TweakTown NewsRefine News by Category:
We know it's coming, but we don't know what exactly to expect. AMD is preparing its mammoth new GPU, which will be a huge 500sq.mm+ GPU, something that will arrive into our hands as the Radeon R9 390X.
AMD's new Radeon R9 390X is being designed by Synapse Design, a company that recently announced new GPU tapeouts - including two 28HPM silicons. The first is the massive 500sq.mm+, which will be as powerful as 20nm SoC technology, which should end up being the new R9 390X. The R9 390X should have a core clock of over 1GHz, but will unfortunately not feature High Bandwidth Memory, or HBM.
We should begin to hear more on the new GPUs as time goes by - but if these rumors are correct, the Radeon R9 390X is shaping up to be a beast.
It looks like NVIDIA isn't going to sit on its hands when it comes to the GTX TITAN, with the GPU maker reportedly working on a GM200-based GTX TITAN II. This new GPU will arrive as soon as early 2015.
According to German tech site 3DCenter.org, specifications of some new GPUs are also here with us, above. We can see the GM200 and GM204 GPUs will reportedly be built on the 28nm process. We heard a couple of days ago that NVIDIA would skip the 20nm process, moving onto the 16nm process, but nothing is in concrete right now.
The GM200 however, will feature over 4,000 CUDA cores, and should come with a much wider, and much anticipated 512-bit memory bus. If we expect it to be baked onto the 28nm process, this is going to be one gigantic, titan of a GPU. NVIDIA is going to have to leverage the energy efficiency of its Maxwell architecture in order to keep the thermals under control. We should expect the GM200-based GTX TITAN II to launch early next year.
Now that we've passed into the second half of 2014, we should expect some new GPUs in the coming months. But what is more exciting, is what is coming in the first half of 2015 according to a new report from SemiAccurate.
We should expect NVIDIA to skip the 20nm process completely, moving directly from 28nm into 16nm. This is just a rumor right now, but if it ends up happening, 2015 is going to be one of the most exciting years for GPUs in a very long time. This would see NVIDIA release the GeForce GTX 990 under the Maxwell architecture, from the GM204B core.
If NVIDIA does skip the 20nm process, we should see some seriously fast GPUs appearing, with much lower than expected power consumption. The core count should also be lower when compared to the GeForce GTX 700 series, as Maxwell is a much more efficient architecture compared to Kepler. So before we have the GTX 800 series, we're already foaming at the mouth about the GTX 900 series. It won't stop there, as we should see NVIDIA unveil the GeForce 1000 series sometime in the next 18-24 months, too.
Legendary overclocker Vince "K|NGP|N" Lucido, along with the help of Illya "Tin" Tsemenko, have smashed a GPU overclocking record. The overclocking duo used the EVGA GeForce GTX 780 Ti Classified K|NGP|N Edition video card, cranking its core to a record-breaking 2GHz.
The GK110-powered GPU was clocked to exactly 2025MHz, which is the highest clock rate reached by a graphics processor, ever. The EVGA GeForce GTX 780 Ti Classified K|NGP|N Edition video card was co-designed by K|NGP|N and Tin, featuring enhanced PLL circuitry, improved GPU power plane, 10CM fans, a special backplate, preinstalled dedicated PWM baseplate, voltage regulation enhancements and more.
Even at the default 1072MHz core clock speed on the GPU, it is faster than most of the GTX 780 Ti GPUs on the market, but at 2GHz, it really flies. K|NGP|N said: "In order to break world records these days you need some serious hardware. This card was engineered to serve one purpose... be the world's best overclocking video card. Expect GPU clockspeeds at over 1.85GHz with extreme cooling".
Chinese website GamerSky has posted up photos that they are calling a GeForce GTX 880 engineering sample, and boy do they paint quite the picture. The prototype that is in the photos below has some incredible specs behind it, with 8GB of RAM to start things off.
We should expect two variants to launch, with a GeForce GTX 880 with 4GB of RAM, and another with 8GB of RAM. From here, we have 3 PCIe power connectors, two 6-pin connectors and a single 8-pin connector for a total power draw of 375W. This is an insane number, but we are looking at a 28nm-based Maxwell GPU, and not the 20nm GPU that will pave the way for lower power draw and temperatures.
We should hopefully see the GeForce GTX 880 materialize before the end of the year, with a refreshed GTX 990 on the 20nm process expected for 2015 - this is me guessing here, but I think we'll see it happen.
According to a report, AMD is all set to use its mid-end 'Tonga' GPU cores on its existing cards, Radeon R9 280 and R9 280X GPUs. The upcoming core will be replacing the existing Tahiti Pro core that used on these cards coming August.
Once the existing two models use the newer core, they will no longer be re-branded versions of AMD Radeon HD 7950 and HD 7970 GPUs. The pricing of the newer cards is not known yet, however the R9 280 retails for $299.
It's still not confirmed if this new core will be based on 20nm fabrication process, though its assumed that it will more efficient than its predecessor. Its unclear if there would be a performance in comparison with the existing R9 280 and 280X cards. It is speculated that the newer cards will use have 2GB GDDR5, though it should be noted that the current variants are with 3GB GDDR5.
It looks like AMD is preparing itself a new GPU, the Hawaii XTX. Right now we have the Hawaii PRO, which is the Radeon R9 290, and the Hawaii XT, which arrives to us in the form of the high-end Radeon R9 290X.
The new Hawaii XTX will arrive in retail form as the Radeon R9 295X, and will feature the full 48 Compute Units - compared to the 40 on the R9 290 and 44 on the R9 290X. Better yet, there will be 3072 cores, this is an increase from the 2816 cores found on the already fast R9 290X. When it comes to TMUs, this will be increased from the 176 on the 290X to 192 on the 295X. The ROP count will stay at 64.
The question now sits with pricing - will AMD be competitive, because these specs will allow the company to compete directly with the GeForce GTX TITAN BLACK Edition, as well as the GTX 780 Ti. Because of the increased specs, we should expect the Radeon R9 295X to arrive with custom cooling from AMD's various partners in SAPPHIRE, MSI, and more.
It looks like we can expect NVIDIA to unleash next-gen GPUs this year, with the GeForce GTX 800 series to be revealed this year. Sweoverclockers is reporting that NVIDIA will launch both the GeForce GTX 880 and GTX 870 sometime in Q4 2014 - so between October and December.
While the GeForce GTX 880 and GTX 870 will be based off of NVIDIA's second generation Maxwell architecture, it will be on the 28nm process - not the 20nm process that we expected. TSMC has started volume production of the 20nm node for everyone, but this should start being spun on new GPUs in 2015 - something that will provide a large jump in performance, thermals, noise and much more for both NVIDIA and AMD.
Another interesting part of this news, is that NVIDIA has reportedly canceled the production of GM100 chips, in favor of the GM200 chips. These new GPUs are still based on the 28nm process, so it will be 2015 before things really ramp up for next-gen GPUs. This means that the new GeForce GTX 800 series are cards 'for now,' and intermediary product before NVIDIA can ramp up 20nm-based products next year.
We might see a few parts based on the 20nm process released under the GeForce GTX 800 series, but time will tell. These parts might arrive as something new, or old - such as NVIDIA bringing back the Ultra branding - which would be based on the 20nm process. This is my personal thought on the future, something I hope to see NVIDIA do. Release the intermediary product - GTX 880 and GTX 870 - and then unleash the GTX 880 Ultra based on the 20nm process before they drop the truly next-gen, 20nm-based GTX 990 (or whatever it arrives as).
It is pretty obvious that the new GeForce GTX 880 will be faster than the GeForce GTX 780, but what is interesting, is that it will be both cheaper, and more power efficient - something nobody can complain about.
The news is slowly floating out that the 20nm-based Maxwell-powered GeForce GTX 880 will be faster, cheaper, and more power efficient - but that is only one GPU in a massive lineup of cards. We should expect the new high-end Maxwell-based 20nm GPU late this year, or early 2015. NVIDIA could be forced to reveal 28nm-based Maxwell GPUs, but right now there's no threat to force their hand.
If NVIDIA do release the Maxwell-based GeForce GTX 880 on the cheap, faster than what we have with the GTX 780, or even the GTX 780 Ti, at around 200W power consumption, it would really put the pressure on AMD. This only ends up being great for consumers, though.
Computex 2014 - The Lightning Series from MSI has been one of our favorites for as long as we can remember. When we saw the GTX 780 Ti Lightning at the corner of our eye when walking around the MSI booth, we found ourselves instantly excited with just what was being shown off.
Carrying the massive triple fan Twin Frozr cooler and sporting Lightning features like GPU Reactor, Military Class 4 Components, Enhanced Power Design and Triple Force Architecture we knew that this was a card we must have. To then be told that the card wouldn't hit the market, though, was quite upsetting. Due to NVIDIA not allowing MSI to increase the voltage specification, MSI have decided to not release the model as it wouldn't give users the Lightning experience that they have come to known.
A small amount of the cards have been made which will be used by overclockers. With a modified BIOS they can increase the voltage to levels that continue to be safe while using LN2 cooling. It's really disappointing we'll never get a chance to test this model.