financials are ok but how are they /really/ doing with their fanboi futures? seems like intel is well ahead of AMD in the rabidity indices and mouth-to-foam ratios.
This being a PC-oriented site, we don't have a lot of console peasants here. So, trolling them would be a lot less successful than trolling AMD fans.
Even if he were spot on, it's unclear what he hopes to accomplish by whining about it here. It's just tilting at windmills. That's yet another reason I think it's just a long-troll.
Are you seriously implying Apple doesn't have 'fanboy' appeal Azix? They probably have the largest cult following of any single company in the history of the universe.
Apple represents about 12% of the entire US stock market cap today. I'm a reluctant Apple owner but I'm interested in their stock price for more than just devices.
I think the reason AMD is being reported is because unlike ALL those other companies you mentioned, along with others, AMD has had a massive turnaround, and AMD is a chip company. None of the other companies make chips. Right now, the global focus is on chip production, hence all the talk of TSMC, Intel and now AMD over the last few months on AT.
This isn't a financial website, and I don't expect them to report financials of every tech company. Just the ones that most specifically pertain to the spirit of Anandtech (which started 23 years ago, effectively, as a computer hardware review website focused on the Intel\AMD war of the late 90's\early 00's.
Right. Apple makes $10b a quarter in Mac sales, often beating or matching Dell, Lenovo, and HP.
$10 billion revenue per quarter selling Macs, lmao.
Clearly, they’ve done quite well. Not to mention the M1 is a massive transition for the computing market, besides its near chart-topping performance-per-watt.
I’d like to see the next “cell phone company” that can break 7.10 on SPECint2017 this cycle. The ranking champ is the 5950X at 7.65, so bloody good luck.
We stopped reporting Apple results a couple of years back. When they stopped reporting unit sales counts and became more cagey about sales data in general, it really put a crimp on our ability to say much of interest about the data. Plus everyone reports on Apple, which makes it hard for us to stand out.
This is actually the first time I've ever seen anyone ask about the matter. Is there demand here for Apple earnings coverage? If so, perhaps I need to look at re-adding that for future quarters.
It's not tech news, per se. If you want financial news, get it from a finance site.
Anandtech already misses enough genuine tech news stories & reviews. Do you want them to miss even *more* of the stuff they're actually good at covering, just because you're too lazy to get Apple news from somewhere else?
Agree with this. AMD financials are interesting to a degree as it's such an impressive turnaround with real implications for end users. But, ultimately, if I want finance news, I'd go to a finance website.
The occasional article makes sense when something stands out but, as someone else mentioned, I'd prefer time and energy put into tech stuff rather than trying to deliver financial analysis on all the major tech companies.
I created an account because I think its important to cover Apple's earnings. The company is the largest technology hardware player and even if it outright hazy about per unit sales, it deserves covering.
Most of the time, your coverage may not be differentiated from the financial service guys. But I am pretty sure that you will find some areas of focus that will be different and add value.
Earnings news is not a big part of Anandtech. But I was pleasantly surprised when it began a few years ago. Apple is worth covering and business reasons for tech decisions are increasingly valuable and not well covered by either financial or tech media.
Because Apple is more of a lifestyle company than a tech company.
All joking aside, Apple is getting big into content, finance, etc. They've outgrown actual technology products, as they search for ways to grow their already enormous revenue.
If you look at pc build forums, the results should be obvious. Everyone has Ryzen in their new builds! Still don't understand how Intel lost such a big lead, what were their engineers upto?
I was shopping for a new system I wouldn’t want to build anything with Intel today plainly on its heat output alone. Not to mention Intel’s tactics and delays and lies and reheated junk for years.
AMD has been killing it and I can only hope they’ll have many more quarters like these so they can really increase their total market share / volume. Intel has been, stubbornly, stuck at 70%+ for two decades. Let’s see a real fight here.
Up to engineering things that clearly weren’t working.
Do you think that AMD is great social company that will make cheapest possible cpu to make folks most beneficial ? Look at numbers again - they are increasing gross margin to level of Intel or more if there will be possibility. When they hit safe ground, then R&D will be decreased, and you will see similar reheated junk, as from Intel.
Absolutely not. When that happens, it's time to switch. All companies go through "ah, we need to actually care this time, lol" phases and that's when you buy. Intel can change its tune, Qualcomm, Apple, etc. They all can.
Anybody that has "loyalty" to billion/trillion-dollar corporations with millionaire/billionaire CEOs has more than a few screws loose.
Whoever makes the most compelling platform + CPU gets the nod. This isn't too complicated...
I don't think they ever will. Not only will they have a resurgent Intel to contend with, but they'll also have to deal with ARM, RISC-V, the disappearance of the Chinese market, and eventually Chinese competitors reaching markets outside of China.
> then R&D will be decreased, and you will see similar reheated junk, as from Intel.
We don't know that. They don't have the same history as Intel of dividends and share buybacks.
It's possible that, if they plateau at some point, they'll face a lot of pressure from investors to do some of those things in lieu of further share growth. However, that will probably trigger the start of their next big decline.
Yeah and that's great, but they're in a fiercely competitive landscape, right now. They will plateau, at some point. And when they do, it'll be tough to stave off the parasites... er, I mean "investors".
Two different vision, two different strategy, two different companies.
AMD always did it the right way. Claiming that AMD is now going to utilize Intel or Nvidia dirty practices when it is clearly not in their philosophy, is nothing more than you spreading nonsense.
While Nvidia and Intel are bribing studios and AIBs, AMD is releasing open source ecosystem... and from your point of view, these companies are all the same...
Their problem is their fabs and they are now making it a business. Their main competitor is not AMD, it is TSMC. If you think TSMC is just going to sit still and let Intel do their things, then you don't get it. Intel will never get the node advantage back, they are 3-4 years behind TSMC.
They need TSMC to drop the ball for 3-4 years to catch them up, this is how bad it is. Putting an Engineer as CEO will not change this. Hope is not a strategy!
> Still don't understand how Intel lost such a big lead, what were their engineers upto?
It's been said many times: Intel has been extracting value from the company for many years. They spend a lot of their profits on dividends and share buy-backs, then turn around and do rounds of layoffs to keep costs down. The result can hardly be surprising.
That's nto all they did though. Intel has also invested billions into the 10nm process trying to get it to work properly, and is still investing billiosn into further R+D and fab capacity.
AMD got really lucky that intel ran into so many roadblocks with 10nm production.
> what AMD has done with zen has no bearing on wht intel wa sspending money on this whole time.
I'm not so sure about that. If Zen wasn't so strong, I think it wouldn't have created quite the same urgency, on Intel's part. Zen could turn out to be the best thing to happen to Intel, if the whole management shakeup and reallocation of funding wouldn't have happened (or at least to the same degree), otherwise.
It is not pure luck though. There's execution. Supposing intel had it's 10nm up and running in 2016, Zen and the chiplet architecture was still going to come out, it could have taken longer for adoption maybe, but all the advantages and disadvantages between the two were going to be compared against each other still, and AMD would still be a tough competitor with that zen uarch
> AMD began initial shipments of their first CDNA 2 architecture-based > Instinct accelerators in Q2,
Of course, you wouldn't know this if you relied solely on Anandtech for tech news. : (
> opening the spigot there for data center GPU revenue going into Q3.
Well, not exactly. They have a "software pipeline" problem that basically none of their consumer hardware is supported by their compute stack, so basically the only ones porting anything to it are AMD employees and probably a few grad students working as interns on the couple HPC wins AMD landed.
The genius move Nvidia made was to fully support CUDA on their entire hardware range, from top to (almost) bottom. Sometimes, their very lowest-end card doesn't support it, but even their Jetson embedded compute modules can run it.
Intel is doing one better than Nvidia, by supporting their compute APIs on CPUs, GPUs, FPGAs, and (where appropriate) hardware AI chips.
As long as AMD considers cloud and HPC as their only compute customers, adoption even in those segments is going to be limited. And as adoption of purpose-built AI hardware starts to gain momentum, AMD is going to see sales of their CDNA products hit a brick wall.
Why would they reveal specific customer sales of the MI200? Have they done that with Radeon? No. The fact the product is ready and targeted for major compute markets, especially Supercomputer clients that have already been published should be a clue who is using them.
> There isn't anything to report on that front. AMD isn't saying a thing about CDNA 2
Thanks for confirming. Not to sound ungrateful, but um, what about CDNA 1 (MI100) coverage? As I recall, there've also been a few other GPUs that launched without coverage on this site.
The "GENIUS" Nvidia move is now biting them in the ass, because ampere is hosting double the cores and offer 40% performance uplift. They are facing a GCN moment where they don't have a gaming uarch anymore... and they lost the entire gaming market. Holding the dGPU business is nothing if studios don't use your hardware beside when you bribe them to implement your tech.
Don't lure yourself, games are made on consoles now. FSR adoption will skyrocket with the console adoption to. Nvidia cornered themselves with DLSS and RTX to paint AMD as the duck, but now with FSR, it literally backfired royally.
They do nerf some their consumer cards in some ways that are important for professional users.
But, the key point is that you can still use consumer cards for development & dabbling with the whole Nvidia software stack. This is important for getting kids in university to port & optimize open source software for your hardware.
Much of the goodwill AMD built by having an open source driver stack, they've lost by stumbling so badly with lacking compute support for consumer hardware. Just checkout any discussion of ROCm on pretty much any site, and you'll find much gnashing of teeth about this.
Goodwill for those who invested when the company was cheap. Goodwill for those who care about efficient CPUs and not high-performance gaming.
Anti-goodwill for the few consumers who have a clue about things like:
— designing GPUs to sell to mining first and foremost
— selling junk at the ‘high end’ so Nvidia can raise prices, the junk that didn’t go to miners will be purchased by foolish ‘team red’s, and the high prices of PC gaming will sell more PCs in disguise — that AMD happens to allocate so many wafers to.
— sandbagging for many years at the midrange and above areas, to, additionally, raise prices with Nvidia via pent-up demand, not be so obvious in the various anti-PC enthusiast schemes (once actually competitive cards are deigned to be produced — just by chance during the great shortage), and be hailed as amazing for the technical wizardry of actually bothering to release something competitive for the first time since Hawaii.
This comment entails the assumption that designing GPUs to compete with Nvidia at the high-end is relatively trivial and something they just decided not to do.
It's not. Just ask Intel, 3DFx, Img tech, Matrox, etc...
" There are no credible excuses for that and for plenty of other very substantive examples that support my case. " oh ? then post them on here so we can see them too, if not, its STILL just more BS
Search for "Differences Between VEGA and Previous Devices", and remember that Polaris came between Fury and Vega. So, for a more complete list of changes, you'd probably also have to consult the Polaris docs.
But that's just the ISA doc. GPU performance is not only about IPC. Vega has other differences, highlighted in the review:
* New memory page management for the high-bandwidth cache controller * Tiled rasterization (Draw Stream Binning Rasterizer) * Increased ROP efficiency via L2 cache * Improved geometry engine * Primitive shading for even faster triangle culling * Direct3D feature level 12_1 graphics features * Improved display controllers
I also remember seeing a few benchmarks where Vega 56 had up to a 2x improvement over Fury.
> There are no credible excuses for that
Vega was dual-focused at gaming and machine learning.
Also, by fabbing it on a smaller process node, AMD was able to turn up the clock speed.
With all that said, I don't want to defend Vega as if it was actually good. Clearly, it was lacking in some key respects. I think its origins were from a time when AMD still thought the GPU race was mostly about brute-force. RDNA was obviously an attempt to course-correct.
BTW, I don't really see your point of calling out Vega, specifically. If you go back through the entire GCN range, each was an evolutionary improvement on the previous generation. Vega was entirely consistent with that.
'That's not really true. They did make some ISA-level improvements.'
Just like Parhelia was a great competitor because it had things like glyph smoothing.
A laundry list of mostly superfluous features is irrelevant to the point. Review sites that did clock-to-clock gaming tests found that Vega's IPC is identical — and there is no excuse for that.
What there is is AMD purposefully not being competitive beyond the low midrange so Nvidia could pump prices up and create artificial demand for 'consoles'. It also kept Polaris-forever overpriced and selling.
> — designing GPUs to sell to mining first and foremost
This just shows the depth of your ignorance on the subject. If AMD cared about miners, they'd actually support GPU compute on those cards!
I know you don't *really* care about facts, if they get in the way of your scapegoating.
That's the biggest problem with your whole conspiracy theory -- it simply doesn't make sense. I get that you're angry and searching for somewhere to target your rage, but you've gotta get past that.
As for empty rhetoric, I've yet to see much else from you. When you do make a case, it's typically weak and not backed by sources or references. The amount of shade you cast, relative to the informational content of your posts is quite high. That's more like a caricature of Oxford than the reality.
On the internet, the only thing that matters is the strength of your argument. I don't care if you're an Oxford prof or a homeless dropout. Since we can't verify your background, it counts for nothing.
AMD is playing a bit of "catch up" in the GPU realm as they devoted most of their resources to Zen, rightfully so, as it probably saved them from bankruptcy. The real question is can they catch up NVIDIA. WCCFTech covered a rumor that RDNA3 will be better than Lovelace, but it is WCCFtech, so who knows.
They are close to NVIDIA with RDNA2 if you exclude extras like RT and DLSS. NVIDIA got the head start though so AMD is playing catch up there.
'AMD is playing a bit of "catch up" in the GPU realm as they devoted most of their resources to Zen, rightfully so,'
It would have bankrupted the company to not release Vega (after such a long delay) with identical IPS to Fury X.
(eyeroll)
Wake up and realize that AMD actively works against the PC gaming market.
It makes mining GPUs and allocates wafers for the 'console' scam. It actively works to keep prices too high in the PC gaming market in a variety of ways, in order to make the 'console' scam function.
It also exploits the dire lack of competition in GPUs, where desperate stupid people will buy its products due to emotional nonsense — like 'It's my donation to keep competition alive!' and 'I'm team red because I like the underdog' — and also because Nvidia keeps prices very high (thanks to help from AMD) so AMD can sell things like 'Polaris forever'.
AMD is also the company that broke AVX when it 'improved' upon Bulldozer (with the Piledriver release) and had the audacity to release the 9000 series, with extremely high prices based on falsehoods like 5 GHz and 8 cores.
It didn't bother to provide even a shred of value to gamers, only hype and lies. The single-thread IPC of Piledriver was so terrible that even at over 5 GHz (don't ask how much power that required, or the cost of the board, or the cost of the cooling, or the noise) it was too slow to prevent lag problems. It could have replaced Piledriver with something that would better justify the extremely high price it put on the 9000 series but... nope. Maximum ridiculousness. Maximum cravenness.
What does it take for people to stop peddling AMD's propaganda? I suppose it takes not having investments in one or more of the following: AMD, Nvidia, Sony, Microsoft... and even Nintendo.
Even Nintendo is helped by the console scam because if there would be robust demand for the PC gaming platform and affordable hardware to go with it that's not locked behind a ton of duplicate walled gardens (aka the 'consoles'), then portable PC gaming systems — recognized for what they are — x86 bog-standard PC gaming hardware ... that even use the same substandard craven joystick mechanism ... Nintendo would have a harder time selling its overpriced substandard hardware complete with its own superfluous parasitic walled garden.
AMD has stated it had no intention of reducing the attractiveness of its cards for mining.
These are the sort of bald facts that jokers on this site like to ignore in favor of ad hominem and flamboyant specious braggartry.
AMD won't sell cards made for PC gaming, rather than cards for mining and wafers for consoles. How can that be? AMD is some sort of benighted altruistic superhero... not a money-grubbing corporation that has absolutely no soul and, like any corporation, is designed to extract maximum profit via delivering the least value to the consumer. Sell less for more. Astroturf the truth away.
And again, this is you not understanding GPU performance. Vega 64 performed a lot better than Fury X, because GPU performance is not simply a function of IPC.
In point of fact, the entire range of GCN GPUs, spanning all the way from the HD 7000-series (launched in 2012) to Radeon VII (launched in 2019), had basically the same instruction throughput per CU per clock.
" allocates wafers for the 'console' scam. " do you have link that shows its AMD's allocation of wafers, and not sony/MS that is buying the wafers ?? are you forgetting the reports that nvidia sold cards directly to minors ? you ONLY seem to be pointing the finger at AMD, while forgetting the prices nvidia and intel also charged for its products. i guess cause you hate " a ton of duplicate walled gardens " i guess Apple's whole product line, makes you catatonic
as i said, post links to this BS. so far, it is ONLY your opinion.
Oxford Guy, I will look into this, but did AVX break or was it slower? Also, we know Bulldozer was a disaster: if they kept on churning out Bulldozer variants today, we've got licence to criticise, but the fact is, they made an architecture from scratch, so excellent that it's turning on the heat beneath Intel's bum; in fact, scorched it. And they've suffered enough from Bulldozer already, so I don't think it's fair to attack them for a mistake from their past.
"Memory writes with the 256-bit AVX registers are exceptionally slow. The measured throughput is 5 - 6 times slower than on the previous model (Bulldozer), and 8 - 9 times slower than two 128-bit writes."
You're right about Piledriver's AVX: it was slow enough to be considered broken. 256-bit stores were the problem, taking 17-20 cycles. However, Steamroller mended that, down to 2 cycles for aligned 256-bit stores and 3 for unaligned ones (faster than Bulldozer's 3 and 10). Excavator, no regression in that regard.
BTW, the fact that it was such a huge regression makes it sound like a hardware bug. Maybe fixing it would've pushed it too close to the market window of Steamroller and not many things were using AVX at the time, anyhow?
The block diagram could shed some light, or maybe not, but I'm going to guess that when they changed the FP backend, possibly adding more units, they were out of time and locked down AVX writes to a certain speed, which they knew worked. Then in Steamroller, had more time to validate and get it working at its right speed. AVX was pretty new back then, so no big loss. A bit slapdash? I suppose. Or: "It'll be fixed in Steamroller, so don't worry, my Piledriver comrades. You know, the Steamroller guys are busy tackling 256-bit stores as we speak. Got an email from the lead architect this morning."
If you remember Zen 1, they set the L2 cache at 17-cycles latency because they were out of time. Then in Zen+ (and I think Raven Ridge), chopped it down to 12 cycles.
All GPUs are mining GPUs. Nvidia's flailing attempts to make its Ampere GPUs *bad* at mining is a perfect example of how hard it is, when it's in fact certain cryptocurrencies that are specifically designed to run well on GPUs.
> allocates wafers for the 'console' scam.
AMD is bound by supply agreements. It's not their choice.
> It actively works to keep prices too high in the PC gaming market in a variety of ways
How?
> in order to make the 'console' scam function.
What's their motivation for this, given that they could be making more $ if they had the freedom to re-target those wafers? This is the critical flaw in your argument, even before we get to the part where you provide zero evidence to back it up.
> had the audacity to release the 9000 series, with extremely high prices > based on falsehoods like 5 GHz and 8 cores.
It's just the same thing we've seen out of Intel. When these companies are stuck in a product line longer than they want to be, it pushes them to take desperate measures, like cranking up clockspeed.
> The single-thread IPC of Piledriver was so terrible
The two interviews Ian recently did with Jim Keller shed some very interesting insight into how much he had to transform the culture of AMD from one of defeatist incrementalism, in order to make Zen happen. If you haven't read those, I'd recommend them.
What I don't understand is how you think this failure to execute is somehow evidence of a conspiracy. It also reveals quite a lot about the strength of your argument (or lack thereof) that you have to go so far back to make your case against AMD.
> Even Nintendo is helped by the console scam
Here's the part where you're bad at conspiracy theories. Any good conspiracy theory treats incentive almost as proof of complicity. Instead of launching on an unrelated tangent about Nintendo, you should find some way of tying Nintendo more deeply into the whole thing.
> AMD has stated it had no intention of reducing the attractiveness of its cards for mining.
Unlike Nvidia, they don't have a separate mining-oriented product line. That's why. Both companies would be punished by their investors (and probably even sued), if they went out of their way to shut out miners and that revenue stream. That's wholly different than saying they set out to build good mining cards.
And lots of people use their cards for mining when they're not gaming, in order to offset the hardware costs. Of course, that only works of you're in a cheap electricity market (or someone else is paying the electric bill, which I'm sure is the case for most kids).
What's weird is that you didn't even touch on AI. That's the one case where there's actual evidence that both companies juiced their hardware for AI. And it's surely consuming quite a bit of Nvidia's die allocation, although I don't think many people are using AMD GPUs for AI.
HPC is another area you could attack. Vega 20 (the 7 nm version) added fp64, in addition to further AI-targeted enhancements. Their CDNA chips (MI100 and now MI200) take these themes even further. And Nvidia is targeting enormous A100 GPUs at HPC, as well as AI. Why go after consoles and not HPC?
> These are the sort of bald facts that jokers on this site like to ignore
They're tenuously-related and really don't support your claims. So much "empty rhetoric", as someone recently said.
Again, I can see that we're beyond facts. You're starting with a scapegoat and trying to cherry-pick facts to support your case. You clearly feel something is deeply lacking in the PC gaming market. You want to recapture some lightning in a bottle, from a bygone era. Maybe it's the lack of better titles. Maybe you work/worked for a indie game developer and sales are/were too weak. Or maybe you're just upset at the high prices and poor availability of GPUs. One thing I can say for sure, this is really an emotional issue.
I don't have a dog in this fight, other than my general disappointment with AMD's *lack* of support for GPU-compute on consumer hardware, in recent years. We waited years for them to rebuild their GPU software stack, and then they screwed us by supporting only CDNA cards with it.
> AMD won't sell cards made for PC gaming
They *just* introduced the RX-6600 XT!
> a money-grubbing corporation that has absolutely no soul
I'm not even saying they're not. However, this doesn't align with your whole argument. With PC GPU prices so high and such strong demand for their CPU products, the profit motive would have them re-target their wafer allocation from consoles to these higher-margin products. So, that's clearly not in their power (which should go without saying, if one has an iota of business accumen).
> via delivering the least value to the consumer.
Um, no. They simply have to behave in profit-maximizing ways. Period. One way to do that is by reducing costs, but there's no direct mandate to minimize the value of their products.
*All* of these corporations are mandated to behave in profit-maximizing ways. Intel, Microsoft, and Nvidia, for sure. I can't really comment on Sony and Nintendo, since they're presumably Japanese-based and I don't really know anything about corporate governance laws and investor rights, in Japan.
> Astroturf the truth away.
Yeah, I'm sure anyone who disagrees with you must be a corporate shill. There can be no other possible explanation.
I had the same question, but I seem to recall someone found a source that indicated such.
Either way, if I'm Sony or MS, it'd be insane to leave my chip supply to the mercy of AMD. So, if AMD was in charge of getting the chips manufactured and packaged, then supply agreements will have been negotiated far in advance.
With something like a console, they will have had an idea of how the sales ramp would look and should've pre-purchased the necessary wafers to support it, probably with options to buy more. Now, where things get interesting is if/how any of the pandemic-fueled demand surge was satisfied. However, I'm guessing even those wafers were purchased not long after it became apparent that most of their target marked would be confined to their homes for the rest of 2020.
'AMD is bound by supply agreements. It's not their choice.'
Great defense, there.
AMD has no choice but to do things like Polaris-forever, Vega has identical IPC to the previous generation, refusal to stop making GPUs designed for mining whilst pretending they're for PC gaming, etc.
It has nothing to do with AMD being the central pillar in the console scam. Consoles are nothing more than PC gaming boxes in disguise, with machinations to create artificial demand — machinations AMD is very much a strong part of.
Regardless of whether you're a neutral enthusiast, or a fanboy of one firm or another... you've got to give a lot of credit to AMD. Their Phoenix-like rise from the ashes has reawakened competition and driven CPU performance way up in PCs and consoles alike. Intel is finally moving again, and AMD hasn't stopped moving since Zen. Things are interesting again.
In the GPU space things are a little more complicated. On the one hand, they have changed course for the better with RDNA and are rapidly becoming more competitive. On the other hand, unless AMD and Nvidia can figure out a REAL way to drastically limit mining performance without harming gaming performance, prices are going to stay massively inflated. A software-based solution isn't likely to work for very long, even IF Nvidia can manage to avoid driver leaks. There's a lot of money at stake, custom drivers will pop up.
Hi Gondalf, glad to see you've at least picked up some English lessons so your posts aren't so incoherent anymore, although the content is still significantly lacking.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
104 Comments
Back to Article
hlovatt - Tuesday, July 27, 2021 - link
Why doesn’t Anandtech any longer report Apple earnings when it still does so for other tech companies?Ashinjuka - Tuesday, July 27, 2021 - link
financials are ok but how are they /really/ doing with their fanboi futures? seems like intel is well ahead of AMD in the rabidity indices and mouth-to-foam ratios.lopri - Tuesday, July 27, 2021 - link
Shouldn't it be foam-to-mouth ratio?artifex - Tuesday, July 27, 2021 - link
That's how much cappuccino makes it to your lips.Azix - Tuesday, July 27, 2021 - link
AMD has the fanboy appealOxford Guy - Wednesday, July 28, 2021 - link
If one ignores the fact that the company actively works against PC gaming and has for many years.mode_13h - Thursday, July 29, 2021 - link
No, they don't. Try blaming Sony, Microsoft, and Nintendo.Oxford Guy - Thursday, July 29, 2021 - link
'No, they don't. Try blaming Sony, Microsoft, and Nintendo.'I will give you the benefit of the doubt and assume you're being intentionally illogical with such statements.
Qasar - Friday, July 30, 2021 - link
wait, so according to you, console makers are NOT actively working against pc gaming ?????????mode_13h - Sunday, August 1, 2021 - link
This being a PC-oriented site, we don't have a lot of console peasants here. So, trolling them would be a lot less successful than trolling AMD fans.Even if he were spot on, it's unclear what he hopes to accomplish by whining about it here. It's just tilting at windmills. That's yet another reason I think it's just a long-troll.
Oxford Guy - Tuesday, August 17, 2021 - link
'So, trolling them would be a lot less successful than trolling AMD fans.'An ad hom a day...
Oxford Guy - Tuesday, August 17, 2021 - link
AMD is pretending to be on the side of PC gamers by improving its Linux CPU support.I am reminded of Nvidia's anti-mining tech that was 'accidentally' broken.
AMD's current execs are good at pretending, playing multiple sides. They're better at it when they put out the FX 9590.
Atari2600 - Saturday, August 7, 2021 - link
Should have went to Cambridge I guess...mode_13h - Sunday, August 8, 2021 - link
:DOxford Guy - Tuesday, August 17, 2021 - link
I assume they teach the basic list of logical fallacies there, too.So, regardless where Mode would have gone... he/she would have likely not been able to get through courses on a wave of fallacies.
Here, though...
Samus - Thursday, July 29, 2021 - link
Are you seriously implying Apple doesn't have 'fanboy' appeal Azix? They probably have the largest cult following of any single company in the history of the universe.WaltC - Tuesday, July 27, 2021 - link
I have no curiosity whatsoever about cell-phone companies. Not even slightly interested.Duwelon - Tuesday, July 27, 2021 - link
Apple represents about 12% of the entire US stock market cap today. I'm a reluctant Apple owner but I'm interested in their stock price for more than just devices.WaltC - Wednesday, July 28, 2021 - link
To repeat, I have no interest in cell-phone companies.Reflex - Wednesday, July 28, 2021 - link
So no interest in Microsoft, Google, Samsung or LG either I guess...Pretty restrictive set of interests there. Virtually everyone in tech is into 'cell phones' on one level or another.
Samus - Thursday, July 29, 2021 - link
I think the reason AMD is being reported is because unlike ALL those other companies you mentioned, along with others, AMD has had a massive turnaround, and AMD is a chip company. None of the other companies make chips. Right now, the global focus is on chip production, hence all the talk of TSMC, Intel and now AMD over the last few months on AT.This isn't a financial website, and I don't expect them to report financials of every tech company. Just the ones that most specifically pertain to the spirit of Anandtech (which started 23 years ago, effectively, as a computer hardware review website focused on the Intel\AMD war of the late 90's\early 00's.
Reflex - Thursday, July 29, 2021 - link
I mean I don't really care, I am just referring to WaltC's ridiculous claim that Apple is just a cell company as a weird excuse to have 'no interest'.Oxford Guy - Tuesday, August 17, 2021 - link
'AMD is a chip company. None of the other companies make chips.'Is AMD making M1 for Apple or is it Intel?
arashi - Wednesday, August 18, 2021 - link
Nubile hermit virgins in the mountains of Appalachia hand manufacture every wafer for Apple.ikjadoon - Tuesday, July 27, 2021 - link
Right. Apple makes $10b a quarter in Mac sales, often beating or matching Dell, Lenovo, and HP.$10 billion revenue per quarter selling Macs, lmao.
Clearly, they’ve done quite well. Not to mention the M1 is a massive transition for the computing market, besides its near chart-topping performance-per-watt.
I’d like to see the next “cell phone company” that can break 7.10 on SPECint2017 this cycle. The ranking champ is the 5950X at 7.65, so bloody good luck.
WaltC - Wednesday, July 28, 2021 - link
I notice you avoid benchmarks where M1 get stomped...;)Ryan Smith - Tuesday, July 27, 2021 - link
We stopped reporting Apple results a couple of years back. When they stopped reporting unit sales counts and became more cagey about sales data in general, it really put a crimp on our ability to say much of interest about the data. Plus everyone reports on Apple, which makes it hard for us to stand out.This is actually the first time I've ever seen anyone ask about the matter. Is there demand here for Apple earnings coverage? If so, perhaps I need to look at re-adding that for future quarters.
hlovatt - Wednesday, July 28, 2021 - link
Thanks for replying. I would be interested, since I use Anandtech for my tech news.mode_13h - Wednesday, July 28, 2021 - link
It's not tech news, per se. If you want financial news, get it from a finance site.Anandtech already misses enough genuine tech news stories & reviews. Do you want them to miss even *more* of the stuff they're actually good at covering, just because you're too lazy to get Apple news from somewhere else?
mode_13h - Wednesday, July 28, 2021 - link
> Is there demand here for Apple earnings coverage?No. Just let them get Apple news from one of the zillions of other publications that fawn over the company.
mode_13h - Wednesday, July 28, 2021 - link
Of course, please keep up the fine coverage of their CPU developments.philehidiot - Wednesday, July 28, 2021 - link
Agree with this. AMD financials are interesting to a degree as it's such an impressive turnaround with real implications for end users. But, ultimately, if I want finance news, I'd go to a finance website.The occasional article makes sense when something stands out but, as someone else mentioned, I'd prefer time and energy put into tech stuff rather than trying to deliver financial analysis on all the major tech companies.
Blahla - Wednesday, July 28, 2021 - link
I created an account because I think its important to cover Apple's earnings. The company is the largest technology hardware player and even if it outright hazy about per unit sales, it deserves covering.Most of the time, your coverage may not be differentiated from the financial service guys. But I am pretty sure that you will find some areas of focus that will be different and add value.
Earnings news is not a big part of Anandtech. But I was pleasantly surprised when it began a few years ago. Apple is worth covering and business reasons for tech decisions are increasingly valuable and not well covered by either financial or tech media.
mode_13h - Thursday, July 29, 2021 - link
The already cover the parts of Apple that are interesting.Unashamed_unoriginal_username_x86 - Wednesday, July 28, 2021 - link
Oh hey, why have I never seen formatting here before?is it <b> writing b in <>? </b> or _underscores,_ or **double asterisks,** maybe [bold] this [/bold] or this or <bold> this but different... </bold>
Spunjji - Wednesday, July 28, 2021 - link
Now I want to know what succeeded 🤣mode_13h - Wednesday, July 28, 2021 - link
So, how did you write "this"? Did you try /slashes/ or what about `back-ticks` ?Heh, maybe we should start trying ANSI escape-sequences or VT100 control codes.
; )
arashi - Wednesday, August 18, 2021 - link
Let's see...<b>: <b>bold</b>
: bold
[bold]: [bold]bold[/bold]
arashi - Wednesday, August 18, 2021 - link
[ b ] and [ /b ] worksgijames1225 - Wednesday, July 28, 2021 - link
I would prefer those reporting resources remain allocated else. It's trivial to get reports on Apple from many other places.mode_13h - Wednesday, July 28, 2021 - link
Because Apple is more of a lifestyle company than a tech company.All joking aside, Apple is getting big into content, finance, etc. They've outgrown actual technology products, as they search for ways to grow their already enormous revenue.
sseemaku - Tuesday, July 27, 2021 - link
If you look at pc build forums, the results should be obvious. Everyone has Ryzen in their new builds! Still don't understand how Intel lost such a big lead, what were their engineers upto?ikjadoon - Tuesday, July 27, 2021 - link
I was shopping for a new system I wouldn’t want to build anything with Intel today plainly on its heat output alone. Not to mention Intel’s tactics and delays and lies and reheated junk for years.AMD has been killing it and I can only hope they’ll have many more quarters like these so they can really increase their total market share / volume. Intel has been, stubbornly, stuck at 70%+ for two decades. Let’s see a real fight here.
Up to engineering things that clearly weren’t working.
TristanSDX - Tuesday, July 27, 2021 - link
Do you think that AMD is great social company that will make cheapest possible cpu to make folks most beneficial ? Look at numbers again - they are increasing gross margin to level of Intel or more if there will be possibility. When they hit safe ground, then R&D will be decreased, and you will see similar reheated junk, as from Intel.ikjadoon - Wednesday, July 28, 2021 - link
Absolutely not. When that happens, it's time to switch. All companies go through "ah, we need to actually care this time, lol" phases and that's when you buy. Intel can change its tune, Qualcomm, Apple, etc. They all can.Anybody that has "loyalty" to billion/trillion-dollar corporations with millionaire/billionaire CEOs has more than a few screws loose.
Whoever makes the most compelling platform + CPU gets the nod. This isn't too complicated...
mode_13h - Wednesday, July 28, 2021 - link
> When they hit safe ground,I don't think they ever will. Not only will they have a resurgent Intel to contend with, but they'll also have to deal with ARM, RISC-V, the disappearance of the Chinese market, and eventually Chinese competitors reaching markets outside of China.
> then R&D will be decreased, and you will see similar reheated junk, as from Intel.
We don't know that. They don't have the same history as Intel of dividends and share buybacks.
It's possible that, if they plateau at some point, they'll face a lot of pressure from investors to do some of those things in lieu of further share growth. However, that will probably trigger the start of their next big decline.
eva02langley - Wednesday, July 28, 2021 - link
They are debt free and about to acquire the biggest player in the FPGA and microcontroller business.Not only this, but they have a dynamic that Intel cannot match due to their size and agility. Intel is now a conglomerate that can't be stern easily.
mode_13h - Wednesday, July 28, 2021 - link
Yeah and that's great, but they're in a fiercely competitive landscape, right now. They will plateau, at some point. And when they do, it'll be tough to stave off the parasites... er, I mean "investors".eva02langley - Wednesday, July 28, 2021 - link
Two different vision, two different strategy, two different companies.AMD always did it the right way. Claiming that AMD is now going to utilize Intel or Nvidia dirty practices when it is clearly not in their philosophy, is nothing more than you spreading nonsense.
While Nvidia and Intel are bribing studios and AIBs, AMD is releasing open source ecosystem... and from your point of view, these companies are all the same...
LordSojar - Tuesday, July 27, 2021 - link
Being run by MBAs who have no grasp of actual technology. Intel is on a POSSIBLE turnaround now, but it remains to be seen by their execution.eva02langley - Wednesday, July 28, 2021 - link
Their problem is their fabs and they are now making it a business. Their main competitor is not AMD, it is TSMC. If you think TSMC is just going to sit still and let Intel do their things, then you don't get it. Intel will never get the node advantage back, they are 3-4 years behind TSMC.They need TSMC to drop the ball for 3-4 years to catch them up, this is how bad it is. Putting an Engineer as CEO will not change this. Hope is not a strategy!
mode_13h - Wednesday, July 28, 2021 - link
> Still don't understand how Intel lost such a big lead, what were their engineers upto?It's been said many times: Intel has been extracting value from the company for many years. They spend a lot of their profits on dividends and share buy-backs, then turn around and do rounds of layoffs to keep costs down. The result can hardly be surprising.
TheinsanegamerN - Wednesday, July 28, 2021 - link
That's nto all they did though. Intel has also invested billions into the 10nm process trying to get it to work properly, and is still investing billiosn into further R+D and fab capacity.AMD got really lucky that intel ran into so many roadblocks with 10nm production.
Teckk - Thursday, July 29, 2021 - link
Doesn’t take away the good designs of Zen and chiplet architecture that Intel mocked to actually do the same thing going forward.TheinsanegamerN - Thursday, July 29, 2021 - link
Nice red herring. I never said it did, and what AMD has done with zen has no bearing on wht intel wa sspending money on this whole time.mode_13h - Sunday, August 1, 2021 - link
> what AMD has done with zen has no bearing on wht intel wa sspending money on this whole time.I'm not so sure about that. If Zen wasn't so strong, I think it wouldn't have created quite the same urgency, on Intel's part. Zen could turn out to be the best thing to happen to Intel, if the whole management shakeup and reallocation of funding wouldn't have happened (or at least to the same degree), otherwise.
Makste - Saturday, July 31, 2021 - link
It is not pure luck though. There's execution. Supposing intel had it's 10nm up and running in 2016, Zen and the chiplet architecture was still going to come out, it could have taken longer for adoption maybe, but all the advantages and disadvantages between the two were going to be compared against each other still, and AMD would still be a tough competitor with that zen uarchAzix - Tuesday, July 27, 2021 - link
Is the word frontloading? Their revenue will drop later as demand drops and competition heats up.eva02langley - Wednesday, July 28, 2021 - link
Next quarter will stay flat... - AMD.No drop here...
Teckk - Thursday, July 29, 2021 - link
Competition goes both ways :)mode_13h - Wednesday, July 28, 2021 - link
> AMD began initial shipments of their first CDNA 2 architecture-based> Instinct accelerators in Q2,
Of course, you wouldn't know this if you relied solely on Anandtech for tech news.
: (
> opening the spigot there for data center GPU revenue going into Q3.
Well, not exactly. They have a "software pipeline" problem that basically none of their consumer hardware is supported by their compute stack, so basically the only ones porting anything to it are AMD employees and probably a few grad students working as interns on the couple HPC wins AMD landed.
The genius move Nvidia made was to fully support CUDA on their entire hardware range, from top to (almost) bottom. Sometimes, their very lowest-end card doesn't support it, but even their Jetson embedded compute modules can run it.
Intel is doing one better than Nvidia, by supporting their compute APIs on CPUs, GPUs, FPGAs, and (where appropriate) hardware AI chips.
As long as AMD considers cloud and HPC as their only compute customers, adoption even in those segments is going to be limited. And as adoption of purpose-built AI hardware starts to gain momentum, AMD is going to see sales of their CDNA products hit a brick wall.
Ryan Smith - Wednesday, July 28, 2021 - link
"Of course, you wouldn't know this if you relied solely on Anandtech for tech news."There isn't anything to report on that front. AMD isn't saying a thing about CDNA 2 other than that they've shipped some parts to someone, somewhere.
mdriftmeyer - Wednesday, July 28, 2021 - link
Why would they reveal specific customer sales of the MI200? Have they done that with Radeon? No. The fact the product is ready and targeted for major compute markets, especially Supercomputer clients that have already been published should be a clue who is using them.Ryan Smith - Wednesday, July 28, 2021 - link
Oh, we're aware of where they're (likely) going. The point is more to underscore how little AMD is saying about CDNA 2 parts right now.mode_13h - Wednesday, July 28, 2021 - link
> There isn't anything to report on that front. AMD isn't saying a thing about CDNA 2Thanks for confirming. Not to sound ungrateful, but um, what about CDNA 1 (MI100) coverage? As I recall, there've also been a few other GPUs that launched without coverage on this site.
eva02langley - Wednesday, July 28, 2021 - link
The "GENIUS" Nvidia move is now biting them in the ass, because ampere is hosting double the cores and offer 40% performance uplift. They are facing a GCN moment where they don't have a gaming uarch anymore... and they lost the entire gaming market. Holding the dGPU business is nothing if studios don't use your hardware beside when you bribe them to implement your tech.Don't lure yourself, games are made on consoles now. FSR adoption will skyrocket with the console adoption to. Nvidia cornered themselves with DLSS and RTX to paint AMD as the duck, but now with FSR, it literally backfired royally.
Spunjji - Wednesday, July 28, 2021 - link
For a company that's "lost" the gaming market, Nvidia sure do seem to be doing just fine selling every gaming card they makemode_13h - Wednesday, July 28, 2021 - link
You can tell they're hurting by the aggressive price breaks. I mean, they can hardly be making any money on those GPUs!; )
mode_13h - Wednesday, July 28, 2021 - link
They do nerf some their consumer cards in some ways that are important for professional users.But, the key point is that you can still use consumer cards for development & dabbling with the whole Nvidia software stack. This is important for getting kids in university to port & optimize open source software for your hardware.
Much of the goodwill AMD built by having an open source driver stack, they've lost by stumbling so badly with lacking compute support for consumer hardware. Just checkout any discussion of ROCm on pretty much any site, and you'll find much gnashing of teeth about this.
Oxford Guy - Wednesday, July 28, 2021 - link
Goodwill for those who invested when the company was cheap. Goodwill for those who care about efficient CPUs and not high-performance gaming.Anti-goodwill for the few consumers who have a clue about things like:
— designing GPUs to sell to mining first and foremost
— selling junk at the ‘high end’ so Nvidia can raise prices, the junk that didn’t go to miners will be purchased by foolish ‘team red’s, and the high prices of PC gaming will sell more PCs in disguise — that AMD happens to allocate so many wafers to.
— sandbagging for many years at the midrange and above areas, to, additionally, raise prices with Nvidia via pent-up demand, not be so obvious in the various anti-PC enthusiast schemes (once actually competitive cards are deigned to be produced — just by chance during the great shortage), and be hailed as amazing for the technical wizardry of actually bothering to release something competitive for the first time since Hawaii.
— Goodwill for those invested in MS and Sony.
Spunjji - Thursday, July 29, 2021 - link
This comment entails the assumption that designing GPUs to compete with Nvidia at the high-end is relatively trivial and something they just decided not to do.It's not. Just ask Intel, 3DFx, Img tech, Matrox, etc...
Oxford Guy - Thursday, July 29, 2021 - link
Vega, which took a long time to hit the market had identical IPC vis-à-vis Fury X. Identical.Zero improvement.
There are no credible excuses for that and for plenty of other very substantive examples that support my case.
Qasar - Thursday, July 29, 2021 - link
" There are no credible excuses for that and for plenty of other very substantive examples that support my case. "oh ? then post them on here so we can see them too, if not, its STILL just more BS
mode_13h - Sunday, August 1, 2021 - link
> Vega, which took a long time to hit the market had identical IPC vis-à-vis Fury X. Identical.That's not really true. They did make some ISA-level improvements.
http://developer.amd.com/wordpress/media/2017/08/V...
Search for "Differences Between VEGA and Previous Devices", and remember that Polaris came between Fury and Vega. So, for a more complete list of changes, you'd probably also have to consult the Polaris docs.
But that's just the ISA doc. GPU performance is not only about IPC. Vega has other differences, highlighted in the review:
* New memory page management for the high-bandwidth cache controller
* Tiled rasterization (Draw Stream Binning Rasterizer)
* Increased ROP efficiency via L2 cache
* Improved geometry engine
* Primitive shading for even faster triangle culling
* Direct3D feature level 12_1 graphics features
* Improved display controllers
https://www.anandtech.com/show/11717/the-amd-radeo...
I also remember seeing a few benchmarks where Vega 56 had up to a 2x improvement over Fury.
> There are no credible excuses for that
Vega was dual-focused at gaming and machine learning.
Also, by fabbing it on a smaller process node, AMD was able to turn up the clock speed.
With all that said, I don't want to defend Vega as if it was actually good. Clearly, it was lacking in some key respects. I think its origins were from a time when AMD still thought the GPU race was mostly about brute-force. RDNA was obviously an attempt to course-correct.
BTW, I don't really see your point of calling out Vega, specifically. If you go back through the entire GCN range, each was an evolutionary improvement on the previous generation. Vega was entirely consistent with that.
Oxford Guy - Tuesday, August 17, 2021 - link
'That's not really true. They did make some ISA-level improvements.'Just like Parhelia was a great competitor because it had things like glyph smoothing.
A laundry list of mostly superfluous features is irrelevant to the point. Review sites that did clock-to-clock gaming tests found that Vega's IPC is identical — and there is no excuse for that.
What there is is AMD purposefully not being competitive beyond the low midrange so Nvidia could pump prices up and create artificial demand for 'consoles'. It also kept Polaris-forever overpriced and selling.
mode_13h - Thursday, July 29, 2021 - link
> — designing GPUs to sell to mining first and foremostThis just shows the depth of your ignorance on the subject. If AMD cared about miners, they'd actually support GPU compute on those cards!
I know you don't *really* care about facts, if they get in the way of your scapegoating.
That's the biggest problem with your whole conspiracy theory -- it simply doesn't make sense. I get that you're angry and searching for somewhere to target your rage, but you've gotta get past that.
Oxford Guy - Thursday, July 29, 2021 - link
'This just shows the depth of your ignorance on the subject.'Don't expect someone to bother reading your posts when you start them off with empty rhetoric like that. Your flamboyant ad hominem is tiresome.
Qasar - Thursday, July 29, 2021 - link
and your crying and whining about the console scam BS, is tiresome as well, point is ?Oxford Guy - Tuesday, August 17, 2021 - link
'and your crying and whining about the console scam BS'Try posting a rebuttal.
mode_13h - Sunday, August 1, 2021 - link
> Your flamboyant ad hominem is tiresome.Same.
As for empty rhetoric, I've yet to see much else from you. When you do make a case, it's typically weak and not backed by sources or references. The amount of shade you cast, relative to the informational content of your posts is quite high. That's more like a caricature of Oxford than the reality.
On the internet, the only thing that matters is the strength of your argument. I don't care if you're an Oxford prof or a homeless dropout. Since we can't verify your background, it counts for nothing.
Oxford Guy - Tuesday, August 17, 2021 - link
'Same.'Tu quoque fallacy.
'I don't care if you're an Oxford prof or a homeless dropout. Since we can't verify your background, it counts for nothing.'
Another ad hom.
Do you have nothing else in your repetoire?
mode_13h - Wednesday, July 28, 2021 - link
Lol, FSR. Good on them for making it open and supporting other hardware, but it's no DLSS.mode_13h - Wednesday, July 28, 2021 - link
They're going to have to do better. Probably not much more is possible on current hardware, though.Thunder 57 - Wednesday, July 28, 2021 - link
AMD is playing a bit of "catch up" in the GPU realm as they devoted most of their resources to Zen, rightfully so, as it probably saved them from bankruptcy. The real question is can they catch up NVIDIA. WCCFTech covered a rumor that RDNA3 will be better than Lovelace, but it is WCCFtech, so who knows.They are close to NVIDIA with RDNA2 if you exclude extras like RT and DLSS. NVIDIA got the head start though so AMD is playing catch up there.
Oxford Guy - Thursday, July 29, 2021 - link
'AMD is playing a bit of "catch up" in the GPU realm as they devoted most of their resources to Zen, rightfully so,'It would have bankrupted the company to not release Vega (after such a long delay) with identical IPS to Fury X.
(eyeroll)
Wake up and realize that AMD actively works against the PC gaming market.
It makes mining GPUs and allocates wafers for the 'console' scam. It actively works to keep prices too high in the PC gaming market in a variety of ways, in order to make the 'console' scam function.
It also exploits the dire lack of competition in GPUs, where desperate stupid people will buy its products due to emotional nonsense — like 'It's my donation to keep competition alive!' and 'I'm team red because I like the underdog' — and also because Nvidia keeps prices very high (thanks to help from AMD) so AMD can sell things like 'Polaris forever'.
AMD is also the company that broke AVX when it 'improved' upon Bulldozer (with the Piledriver release) and had the audacity to release the 9000 series, with extremely high prices based on falsehoods like 5 GHz and 8 cores.
It didn't bother to provide even a shred of value to gamers, only hype and lies. The single-thread IPC of Piledriver was so terrible that even at over 5 GHz (don't ask how much power that required, or the cost of the board, or the cost of the cooling, or the noise) it was too slow to prevent lag problems. It could have replaced Piledriver with something that would better justify the extremely high price it put on the 9000 series but... nope. Maximum ridiculousness. Maximum cravenness.
What does it take for people to stop peddling AMD's propaganda? I suppose it takes not having investments in one or more of the following: AMD, Nvidia, Sony, Microsoft... and even Nintendo.
Even Nintendo is helped by the console scam because if there would be robust demand for the PC gaming platform and affordable hardware to go with it that's not locked behind a ton of duplicate walled gardens (aka the 'consoles'), then portable PC gaming systems — recognized for what they are — x86 bog-standard PC gaming hardware ... that even use the same substandard craven joystick mechanism ... Nintendo would have a harder time selling its overpriced substandard hardware complete with its own superfluous parasitic walled garden.
AMD has stated it had no intention of reducing the attractiveness of its cards for mining.
These are the sort of bald facts that jokers on this site like to ignore in favor of ad hominem and flamboyant specious braggartry.
AMD won't sell cards made for PC gaming, rather than cards for mining and wafers for consoles. How can that be? AMD is some sort of benighted altruistic superhero... not a money-grubbing corporation that has absolutely no soul and, like any corporation, is designed to extract maximum profit via delivering the least value to the consumer. Sell less for more. Astroturf the truth away.
Oxford Guy - Thursday, July 29, 2021 - link
'It would have bankrupted the company to not release Vega (after such a long delay) with identical IPS to Fury X.'Oxford Guy - Thursday, July 29, 2021 - link
'It would have bankrupted the company to not release Vega (after such a long delay) with identical IPC to Fury X.'IPC... Ugh. It took three posts but that was the original idea.
mode_13h - Sunday, August 1, 2021 - link
> identical IPC to Fury XAnd again, this is you not understanding GPU performance. Vega 64 performed a lot better than Fury X, because GPU performance is not simply a function of IPC.
In point of fact, the entire range of GCN GPUs, spanning all the way from the HD 7000-series (launched in 2012) to Radeon VII (launched in 2019), had basically the same instruction throughput per CU per clock.
Qasar - Thursday, July 29, 2021 - link
" allocates wafers for the 'console' scam. "do you have link that shows its AMD's allocation of wafers, and not sony/MS that is buying the wafers ??
are you forgetting the reports that nvidia sold cards directly to minors ? you ONLY seem to be pointing the finger at AMD, while forgetting the prices nvidia and intel also charged for its products.
i guess cause you hate " a ton of duplicate walled gardens " i guess Apple's whole product line, makes you catatonic
as i said, post links to this BS. so far, it is ONLY your opinion.
Oxford Guy - Tuesday, August 17, 2021 - link
'do you have link that shows its AMD's allocation of wafers, and not sony/MS that is buying the wafers ??'Irrelevant.
GeoffreyA - Sunday, August 1, 2021 - link
Oxford Guy, I will look into this, but did AVX break or was it slower? Also, we know Bulldozer was a disaster: if they kept on churning out Bulldozer variants today, we've got licence to criticise, but the fact is, they made an architecture from scratch, so excellent that it's turning on the heat beneath Intel's bum; in fact, scorched it. And they've suffered enough from Bulldozer already, so I don't think it's fair to attack them for a mistake from their past.GeoffreyA - Sunday, August 1, 2021 - link
"Memory writes with the 256-bit AVX registers are exceptionally slow. The measured throughput is 5 - 6 times slower than on the previous model (Bulldozer), and 8 - 9 times slower than two 128-bit writes."You're right about Piledriver's AVX: it was slow enough to be considered broken. 256-bit stores were the problem, taking 17-20 cycles. However, Steamroller mended that, down to 2 cycles for aligned 256-bit stores and 3 for unaligned ones (faster than Bulldozer's 3 and 10). Excavator, no regression in that regard.
https://www.agner.org/optimize/blog/read.php?i=285...
https://www.agner.org/optimize/microarchitecture.p...
mode_13h - Sunday, August 1, 2021 - link
> You're right about Piledriver's AVX: it was slow enough to be considered broken.Interesting. Thanks for the info. Irrelevant to any topic at hand (not your fault, I mean the one who raised it), but still interesting.
mode_13h - Sunday, August 1, 2021 - link
BTW, the fact that it was such a huge regression makes it sound like a hardware bug. Maybe fixing it would've pushed it too close to the market window of Steamroller and not many things were using AVX at the time, anyhow?GeoffreyA - Sunday, August 1, 2021 - link
The block diagram could shed some light, or maybe not, but I'm going to guess that when they changed the FP backend, possibly adding more units, they were out of time and locked down AVX writes to a certain speed, which they knew worked. Then in Steamroller, had more time to validate and get it working at its right speed. AVX was pretty new back then, so no big loss. A bit slapdash? I suppose. Or: "It'll be fixed in Steamroller, so don't worry, my Piledriver comrades. You know, the Steamroller guys are busy tackling 256-bit stores as we speak. Got an email from the lead architect this morning."If you remember Zen 1, they set the L2 cache at 17-cycles latency because they were out of time. Then in Zen+ (and I think Raven Ridge), chopped it down to 12 cycles.
mode_13h - Sunday, August 1, 2021 - link
> Vega ... with identical IPS to Fury X.I dealt with this, above.
> It makes mining GPUs
All GPUs are mining GPUs. Nvidia's flailing attempts to make its Ampere GPUs *bad* at mining is a perfect example of how hard it is, when it's in fact certain cryptocurrencies that are specifically designed to run well on GPUs.
> allocates wafers for the 'console' scam.
AMD is bound by supply agreements. It's not their choice.
> It actively works to keep prices too high in the PC gaming market in a variety of ways
How?
> in order to make the 'console' scam function.
What's their motivation for this, given that they could be making more $ if they had the freedom to re-target those wafers? This is the critical flaw in your argument, even before we get to the part where you provide zero evidence to back it up.
> had the audacity to release the 9000 series, with extremely high prices
> based on falsehoods like 5 GHz and 8 cores.
It's just the same thing we've seen out of Intel. When these companies are stuck in a product line longer than they want to be, it pushes them to take desperate measures, like cranking up clockspeed.
> The single-thread IPC of Piledriver was so terrible
The two interviews Ian recently did with Jim Keller shed some very interesting insight into how much he had to transform the culture of AMD from one of defeatist incrementalism, in order to make Zen happen. If you haven't read those, I'd recommend them.
What I don't understand is how you think this failure to execute is somehow evidence of a conspiracy. It also reveals quite a lot about the strength of your argument (or lack thereof) that you have to go so far back to make your case against AMD.
> Even Nintendo is helped by the console scam
Here's the part where you're bad at conspiracy theories. Any good conspiracy theory treats incentive almost as proof of complicity. Instead of launching on an unrelated tangent about Nintendo, you should find some way of tying Nintendo more deeply into the whole thing.
> AMD has stated it had no intention of reducing the attractiveness of its cards for mining.
Unlike Nvidia, they don't have a separate mining-oriented product line. That's why. Both companies would be punished by their investors (and probably even sued), if they went out of their way to shut out miners and that revenue stream. That's wholly different than saying they set out to build good mining cards.
And lots of people use their cards for mining when they're not gaming, in order to offset the hardware costs. Of course, that only works of you're in a cheap electricity market (or someone else is paying the electric bill, which I'm sure is the case for most kids).
What's weird is that you didn't even touch on AI. That's the one case where there's actual evidence that both companies juiced their hardware for AI. And it's surely consuming quite a bit of Nvidia's die allocation, although I don't think many people are using AMD GPUs for AI.
HPC is another area you could attack. Vega 20 (the 7 nm version) added fp64, in addition to further AI-targeted enhancements. Their CDNA chips (MI100 and now MI200) take these themes even further. And Nvidia is targeting enormous A100 GPUs at HPC, as well as AI. Why go after consoles and not HPC?
> These are the sort of bald facts that jokers on this site like to ignore
They're tenuously-related and really don't support your claims. So much "empty rhetoric", as someone recently said.
Again, I can see that we're beyond facts. You're starting with a scapegoat and trying to cherry-pick facts to support your case. You clearly feel something is deeply lacking in the PC gaming market. You want to recapture some lightning in a bottle, from a bygone era. Maybe it's the lack of better titles. Maybe you work/worked for a indie game developer and sales are/were too weak. Or maybe you're just upset at the high prices and poor availability of GPUs. One thing I can say for sure, this is really an emotional issue.
I don't have a dog in this fight, other than my general disappointment with AMD's *lack* of support for GPU-compute on consumer hardware, in recent years. We waited years for them to rebuild their GPU software stack, and then they screwed us by supporting only CDNA cards with it.
> AMD won't sell cards made for PC gaming
They *just* introduced the RX-6600 XT!
> a money-grubbing corporation that has absolutely no soul
I'm not even saying they're not. However, this doesn't align with your whole argument. With PC GPU prices so high and such strong demand for their CPU products, the profit motive would have them re-target their wafer allocation from consoles to these higher-margin products. So, that's clearly not in their power (which should go without saying, if one has an iota of business accumen).
> via delivering the least value to the consumer.
Um, no. They simply have to behave in profit-maximizing ways. Period. One way to do that is by reducing costs, but there's no direct mandate to minimize the value of their products.
*All* of these corporations are mandated to behave in profit-maximizing ways. Intel, Microsoft, and Nvidia, for sure. I can't really comment on Sony and Nintendo, since they're presumably Japanese-based and I don't really know anything about corporate governance laws and investor rights, in Japan.
> Astroturf the truth away.
Yeah, I'm sure anyone who disagrees with you must be a corporate shill. There can be no other possible explanation.
Qasar - Sunday, August 1, 2021 - link
" > allocates wafers for the 'console' scam.AMD is bound by supply agreements. It's not their choice. "
but is there anything that shows its its AMD's wafers, or sony/MS ??
mode_13h - Monday, August 2, 2021 - link
I had the same question, but I seem to recall someone found a source that indicated such.Either way, if I'm Sony or MS, it'd be insane to leave my chip supply to the mercy of AMD. So, if AMD was in charge of getting the chips manufactured and packaged, then supply agreements will have been negotiated far in advance.
With something like a console, they will have had an idea of how the sales ramp would look and should've pre-purchased the necessary wafers to support it, probably with options to buy more. Now, where things get interesting is if/how any of the pandemic-fueled demand surge was satisfied. However, I'm guessing even those wafers were purchased not long after it became apparent that most of their target marked would be confined to their homes for the rest of 2020.
Oxford Guy - Tuesday, August 17, 2021 - link
'AMD is bound by supply agreements. It's not their choice.'Great defense, there.
AMD has no choice but to do things like Polaris-forever, Vega has identical IPC to the previous generation, refusal to stop making GPUs designed for mining whilst pretending they're for PC gaming, etc.
It has nothing to do with AMD being the central pillar in the console scam. Consoles are nothing more than PC gaming boxes in disguise, with machinations to create artificial demand — machinations AMD is very much a strong part of.
Alexvrb - Sunday, August 8, 2021 - link
Regardless of whether you're a neutral enthusiast, or a fanboy of one firm or another... you've got to give a lot of credit to AMD. Their Phoenix-like rise from the ashes has reawakened competition and driven CPU performance way up in PCs and consoles alike. Intel is finally moving again, and AMD hasn't stopped moving since Zen. Things are interesting again.In the GPU space things are a little more complicated. On the one hand, they have changed course for the better with RDNA and are rapidly becoming more competitive. On the other hand, unless AMD and Nvidia can figure out a REAL way to drastically limit mining performance without harming gaming performance, prices are going to stay massively inflated. A software-based solution isn't likely to work for very long, even IF Nvidia can manage to avoid driver leaks. There's a lot of money at stake, custom drivers will pop up.
Oxford Guy - Tuesday, August 17, 2021 - link
'you've got to give a lot of credit to AMD'I'm sure the people who paid full price for the FX 9590 agree... whilst they're playing on their 'console'.
arashi - Wednesday, August 18, 2021 - link
Hi Gondalf, glad to see you've at least picked up some English lessons so your posts aren't so incoherent anymore, although the content is still significantly lacking.