Site seems to only be readable on mobile if you force PC mode display. Otherwise if you try to read pass the abstract the article gets replaced with a message to download their app.
It's also 8 pages so they can show a different ad on each page. The content might be OK, but it's the same slimy tactic that Facebook listicle sites use to boost "page views" and ad views. To me, this massively devalues the content.
Other sites use identical strategies except they refresh an ad slot every 15 seconds. The 8 page thing is either bad/weird web design or serves another purpose.
GPU companies seem to be notoriously closed and in general secretive about their products; nVidia is only one of the more prominent current examples, but AMD/ATI wasn't that nice in the past either. AFAIK Intel was the first to release detailed specs on its GPUs.
Ideally, one should be able to go to their websites and download full documentation (and drivers with source) on any of their products, as well as actually order the bare GPUs chips. Instead, thanks to their desire to control the entire manufacturing and distribution chain, it's a convoluted NDA-filled process. Personally, I don't really get it; why would a company put all these barriers in place to would-be customers of their products, and thus restrict their revenue? I'm almost willing to bet that there would be many more buying and using their GPUs if they left the market more open.
Microchip is one prominent example of the "other way" --- full documentation for their microcontrollers, and even direct ordering from the website. They are very successful. Unfortunately no GPU manufacturer seems to be doing this.
NVidia and ATI/AMD experienced more than a decade as a duopoly where everyone who wanted a powerful graphics chip bought one from one of those two companies. Their only big growth opportunities were to expand the number of gamers. Open specs don't help much with that.
In the meantime, computer graphics has long been a patent minefield, and Hollywood has been pressuring the chip makers to play along with their DRM schemes. Both of those are strong incentives to not be fully open.
Their only big growth opportunities were to expand the number of gamers. Open specs don't help much with that.
It would increase the number of companies who could produce their own designs and thus diversity the market; and GPGPU might've come sooner if the documentation was far more widely available.
> It would increase the number of companies who could produce their own designs
Their own designs for competing GPUs, or their own designs for graphics cards incorporating NVidia and ATI GPUs? There's never been a shortage of the latter, and there's never been an incentive for NVidia or ATI to help enable the former.
I don't think GPGPU would have happened much earlier, but I do think open hardware specs would have drastically changed the evolution of GPGPU tooling. There would have been upstream backends for GCC and LLVM before closed frameworks like CUDA could have gotten a foothold. But none of that was going to happen until the GPUs were actually capable of running general purpose code.
I was rather hoping one day GPU could be used as simple as CPU, and Drivers is barely needed. The problem with today's graphics pipeline is the amount of code and work required in Drivers. In the past have many GPU vendors, most of them could not get their drivers working right or performance enough. And Nvidia and AMD has no incentive of opening up, because their years of work in the drivers set are a barrier of entry for anyone who wants to step in to make a decent GPU.
May be Intel's forthcoming GPU will be a little more opened? May be Apple's has a Metal GPU in the making for Mac?
This seems like no big deal. Nvidia creates a preferred partner co-branding program, and insists that an OEM’s half of the branding effort be Nvidia-specific, presumably so Nvidia isn’t spending marketing dollars promoting OEM brands that include AMD GPUs.
This program goes beyond branding efforts, according to the report[1] the partner program includes:
- high-effort engineering engagements
- early tech engagement
- launch partner status
- game bundling
- sales rebate programs
- social media and PR support
- marketing reports
- Marketing Development Funds (MDF)
Conditionally including MDF as part of this program is behavior similar to what caused the government action against Intel by the FTC:
"The FTC settlement applies to Central Processing Units, Graphics Processing Units and chipsets and prohibits Intel from using threats, bundled prices, or other offers to exclude or hamper competition or otherwise unreasonably inhibit the sale of competitive CPUs or GPUs. The settlement also prohibits Intel from deceiving computer manufacturers about the performance of non-Intel CPUs or GPUs."[2]
Yes, but Intel still didnt pay the fine (in EU) and keeps doing the same old under new clothes. As a reseller you regularly get extraordinary deals from Intel. For example for every highend CPU + mobo combo sold intel "gifts" you an SSD = hidden heavy rebate on expensive CPU, had this one at least couple of times.
It's up to those companies if they want to go that far and sign on with GPP then, which seems fine to me. I'm not a fan of these brutal tactics on NV's part, but I can't say I have strong feelings on it. It's not illegal so it's either worth it to the vendors or it's not sufficient enough to be.
Disclaimer: I'm not really partial to any of these companies (you can hardly state anything without some fan of either coming after you). I started with a Commodore, then mostly Matrox & S3 graphics from '89 till the 3dfx Voodoo 4MB ('95), used Nvidia from 00-07 and AMD from 09-17, now I'm back on NV and kind of regretting it. Mostly because of driver issues, after a good long pretty much problem free stint on my last Radeon that lasted me a whopping 8 years. I've purchased dozens of NV cards, by the most from any vendor that I've used.
If anything, my favorite is 3dfx since seeing GLQuake in 1995 on a Voodoo card is an experience that simply can't be duplicated. Everything is a letdown and iterative since then. I'm thankful I could experience that at the time when it was a meaningful improvement. Zeus knows gameplay hasn't improved much since the 90s. But I certainly miss my nice 8 year stint of updating to each WHQL release without major issues and have zero reservations about going back to AMD.
You still can see that something anti-competitive is happening here, a company has a dominant position and abuses it, ASUS,MSI and the others have a free choice between selling NVIDIA products with all the benefits or being pushed at the end of the queue so probably always having not enough GPUs to sell, no discounts etc, so basically have a free choice between continuing to make money or getting a bit financial hit, it is a bad choice and a for profit company will chose profit not the right thing to do.
The situation got so ridiculous that AMD cards were removed from gaming section on the sellers pages, one such seller commented that AMD cards are not for gaming even if they themselves packaged it with the name "AMD gaming"
And other big issue is that this GPP program that NVIDIA calls it transparent is under an NDA and no one come forward with details to calm things down.
It is not AMD vs NVIDIA but about preserving this poor competition, we only have 2 players, let keep as it is if we can't have more.
This does feel like the equivalent of Microsoft's old "pay way less per copy if you just pay us for every computer you ship instead of pay us for every computer that actually ship's with Microsoft". I was at microsoft when that was going (working as a dev with nothing to do with this). but that was really effective at getting companies to be less interested in shipping other OS'.
Between this and the "no GeForce usage in datacenters" thing, I'm really starting to dislike NVidia.
I was hoping AMD would be able to be more competitive with them on both the gaming and deep learning fronts, but manufacturers already seem to be going along with this which could hurt AMD even more.
NVidia has been pretty crappy for a while now, they've made some exceptional products but they've also been complete dicks to gamers at the same time (see: GameWorks, along with their general history of implementing driver hacks to improve performance in games while reducing correctness).
I was really hoping AMD could be more competitive this generation too, I had only used AMD GPU's since 2011 when I built my first gaming system from scratch (as opposed to the cobbled up upgrades I had when I was still in middle/high-school). My R9 290 was looking long in the tooth, Vega failed to impress and the RX 580 was barely an upgrade (more VRAM, lower power, and due to cryptomining horrendously overpriced) - I ended up getting an ASUS Strix 1080Ti.
All of us should hope that AMD at least catches up with Navi, the current dominance of NVidia in the GPU market (both for gaming and the datacenter) is bad for everyone - even if you're loyal to them for some reason no real competition doesn't give them much reason to keep innovating.
yields:
"And, our friends at EPIC, with collaboration from ILMxLAB and NVIDIA, have also put together a stunning technology demo with some characters you may recognize."
Nvidia being dicks to gamers because of Gameworks is like saying Apple are dicks to business people because of iOS. Both Nvidia and Apple create software products and only want their products to be run on their walled garden. The fallout of Gameworks wasn't even that bad as Nvidia released the source code under a restricted license and allowed AMD to optimize their drivers for the API. So like... what...?
Given the stated goal of Gameworks was to get usage out of high end GPUs using the latest effects in games that were otherwise ported poorly from console I think it has really been delivering. Now we have the opposite problem where these special effects can genuinely be hard to run on anything but the latest and greatest. It is a better situation than the alternative where you could 4x SSAA a console ported game and max out the FPS.
As long as AMD can make a card that is ballpark competative with nVidia on price/performance, I'll buy AMD. nVidia drivers, at least in my experience, have been pretty awful for awhile now. I hot-swap displays daily on a 7970 rig with no issues. I am terrified to swap the hdmi or dvi cables on my GTX 980 because it will completely screw up the display config and I'll spend an hour fixing it. Both rigs are due for upgrades, and we'll be going with AMD for both. nVidia just isn't good enough on any level for me to support this kind of tactic.
> In the past this strategy backfired for Intel (INTC) and the same may be the fate of Nvidia.
It sure as shit didn't backfire for Intel - they settled with AMD for $1,000,000,000.
For this price they completely hobbled AMD's ability to sell the Athlon 64, which hobbled profits (both directly and in terms of long-term branch image), which ultimately hobbled R&D.
AMD never fully recovered - and sold their foundries in order to survive.
Intel, on the other hand, got full control over the pricing of pretty much the entire x86 market for over a decade - it was well worth what they settled for.
Bullshit. The EU has exactly the same powers as the US government regarding antitrust enforcement against multinationals corporations.
But, as in the US or any other society ruled by law, antitrust fines and all other executive decisions can be challenged in court. That’s what Intel did, and the case is still pending.
The cost of rebates to HP and Dell far exceeded the fine as they made up the majority of those companeis revenue at one point. The fines were an absolute bargain especially since Intel hasn't even paid the EU one yet.
Allowing politicians so heavily regulate business owners has put the EU where it is: in big trouble. It's going to continue to break up. Poland and Hungry are probably next.
Well I for one will certainly avoid any product brand that is now part of the GPP. It is great getting discounts from things like Intel but like the age old saying, if it's too good to be true it probably is.
And this definitely got exposed with the recent serious vulnerabilities in all modern Intel x86 CPUs. In the attempt to get ever faster and make ever more sales, they had to lower quality.
Pretty much the same path with Nvidia in my opinion.
>However, OEM computer manufacturers are a whole different story altogether. This is where AMD might get hurt by GPP. You see, GPP doesn't only apply to AIB manufacturers, it also applies to OEMs such as Dell, Apple, Lenovo or HP.
Perhaps in theory (emphasis on the perhaps, IANAL and neither is the author), but not in practice. Nvidia has a great deal of power over ATI, Asus etc. They have no power over Dell, much less Apple. Nvidia needs Dell and Apple far, far more than Dell or Apple needs Nvidia (they don't).
Nvidia and AMD are selling, if not identical products, then near enough for government work. If Dell decided to drop Nvidia from all their product lines, the impact on Dell would be negligible at worst, but it would be a catastrophe for Nvidia (while AMD would be making out like Scrooge McDuck). Apple? Nvidia sends Apple the corporate equivalent of drunken post-breakup apology voicemails (releasing Mac drivers for cards that aren't offered on Macs).
Nvidia's ability to enforce branding guidelines on Dell is nil. If Dell decided they'd rather brand "Nvidia" cards as "Dell GraphicsMaster 6000 SUX," Nvidia would have to roll with it.
>Nvidia and AMD are selling, if not identical products, then near enough for government work.
This is an incredibly ignorant statement. While they both sell GPUs for consumer, and business (Quadro/Fire) AMD sells semi-custom CPUs, and has merged CPUs & GPUs into APUs. NVDA sells SOC, and AI focused hardware in addition to GPUs. The GPUs have huge differences between them, particularly with the newer APUs which are more, or less, customized for OEMs.
Summary: they overlap on GPUs, but their businesses are very different elsewhere.
59 comments
[ 4.4 ms ] story [ 110 ms ] threadIdeally, one should be able to go to their websites and download full documentation (and drivers with source) on any of their products, as well as actually order the bare GPUs chips. Instead, thanks to their desire to control the entire manufacturing and distribution chain, it's a convoluted NDA-filled process. Personally, I don't really get it; why would a company put all these barriers in place to would-be customers of their products, and thus restrict their revenue? I'm almost willing to bet that there would be many more buying and using their GPUs if they left the market more open.
Microchip is one prominent example of the "other way" --- full documentation for their microcontrollers, and even direct ordering from the website. They are very successful. Unfortunately no GPU manufacturer seems to be doing this.
In the meantime, computer graphics has long been a patent minefield, and Hollywood has been pressuring the chip makers to play along with their DRM schemes. Both of those are strong incentives to not be fully open.
It would increase the number of companies who could produce their own designs and thus diversity the market; and GPGPU might've come sooner if the documentation was far more widely available.
Their own designs for competing GPUs, or their own designs for graphics cards incorporating NVidia and ATI GPUs? There's never been a shortage of the latter, and there's never been an incentive for NVidia or ATI to help enable the former.
I don't think GPGPU would have happened much earlier, but I do think open hardware specs would have drastically changed the evolution of GPGPU tooling. There would have been upstream backends for GCC and LLVM before closed frameworks like CUDA could have gotten a foothold. But none of that was going to happen until the GPUs were actually capable of running general purpose code.
There are SIGRAPH papers from 2003 about how to use GPU Assembly with texture data for doing non graphics related computations.
Nvidia still deploys a binary blob (which might work with your kernel or not)
May be Intel's forthcoming GPU will be a little more opened? May be Apple's has a Metal GPU in the making for Mac?
[1] https://en.wikipedia.org/wiki/Larrabee_(microarchitecture)
That won't help anyone cause sure as hell, Apple won't allow you to order the chips from their site.
- high-effort engineering engagements
- early tech engagement
- launch partner status
- game bundling
- sales rebate programs
- social media and PR support
- marketing reports
- Marketing Development Funds (MDF)
Conditionally including MDF as part of this program is behavior similar to what caused the government action against Intel by the FTC:
"The FTC settlement applies to Central Processing Units, Graphics Processing Units and chipsets and prohibits Intel from using threats, bundled prices, or other offers to exclude or hamper competition or otherwise unreasonably inhibit the sale of competitive CPUs or GPUs. The settlement also prohibits Intel from deceiving computer manufacturers about the performance of non-Intel CPUs or GPUs."[2]
1: https://www.hardocp.com/article/2018/03/08/geforce_partner_p...
2: https://www.ftc.gov/news-events/press-releases/2010/08/ftc-s...
The main problem was not the gifts or discounts for selling Intel CPU, there was conditions like do not sell any AMD or less then 5% AMD.
Disclaimer: I'm not really partial to any of these companies (you can hardly state anything without some fan of either coming after you). I started with a Commodore, then mostly Matrox & S3 graphics from '89 till the 3dfx Voodoo 4MB ('95), used Nvidia from 00-07 and AMD from 09-17, now I'm back on NV and kind of regretting it. Mostly because of driver issues, after a good long pretty much problem free stint on my last Radeon that lasted me a whopping 8 years. I've purchased dozens of NV cards, by the most from any vendor that I've used.
If anything, my favorite is 3dfx since seeing GLQuake in 1995 on a Voodoo card is an experience that simply can't be duplicated. Everything is a letdown and iterative since then. I'm thankful I could experience that at the time when it was a meaningful improvement. Zeus knows gameplay hasn't improved much since the 90s. But I certainly miss my nice 8 year stint of updating to each WHQL release without major issues and have zero reservations about going back to AMD.
The situation got so ridiculous that AMD cards were removed from gaming section on the sellers pages, one such seller commented that AMD cards are not for gaming even if they themselves packaged it with the name "AMD gaming"
And other big issue is that this GPP program that NVIDIA calls it transparent is under an NDA and no one come forward with details to calm things down.
It is not AMD vs NVIDIA but about preserving this poor competition, we only have 2 players, let keep as it is if we can't have more.
I was hoping AMD would be able to be more competitive with them on both the gaming and deep learning fronts, but manufacturers already seem to be going along with this which could hurt AMD even more.
I was really hoping AMD could be more competitive this generation too, I had only used AMD GPU's since 2011 when I built my first gaming system from scratch (as opposed to the cobbled up upgrades I had when I was still in middle/high-school). My R9 290 was looking long in the tooth, Vega failed to impress and the RX 580 was barely an upgrade (more VRAM, lower power, and due to cryptomining horrendously overpriced) - I ended up getting an ASUS Strix 1080Ti.
All of us should hope that AMD at least catches up with Navi, the current dominance of NVidia in the GPU market (both for gaming and the datacenter) is bad for everyone - even if you're loyal to them for some reason no real competition doesn't give them much reason to keep innovating.
Just looking here and searching for NVIDIA: https://blogs.msdn.microsoft.com/directx/2018/03/19/announci...
yields: "And, our friends at EPIC, with collaboration from ILMxLAB and NVIDIA, have also put together a stunning technology demo with some characters you may recognize."
Which was rendered with DirectX Raytracing using RTX hardware. Link: https://www.youtube.com/watch?v=J3ue35ago3Y&feature=youtu.be
It sure as shit didn't backfire for Intel - they settled with AMD for $1,000,000,000.
For this price they completely hobbled AMD's ability to sell the Athlon 64, which hobbled profits (both directly and in terms of long-term branch image), which ultimately hobbled R&D.
AMD never fully recovered - and sold their foundries in order to survive. Intel, on the other hand, got full control over the pricing of pretty much the entire x86 market for over a decade - it was well worth what they settled for.
https://www.bloomberg.com/news/articles/2017-09-06/intel-get...
Basically they said please be nice, please pay us 1B.
But, as in the US or any other society ruled by law, antitrust fines and all other executive decisions can be challenged in court. That’s what Intel did, and the case is still pending.
I will no longer recommend or buy Nvidia hardware for anything and I hope reviewers will start to not review NV hardware over this.
It's a real dick move.
And this definitely got exposed with the recent serious vulnerabilities in all modern Intel x86 CPUs. In the attempt to get ever faster and make ever more sales, they had to lower quality.
Pretty much the same path with Nvidia in my opinion.
Perhaps in theory (emphasis on the perhaps, IANAL and neither is the author), but not in practice. Nvidia has a great deal of power over ATI, Asus etc. They have no power over Dell, much less Apple. Nvidia needs Dell and Apple far, far more than Dell or Apple needs Nvidia (they don't).
Nvidia and AMD are selling, if not identical products, then near enough for government work. If Dell decided to drop Nvidia from all their product lines, the impact on Dell would be negligible at worst, but it would be a catastrophe for Nvidia (while AMD would be making out like Scrooge McDuck). Apple? Nvidia sends Apple the corporate equivalent of drunken post-breakup apology voicemails (releasing Mac drivers for cards that aren't offered on Macs).
Nvidia's ability to enforce branding guidelines on Dell is nil. If Dell decided they'd rather brand "Nvidia" cards as "Dell GraphicsMaster 6000 SUX," Nvidia would have to roll with it.
This is an incredibly ignorant statement. While they both sell GPUs for consumer, and business (Quadro/Fire) AMD sells semi-custom CPUs, and has merged CPUs & GPUs into APUs. NVDA sells SOC, and AI focused hardware in addition to GPUs. The GPUs have huge differences between them, particularly with the newer APUs which are more, or less, customized for OEMs.
Summary: they overlap on GPUs, but their businesses are very different elsewhere.