Two days ago twitter user @CyberCatPunk shared a photo of what appeared to be spec sheets from SK Hynix about a new GPU that appeared to contain an impressive HBM2e memory configuration. The news was reported by several tech outlets, drawing the ire of Sk Hynix.
The rumor stated that the Big Navi GPU from AMD would contain a 24 GB HBM2e (opens in new tab)memory configuration with a 4096-bit memory interface, leading to a bandwidth of 2 TB/s.
Because we had our doubts about the legitimacy of this story, we decided not to run it. In hindsight, this was the correct choice: SK Hynix offered up a press release today stating that the news is fake.
"After a thorough investigation, we conclude that the screen capture of the allegedly internal document is fabricated." read the press release.
An earlier version of this press release, spotted by VideoCardz.com, even stated that "We can assure you that SK Hynix will take every necessary measure against such media, including all available legal actions."
The twitter account associated with @CyberCatPunk has since been deleted.
Situations like these highlight the necessity to ensure that information is gathered from trusted sources, and to remind ourselves that any information not directly from the official manufacturer should always, no matter the track record of the source, be taken with a pinch of salt.
For now, we do know that Big Navi is coming, as AMD has confirmed the GPU multiple times. However, we must not fool ourselves in knowing its specifications and/or performance levels. Many sources around the internet call the Big Navi an 'Nvidia Killer,' but without knowing its specifications or what Nvidia has in store for us with Ampere, such statements should ideally be avoided.
The specs seemed way high and for what they would be would only be for HPC uses. Otherwise it would be way too expensive for consumers with 24GB of the latest and greatest HBM memory.
You don't sue over rumors if they are totally bogus. It is something else when you screw up and leak information under NDA, then you sue.
Ahhh yes because one review thats meh means nothing reported is correct. I guess WCCFTech was right about the 5GHz 16 core Ryzen CPUs......
Rumors are rumors and have no place on a site like this. WCCFTech will throw multiple rumors out in a week about the same hardware. The only thing that should be on a legitimate tech site are actual specs and hard confirmed information.
You're right, but this site has been posting increasing amounts of leaked/rumor news as time has moved on. They long ago left the realm of only official and verifiable news posts.
And I comment such on those as well and say the same thing. If the article rates it as a rumor its normally OK though as rumors will come and go. There are other sites though that post it as just news.
Honestly to fight this the site should have a Rumors section dedicated to just that and not allow it to be mixed up in the normal news.
The reviews, real ones, are still mostly well done and legit and thats what I like TH for. But there has been a drop in names I know and trust to do them which is disappointing.
If there were no such product in the works, why would a "thorough investigation" even be required? It seems like it would simply be a matter of calling up the relevant people in the company and asking them if such a product is being worked on.
And they didn't actually deny that such a card is in the works. They denied that they created or distributed the specifications document in question, and claimed the HBM2e specifications listed therein to be inaccurate.
One possibility might be that the document is in fact fabricated, but that a lot of the specifications are real. There was already a prior leak about the alleged specifications of the card that this lined up with, and only the memory specifications were new, so that could very well be the case.
I agree about that being an excessive amount of HBM memory for consumer use, though the Radeon V also arguably had an excessive amount of HBM memory when it launched over a year ago. If this were in fact a top of the line card outperforming Nvidia's current $1000+ models, then there might be a fair amount of room to work with for pricing. Of course, unless there's some technical reason requiring additional HBM packages to avoid bandwidth limitations that would significantly limit performance, AMD would undoubtedly be better off limiting the VRAM to no more than 16GB, if not less, to give them more price flexibility and greater profit margins.
While the clock speeds might not have matched the rumors, if you take into account the 15% or so IPC uplift that had not yet been announced, the current Zen 2 based Ryzen 7 and 9 processors are capable of meeting or exceeding the performance that a Zen+ based processor would have offered at 5GHz. And we did in fact get a 16-core processor running at relatively high clocks on the mainstream AM4 platform. The rumored pricing that accompanied those specifications was quite unlikely though. Of course, Tom's Hardware reported on those rumors too, despite those pricing details not making much sense. : P
It's interesting that SK Hynix didn't actually deny existence of the graphics card (although there's no way a card with those specs would be gaming). They said that the HBM2e specs were wrong, the document itself was fabricated and factually incorrect (wrong HBM specs), and said that is was illegal to reuse content from their Newsroom without permission.
I don't know how SK Hynix normally denies rumors, but this if reads like the kind of careful wording that Nintendo has previously used to "deny" the existence of new products. Nintendo words things in a way that expects you to infer that something doesn't exist, and then they announce it a few weeks later.
Or it could read like the wording was careful, if SK had a better translator. I don't think this denial was done with that kind of care.
For the mass majority 16GB would be overkill. For the top of the crop people it would be usable but it wouldn't be a hot seller and even 16GB would put costs on a new memory standard way up there. It would be hard for AMD to price it competitively with a GPU that uses GDDR6 which is much cheaper than HBM.
And the rumors were proclaiming a 5GHz chip, nothing to do with IPC uplift. WCCFTech is a major rumor mill.
And yes TH did report on them. I think its stupid to do so unless they have a section dedicated to just rumors. I think rumors do not belong on this site but this site has changed a lot and the ones running it are not quite the same enthusiasts that started it long ago.
Manipulation of stock prices through false information is a serious crime. We aren't talking about idle speculations here. The guy went through the trouble of fabricating internal documents.