KitGuru KitGuru.net – Tech News | Hardware News | Hardware Reviews | IOS | Mobile | Gaming | Graphics Cards
Rumour has it that AMD might bump up the release of its ‘Vega' architecture graphics card. Just last month, AMD updated its GPU roadmap, showing Polaris launching this Summer and Vega coming in 2017 but following on from Nvidia's recent Pascal launch, some believe that AMD may now launch Vega in October this year. While this rumour has caught on with multiple sites reporting it, you probably shouldn't get too excited about this idea yet.
Vega is set to be the first of AMD's graphics cards to make use of HBM 2 and will apparently be a high-end card, while Polaris is targeting the ‘mainstream' GPU market this year.
Word that AMD might be planning to bring forward its Vega release date comes from a 3DCenter report, which is in-turn based on a comment from a forum poster. It is really shaky ground and highly speculative as there is no legitimate source.
I did some digging around and I managed to scrounge up two other instances of Vega being mentioned recently, the first was in a SemiAccurate comment thread, in which one user claims that the GPU is already being shown to some behind closed doors but doesn't specify who exactly. The second is in the patch notes for AIDA64, which was recently updated with preliminary GPU information for AMD Vega 10 (Greenland). Neither of these additional sources really do much to back up the rumour.
If Polaris is indeed targeting the ‘mainstream' GPU market, then it would make sense for AMD to bring out something this year to tackle the GTX 1080 in the enthusiast market. However, it is worth considering that there are other factors in play, like the availability of manufacturers and HBM2 chips that would impact AMD's plans to bring out a Vega graphics card this year.
KitGuru Says: If AMD could bring out a powerful Vega GPU this year then that would be great to see but this rumour sits on top of super shaky, practically non-existent ground, so don't get too excited about the idea until something more concrete comes along.
AMD seem to be putting their eggs in the HBM2 market and they’re probably right to do so if they want to compete or even overtake the GTX 1080, but isn’t the whole problem surrounding this that HBM2 isn’t exactly ready yet? (or the yields aren’t good enough, I can’t remember). How can they push the release date closer without making a huge loss overproducing HBM2 chips?
I need Vega before the release of Star Citizen PU. That may mean, next year. I want to know what is this card capable of in DX12/Vulkan.
HBM2 is ready, but it’s not in high quantities yet.
You know that every rumor says the specs will use HBM2, and nothing beyond that. GTX 1080 has proven to be overclock-able to 2.1Ghz which is more important to games than the type of memory used. It also uses 8GB GDDR5X, and HB SLI, which will bump up the SLI performance of these cards.
My point is, to decide to go with Vega purely based on the point that it will use HBM2 is near-sighted.
I like both AMD and NVIDIA. My current card is AMD. I’ll be looking for the real indicators of a card’s success: HBM2, 2.1Ghz OC speed, AMD has better CrossFire than NVIDIA, etc.
Don’t wait too long, these types of tech move forward at an alarming rate. Buy a VEGA card just to watch another card (AMD or NVIDIA) come and make that obsolete.
I’m totally with you on that. I’m aiming for an aftermarket 1080 and I’m dead set on that solely because if I wait for the Ti then the Volta will be just around the corner and that process will infinitely repeat.
To my understanding GDDR5X isn’t exactly too far off performance-wise of HBM2 So the news on the 1080 not using it, while a little annoying, isn’t exactly a game changer for me (and lets be honest, G5X Is still much better than what we’ve had previously).
With that being said there has been a lot of hype surrounding HBM2 and I guess I can see why that is but at the end of the day you are indeed right and memory is but one feature of many. We’ll just have to see how they both perform in the real world
Ah fair enough. So it’s a matter of potential low stock in the initial wave? I suppose AMD know what they’re doing in that case and can probably predict suitable quantities in time
Volta is planned for last quarter of 2018.
Oh is that so? A little further out than I expected then. Even so, the Ti probably won’t come out until half a year after the 1080 and that’s a long time, and then there’s the question of whether it will even be worth the wait vs aftermarkets and pricing. I guess I might hold out a little bit and see what information pops up. I never exactly intended to pre order a 1080 anyway, but it has been my intention for a long time to make this generation my next upgrade milestone
At 4k HBM 2.0 will come into its own.
Fury X, although restricted to 4GB is proof of that.
maybe a paper launch is likely here, I am itchy to see professional and laptop graphics cards with HBM, it really could be basically a die shrunk Fury built on 14nm silicon using HBM1, even a shrunk Fiji cut in half it would make for one great chip in efficiency focused applications. Come to think of it, for a lot of server type accelerators (not in scientific applications, I’m thinking for like streaming services), they are not in as dire need for compute as far as raw efficiency, scalability and bandwidth so a mini Fiji with a bit more memory would come in handy
Highly Volatiles Rumors
there is a substantial performance and power difference between GDDR5X and HBM2, and in regards to 1080 being so “Fast” if it was not able to make use of the substantial clock speeds, it would NOT be as fast as it appears to be, who knows how fast Polaris will be but take the 1070/1080 and run them at more “normal” clocks and they are not nearly as substantial a performance uplift from the 900 series.
Folks really need to stop drinking the Nvidia Kool-Aid and pay attention to how destructive to the industry they have been over the years let alone the massive cost per generation often with them reducing the quality of the components used to inflate the revenue they get to make which in turn boost their “apparent” market-share.
Anyways, until there is substantial 3rd party verification that is not an Nvidia paid promoter, I would take the information given with a massive bucket of salt, as often is the case, once they hit the open market then comes the proof of how well they are made and what if any BS tricks they pull to “alter” the given performance or actual power use, they often downplayed actual power use and overplayed quality of component selection and if there is ANY company that forces their ODM/OEM or whatever to follow VERY specific guidelines for bill of materials, clock speeds etc, it is Nvidia big time.
apparently HBM2 is first served to AMD for them helping to get HBM to market, so, I would believe their time to market(possibly even further refined for their specific needs then Pascal will be which is just using a bunch of the chips i.e sledgehammer instead of the scalpel which is what it truly needs to shine) also micron just announced mass availability of GDDR5x at higher speeds/lower voltage then initially anticipated, as well Samsung announced on track for 2018 7nm UV lithography(GlobalFoundries of course being a direct partner of this as well)
I personally think Zen/Polaris/AM4 is going to be a very streamlined level of class leading performance and a great reduction in power/temps more then folks are thinking, and seeing as AMD is essentially unifying the platform top to bottom for the first time in a VERY long time, as well Lisa Su and her team have been delivering exactly what she/they have promised(or better) since she took over, I have no doubt many people are going to be full jaw drop once it all “clicks in place”
GDDR5x is more than 50% slower than HBM 2 so there’s a big difference between the two. G5x – up to 448 GB/s – HBM 2 up to 1TB/s. It will be interesting to see if there will be much difference between Nvidia’s 16nm vs AMD’s 14nm Vega
I totally understand what you are saying, I’m in the same boat, if I wait for a professional version of Vega, I might as well wait for Navi, and the cycle repeats, I might as well go with a Polaris pro use card (preferably a full fat Polaris 11 low profile with 8 gigs of RAM) if they are offered at launch, if not just grab a full fat Polaris 10 with as much V-RAM as I can get (I do CAD work, video editing, small scale scientific work, mathematical comp and gaming on the side, I find AMD stuff easier to work with)
I would put GDDR5x closer to HBM1 in terms of performance, which still is impressive in itself
Yeah, that’s my guess too.
Yoy mean like how nvidea use average max power draw rather then the actual max power like amd
HBM2 is 8x faster per pin than DDR5X memory and 16x faster than DDR5. HBM2 also has 16 channels of interface compared to DDR5X’s 2. Think of it not only as faster memory (bandwidth throughput), but more efficient memory. Not only can you fit more onto a GPU because of it’s smaller architecture, but you also require less of it because of it’s efficiency.
How this will improve gaming performance is speculative at the moment, but I’m willing to wait for the 1080TI, or some other card to come out before I wait…Then again, I might buy a 1080 and switch to another card if HBM2 makes a bigger difference in gaming performance than I suspect it will.
Just a little correction, HBM2 is slower per pin than DDR5/X, but it is wider, having more data pins means the bandwidth is greater overall.
Oops!
The GP100 chip is a monster. It will for sure be worth waiting for if you already have a 970 or better. NVIDIA will save the 1080ti and Titan to counterpunch AMD’s Vega which I predict will also blow us away. Christmas time or early next year.
Star Citizen PU, while worth the wait, will probably be well after the release of Vega. The timing should be good for SQ42.
Yeah you are probably right, I am talking about Vega 11, just so you know so that is definately next year, perhaps in the holiday season – I expect it by then, HOWEVER, if Nvidia does decide to release a card with HBM2 next year and is powerful enough, I might as well get that…
SO glad you caught that, I didni’t want to be the one to have to waste my time typing to explain why in the actual F HBM2 is blatantly superior to 5X