KitGuru KitGuru.net – Tech News | Hardware News | Hardware Reviews | IOS | Mobile | Gaming | Graphics Cards
Advanced Micro Devices uses yet unannounced Radeon R9 graphics card for a virtual reality demonstration at the Game Developers Conference, the company confirmed on Thursday. The GPU developer reveals no details about its new flagship graphics card, but those, who have seen it in action, notice impressive performance.
Although AMD decided to postpone the launch of its new flagship Radeon R9 graphics card to Computex Taipei 2015 trade-show because of business reasons, according to unofficial sources, the product seems to be ready and AMD and its partners already use it for demonstrations. During a press event at the GDC trade-show AMD ran a virtual reality demonstration called “Showdown” that used Oculus Rift “Crescent Bay” VR headset and was rendered using a PC powered by “unannounced flagship AMD Radeon R9 graphics card,” a spokesman for the company revealed.
The Showdown demonstration powered by Unreal Engine 4 and developed by Epic Games is designed to show capabilities of virtual reality technologies and hardware. According to a report from Tom’s Hardware, “the performance of this demonstration was remarkably responsive” on the system powered by AMD’s next-generation Radeon R9 graphics adapter, latencies were low and graphics quality was very high. It is unknown whether AMD used its LiquidVR technologies to improve experience, but it is a possible scenario.
High performance and lack of any problems with this particular demo indicate that drivers for AMD’s new Radeon R9 graphics cards are in good condition. This points to the fact that AMD’s next-generation Radeon R9 is generally ready to ship and its software is fine for a commercial product.
Previously it was reported that AMD’s next-generation top-of-the-range Radeon R9 features GCN 1.2 or even better architecture, 4096 stream processors and 256 texture units. The memory sub-system of the new Radeon flagship relies on SK Hynix’s stacked high bandwidth memory (HBM) chips that provide up to 640GB/s of bandwidth (thanks to 4096-bit memory bus and 1.25GT/s transfer rate). The graphics boards will be equipped with 4GB of HBM memory.
Some estimate that the Radeon R9 380X will provide over 50 per cent performance improvement compared to the Radeon R9 290X and will demonstrate particularly high results in ultra-high-definition resolutions, such as 4K (3840*2160, 4096*2160) or 5K (5120*2880).
Discuss on our Facebook page, HERE.
KitGuru Says: It is pretty obvious that AMD’s next-generation Radeon R9 is here and its launch is inevitable. The only reason why AMD postpones the official introduction of its all new graphics card with innovative type of memory is the company’s intention to sell through existing Radeon R9 280-/290-series inventory. But who will buy an outdated graphics board knowing that something brand new is just around the corner?
If its ready and running then at least release the specs for the people waiting for some kind of word on how long they are going to have to wait. At this point they still haven’t said if they are going to release it fully at Computex or if they are just going to release the specs on the new line. If they don’t do a full release at Computex then what? We’re going to have to wait another month or more to see it on shelves? I was looking forward to hearing something at GDC on a timeframe not telling their customers that they were going to have to wait even longer.
well i looked at the Computex calendar and the closest event they have is on june 1st…that is almost 3 more months just to announce the god damn thing! who knows when it’s gonna be available,or the reference coolers are good enough and availability won’t be crap cuz of greedy miners…i really hope with all these delays they put up a fight!
I have to agree, even with all the 970 events nVidia are gonna run away with it with their 900 series if AMD take too long (unless the R9 300s are SUPER EPIC). And now Titan X has been announced as well. I like competition, it brings prices down and forces innovation. I need a second card for SLI at half price plz 😀
I’m willing to bet that last article about the delay was complete BS. Either that or the management at AMD is made up of idiots. Who demos a full-feature product an entire quarter before it’s released?
Actually this is common practice in the gpu business. the 290x was showed almost a full year ahead of release at a convention. It sucks but this news isn’t speculation , it’s fact. It has came from several credible AMD sources (distributers) in the last few days.
Stop posting.
They will have production samples they can demo before the product has been in mass production long enough to have a retail release. No point releasing it and then having it sold out for the next 6 months.
If nvidia’s titan launches before AMDs new card, it allows AMD to directly compare their card to NVIDIAs straight out of the gate, because these rumored specs have AMDs card trumping nvidia’s new titan.
I am Confused,
Fiji XT is 380X or 390X?
Oh man June is a long ways away…
We know Lisa said back end of January, they intend to clear the channel of inventory, but such Fiji products are so much higher in the stack I wouldn’t see that directly impacting Hawaii or say the 380X (Grenada). So I would think two things; AMD is readying a big “war-chest” of Fiji chips to understand how best they can deactivate what (if the intention is to have 3 variants, along with dual card) to figure out the most effective mix. The other could be receiving enough HBM to really able to saturate day one of release. Finally they’re not in any different position than they’ve been since the 980/970 came to market last October. They can wait… get closer to what Nvidia may offer with a GM200 to best position it (Fiji) in the market.
Now the rest of the line-up I can’t tell you what AMD doing there, it’s not like they are new designs… they’re rebrands. I don’t think AMD is waiting to see what a 960Ti turns out like, that’s not going to be an earth-shattering affair. Heck I don’t see Nvidia going there until after AMD takes their turn. Other than still clearing the channel of inventory (without dumping the price), which is what AMD looks to have been trying, there’s not much of a purpose in holding off releasing everything below Fiji.
I hope they start with the Tonga (Pro $170/XT $200) and knock the GTX960 down a rung or three. Then Grenada XT for $280, and have the 970 feel some heat (pun-intended). Trinidad twins at end of April-May to really grab hearts of all the summers young gamers looking to upgrade with little funds ($100-140). Then finally Fiji as the “Icing on the Cake” is announced 1st of June at Computex; the Pro at $340-350; XT for $500.
Sure AMD can’t best on efficiency, or perhaps maintain some illustrious crown… this round. They just need to overwhelm Nvidia at every price point for the next year and half, and not bother about Titan X. Stay in the trenches and vie on perf/$.
Keep Calm & Carry On…
Depends on how much higher the IPC of their new architecture is. If it’s 5-10% faster than Tonga then it could easily outperform the new Titan. If not though, it could struggle to compete and their decision to wait to release it could backfire badly.
I truly believe this is misimformation purposely spread from amd to force nvidia’s hand. All rumors pointed to an early april launch, hoping nvidia would show their hand if they believed amd would be releasing fiji xt in april. Now that nvidia have announced titan x they will wait for performance and pricing before making their own move. I still believe the 390x will launch in april if titan x is launched this month. They can get the 390x in the market first allowing it to build hype for the cut down versions to come in may and june.
I’m afraid the 390x will cost more than previous amd flagships. If it contains 4096 shaders it will undoubtedly be on a very large die, probably the largest amd have ever produced. They also referred to it as an ulteultra enthusiast solution when confirmimg its existence to pcper, I wouldn’t be surprised to see a 699 or higher price tag.
380X is Fiji XT
390X is Bermuda XT
395X2? Theres no information out about such a chip at all.
–
Some people are saying Bermuda is the dual chip but that dose not match up with the benchmarks. The Bermuda benchmark lists it as slower then the 295X2. Right where you’d expect a single 4096SP chip to be… dual 4096SP chip should be much faster then 295X2.
Titan X is a confirmed flop no market for the card.
Maybe not if there doing it at Glifo. Glifo 28nm is more optimized for GPU & APU instead of CPU like TSMCs is it should come with some size benefits.
The 290X hit the market at $550, but now as low as $300. So, that was 438 mm² part with 2816sp; add 45% = ~4083sp, while 438 mm² + 45% = 635 mm². I can’t see it being that size, and I not sure if it end up a 4096sp part more how they count a shader, just as Nvidia changed what they called Cuda Core counts.
The GTX110 was a 561 mm², but said to use the more costly HP process and they sold off cards as low $300 (salvaged 780 parts). First we don’t absolutely know what “Fiji” will be and yea $500 is very light on price. but $700 from AMD IDK. I think they won’t price to via TitanX, but set a price that is meant to make a 980Ti come about.
If as the guest below indicates “if” GloFo parts come about then price is totally unknown. It be nice to have AMD at GloFo this go around, but I yet to believe that’s going to come about this round.
Using GloFo for this chip is how they will achieve 4096 shaders, their 28nm process is rumored to allow for far denser transistor design than tsmc at the same size.
Honestly, if AMD has been using this time to flip their designs to GloFo… Oh man would I be happy! If they return to a channel that’s cleaned out, and product on new silicon from GloFlo, it would be hard to curb the enthusiasm.
Now can’t say the rumor mill backs this thinking up, but if that transpires I hope it works well for AMD (risk=reward). Even if only moderately improved, if they have Grenada, Fiji and Trinidad from GloFo, it could make things interesting. Something both being subservient to TMSC has not afforded the market.
You are making several omissions in your post.
1) R9 290X increased shaders from 2048 of Hawaii to 2816 but the die size went up from 354mm2 to 438mm2 (+24% increase in die size for 37.5% more shaders). In other words, the number of shaders does not need to scale linearly with die size as you have assumed.
2) R9 290X had a 20% smaller in die size memory controller despite it being a 512-bit one over Tahiti’s 384-bit version. Since HBM is a completely different memory technology, we can’t rule out that 390X’s memory controller won’t be even smaller than R9 290X’s which provides even more space.
I think a more reasonable die size will be 540-560mm2, not 635mm2 as you have described.
Re-read the post I said… I *can’t* see it being that size (635 mm²) … I was just giving some correlation to Trent Foley post, while making the further point “that it’s more how they count a shader”.
You meant Tahiti to Hawaii it went 2048 / 2816, while yes they packed more shaders on a proportionally smaller die, and yes it was memory controller and other things that were cut back offer the savings. I don’t see them packing the die even more again, as the density of components was a direct result of Hawaii being hot and that heat was lost in inefficiency. AMD took for their word TSCM and their specified engineering parameters for their 28nm process and it bit them.
There’s no secret sauce… current GCN shaders use a fix amount of real estate and why i indicated if they intend increase by ~45% the number (based on what a GCN shader is today) the die will grow, and well beyond any logical size. It was why I said the GCN shader as we know it today is almost undoubtedly be revised.
Misinformation? We are talking about a marketing division, not a tin foil hat conspiracy…