll pay too much for it to...the future is bright!
Yeah, AMD had to sharply discount some of their CPUs several years ago in order to sell them. But that's not the case today. AMD just had their highest quarterly revenue ever, even though console sales were down and their coming flood of sales from Ryzen Mobile 4000 series and EPYC Rome mostly hasn't kicked in yet.
That Core i5-9400F that you left unnamed but seem to be referring to gets smoked by a Ryzen 5 3600 for about the same price. Singled-threaded performance will at least be close, but once you push a lot of cores, the AMD CPU will win by a lot.
Intel does still have the best single-threaded performance if you're willing to buy one of their top few mainstream consumer parts (basically, a $370 Core i7-9700KF or better), but not by a very large margin. And even that is likely to go away later this year when Zen 3 arrives.