At the rate things are going, Pascal may end up the most hyped graphics card release in recent memory. So far we’ve been treated to die shots of GP106, the low-end chip used in the Drive PX 2. The GP104, set to release a little over a month from now, has remained hidden from view, till today.
Coming out from Chinese forums, we can finally see what GP104 looks like. The SKU we have before us is the GP104-200, the one that is expected to bed used for the GTX 1070, the middle of the pack for GP104. The die size comes in at about 333 mm2 which is remarkably similar to the GM204. Around the die, we are able to see some Samsung 8Gbps 8Gb modules. This is what we have expected for the GTX 1070 and suitable for the GTX 1080 if it doesn’t use GDDRRX.
Overall, the leak today confirms what we already know about Nvidia’s plans. They will start off with the small GXx04 die as the “flagship” and follow it up with the bigger die later on. If the die size is correct, we maybe seeing a decently larger chip than the GM204 due to the die shrink. However, Pascal may bring back some of the compute cut out in Maxwell so it remains to be seen how the chip will truly perform.
When AMD and Nvidia release all of those TFLOPs numbers, it’s important to realize that those are theoretical maximums. In order of a chip to reach that number, its architecture has to be extremely efficient and powerful to use. When Pascal first launched, Nvidia released some details about what it would look like. With the release of the white paper for the architecture, there are a few additional highlights worth noting.
First off, we know that Pascal has cut the SM (Stream Multiprocessor) down from 128 FP32 cores to 64. This allows for better distribution of processing power to tasks and as each SM keeps the same amount of register files and other support hardware, throughput is increased overall. Nvidia has also tweaked the SM so datapaths are more streamlined and sharing information within the SM takes less power and hardware. The scheduler has also seen some improvements and updates to ensure the SM is constantly being fed.Cache sizes have also been increased from 3MB to 4MB and a dedicated shared memory space of 64KB per SM has been added This is lower than the 96KB per SM in Maxwell but if you consider the doubled SM count relative to a same size Maxwell chip, it’s actually an increase of 16KB per SM. Finally, Nvidia detailed the P100 interposer layout for HBM2, something for us to look forward to when HBM2 finally arrives.
For other details, be sure to check out our earlier write up on the Pascal architecture.
With Computex just over a month away, leaks have been popping up daily about Nvidia’s upcoming Pascal graphics cards. So far, we’ve seen a lot of leak around the physical card but nothing yet on core specifications or pricing. Today, according to Taiwanese insiders, we are getting a look at how much GP104 will set buyers back. The GTX 1070 will reportedly sell for NT $ 19990 and the GTX1080 will sell for NT $ 27990.
Converting the prices, you get around $620 and $870 for the 1070 and 1080 respectively. That seems a bit high considering that you can find the 980 at about $500 and the 980Ti at $700. However, if you only consider NT prices, that’s exactly in line with what the GTX 980 and 980Ti cost in Taiwan respectively. This suggests to me that the GP104 cards would be priced where the GTX 980 and 980Ti are right now based on the market they are selling in
These prices are in line with what we’ve come to expect from Nvidia and mirror the GTX 670/680 and GTX 970/980 launches. The GTX 1070 will significantly less and offer competitive performance and the GTX 1080 will command the extra flagship premium. While high the prices are understandable given the new process and architecture. It’s just a shame that the top end GP100 won’t be the true flagship anymore at launch.
Both AMD and Nvidia have tended to launch higher end flagships before moving to fill the rest of the lineup. The same is expected for Nvidia’s Pascal launch, with the “flagship” GP104 based cards set to arrive before the midrange GP106 does. As we know GP104 will arrive with the GTX 1080 and 1070 at Computex and reports are surfacing that we won’t see GP106 till the autumn of this year.
By holding back the release, Nvidia balances out their sales and gives a chance for last gen cards to get be cleared out of stock first. This is especially true of the low-end as GM206, the current Maxwell midrange chip, is still in production. Many of Nvidia’s AiB partners also recently launched new 75W GTX 950 cards, raising suggestions that Nvidia is preparing cards to counter the more power efficiency Polaris 11 which should arrive before GP106.
With GP104 reaching down to the 1060Ti, the GP106 will likely power the GTX 1060, 1050Ti and maybe even the GTX 1050 as well. According to the source, the chip will also not require a PCIe power connector, with a TDP of at least under 75W. This would put is TDP near that of Polaris 11 which is also under 75W. With GP106 already spotted on the Drive PX2, it’s only a matter of time though till Nvidia launches the cards,
NVIDIA’s upcoming architecture codenamed Pascal is rumoured to launch in June and apparently invites have already been sent out to members of the press for an early briefing on its unique features. The new range is built on the 16nm manufacturing process and could utilize the latest iteration of high bandwidth memory. Although, the mid tier options might opt for GDDR5X to cut costs and maintain a lower price point. Of course, there’s always leaks whenever a new architecture is being readied from either graphics company. However, NVIDIA has kept the information extremely secret and there’s not much to go on when trying to work out the specification across various products. Some time ago, a leaked image supposedly showcased the new cooler shroud and GTX 1000 branding. This looked fairly credible given the environmental setting and high quality of workmanship.
Today, a brand new image has arrived courtesy of Chinese website Baidu which apparently shows the GTX 1080 in full for the first time:
The cooler opts for a dynamic appearance with sharp edges compared to the older reference design. It also corresponds with the previous leak which suggests both images are credible. On the other hand, it’s important to adopt a cynical approach just in case someone made the effort of modding the stock cooler to foil people around the globe. Honestly, this is very unlikely and it’s probably the final design for the GTX 1080. Sadly, it’s impossible to tell if the GPU contains a stylish backplate. Although, this should be revealed soon if the release date reports are correct. Whatever the case, it looks like NVIDIA has tweaked the cooler design and opted for the GTX 1000 series name. I’m not so sure this is a clever move as the 1080 might cause some confusion.
Do you think the image discussed in the article is genuine?
Winding down production of a soon to be discontinued product is industry standard practice. When you see production of current models cease, that’s when you know the next generation is just around to corner. Two weeks ago, we brought you news that Nvidia may have stopped GTX 980Ti production. Now it seems that Nvidia has also ceased production of 2 other Maxwell-based GPUs in preparation of Pascal.
According to HWBattle, both of the GM204 based GPUs are no longer being supplied to AiB partners. This means the GTX 970 and 980 will disappear off store shelves sometime between the next 2 months given a normal logistics situation. At the same time, this means the replacements for GM200 and GM204 are well on their way and may arrive within 2 months as well. This is perfect for the late May launch at Computex.
As we’ve reported before, the replacements will be the 3 GP104 chips that will be called GTX 1080, 1070 and 1060Ti. These are the GP104-400, GP104-200 and GP104-150 respectively. Furthermore, the usual practice of launching reference cards first followed by custom ones will not be followed by the GTX 1070. Instead, the custom cards will launch at the same time as the reference models. With Computex just around the corner, we’ll all find out soon what GeForce Pascal truly is like.
For those of you hoping for massive performance jump with the launch of Pascal, prepare to be disappointed. Every new generation tends to improve performance but some more than others. According to previous rumours, Nvidia is using their GP104 die to replace the GTX 980Ti with the GTX 1080 and 1070. Now, the latest reports are suggesting that Nvidia will launch 3 different Pascal SKUs, all based off of GP104, at Computex.
As the xx4 die, GP104 has traditionally been viewed as the smaller chip to the larger x10 or x00 dies that traditionally power flagships. Due to this, don’t expect Pascal to surpass the 980Ti by any large amount. Today’s news also furthers that impression. By splitting GP104 into 3 SKUs, we can expect performance between the 3 cards to be pretty close. It wouldn’t make sense to have so many close performing cards to the flagship which suggests that GP104 won’t be real flagship material.
By slipping GP104 into 3 SKUs, we will likely run into the same situation as the GTX 560Ti 448/570/580 and the 660Ti/670/680. If we take our past experience with those cards as the guideline, we can expect differentiation, not just on the core but the memory bandwidth as well. This makes the previous rumours about the GTX 1070 using GDDR5 while the GTX 1080 will use faster GDDR5X. The 1060Ti as I am calling it may feature either a gimped 192bit bus or the same situation faced by the GTX 970 with a section of VRAM being slower.
Right now, all we have to differentiate the 3 SKUs are the model numbers, the GTX 1080 will be GP104-400-A1, the GTX 1070 GP104-200-A1, and the 1060Ti will be using the GP104-150-A1. It will be interesting to see how Nvidia will differentiate the cards and how they compete against current Maxwell models. Computex can’t come soon enough!
If there is any doubt in our minds that real Pascal cards are coming soon, this latest report clears that all out. According to PCTuning, Nvidia has started sending out invites to select press outlets for a Pascal event. This suggests that Nvidia already has all of their Pascal chips that are launching at Computex ready for the press to review and test out. As of right now, we still don’t know the exact date of Pascal’s launch.
Right now, whether or not the event will be a private event for press only or a live webcast sort of affair remains to be seen. This may well be the event Nvidia sometimes holds with the press beforehand in order to brief them on the chip a bit before shipping out review samples.
What’s more, the leaker is suggesting that Nvidia is going to allow reviews and other information to be released in as soon as three weeks and four weeks at the latest. This would place it at least 2 weeks ahead of Computex which has been the suggested launch. The possibilities are that the leaker is wrong, Computex will be the hard launch with a soft launch before; either way, Pascal is coming soon.
NVIDIA showed of its DRIVE PX 2 system – the new iteration of its autonomous and driver assistance AI platform – at last week’s GTC 2016 conference, and eagle-eyed viewers may have noticed that the board shown to the audience by CEO Jen-Hsun Huang was sporting a pair of integrated GP106 GPUs, eschewing the two Maxwell-based NVIDIA Tegra X1 chips that powered the original DRIVE PX, and confirming a rumour that we reported last week.
The GP106 runs on NVIDIA’s new Pascal architecture – set to hit the market in the latest line of GeForce graphics cards this Summer – which can perform at 24 DL TOPS or 8 TFLOPS, and features up to 4GB GDDR5.
NVIDIA hopes that the new DRIVE PX 2 will power the next generation of driverless cars – the DRIVE PX has so far only be used to power the ‘infotainment’ system on-board a Tesla, for example – and has already shipped to a number of unnamed Tier 1 customers.
“DRIVE PX platforms are built around deep learning and include a powerful framework (Caffe) to run DNN models designed and trained on NVIDIA DIGITS,” according to NVIDIA. “DRIVE PX also includes an advanced computer vision (CV) library and primitives. Together, these technologies deliver an impressive combination of detection and tracking.”
Some of the first cards to run utilizing the all new “Pascal” architecture made by Nvidia, may debut at Computex 2016. The show is going to be in Late may / Early June in Taipei and is one of the biggest ICT shows in the world and you can be sure the eTeknix team will be there to catch the latest news from the event!
Mass shipments should start sometime in July according to Digitimes, the Taiwan based industry observer. With Nvidia unveiling the new cards via its add-in card (AIC) partners, with large manufacturers such as ASUS, MSI, and GIGABYTE being at the event.
The new GPU will be based on the GP104 chip and utilize GDDR5X VRAM; a whopping 8GB is rumored to be the amount. The leaked specs show it having a single eight-pin power connector, meaning that (due to electrical capacity) the max power usage would be 225W, though it could use a lot less power. The 980 is only 165W so this card can’t be a huge amount more. The leaked specs also tell us that it could feature up to 6144 CUDA cores and a whopping 12.6 Teraflops. We’re not sure how accurate these specs are as they have been sourced from various places, only time will tell. Either way, Computex 2016 is going to be huge this year.
So far, we can accurately say:
2x performance per watt estimated improvement over Maxwell
DirectX 12_1 or higher
Successor to the GM200 GPU in the 980TI
Built on 16nm manufacturing process
It will be interesting to see the Polaris release too, as there is going to be some very tough competition on the GPU market shortly and that’s obviously great news for consumers.
Which cards are you most excited about this year, AMD’s or Nvidia’s latest? Let us know in the comments section below.
With rumours pointing to a May/June launch at Computex, more information is coming about Nvidia’s upcoming GTX 1070 and 1080. At times taking on an x70 and x80 moniker, the two chips are slated to use the GP104 Pascal die and take up the role currently filled by the GTX 970 and 980. Today, another leak has come out detailing what the two cards will look like and it seems the GTX 1080 will have a lot more memory bandwidth than the GTX 1070.
According to the leak, both the GTX 1080 and GTX 1070 will be based on the GP104 die. This will slot into the Pascal lineup just like the GK104 and GM204. While previous cards have mostly differentiated in the core specifications, it looks like this time, memory bandwidth and a lot of it will be the difference. The GTX 1080 will reportedly use GDDR5X while the 1070 will use GDDR5. This should give potentially up to 100% more bandwidth for the GTX 1080 and better energy efficiency to boot. Due to the use of different memory, the 1080 will boast 20 more pins. The 1080 will use the GP104-400-A1 and the 1070 the GP104-200-A1.
For two similar cards based of off the same GP104, there seems little reason to split the memory between two different types since the required memory bandwidth should be similar. One possibility is that the GTX 1080 may be using relatively slower GDDR5X that isn’t much faster than the fastest GDDR5. This makes sense if GDDR5X is supply constrained for the top end models. Another possibility is that both were meant to get GDDR5X but supply meant only one of them could. Finally, it could also be a way to differentiate the GTX 1080 as it may remain the GeForce flagship for quite a while.
Last of all, we have also gotten information about the connectors, Both cards will feature 3x Displayport, 1 HDMI, and 1 DVI for display connectivity. Power will be provided by 2 x 8pin PCIe power connectors which is actually more than what the Titan X has. Given the efficiency from moving to 16nm, this points to either a monster chip or lots of dual precision hardware being left in. With only several weeks left, will be interesting to finally see what Nvidia has cooked up for us.
All throughout this week, we’ve been treated to leak after leak and release after release of new Polaris and Pascal information. After we’ve gotten die shots of GP100 and GP104, what may be another Pascal chip has surfaced. According to Videocardz, the Drive PX 2 module Jen-Hsun Huang waved in front of our faces might actually be Pascal this time around. More specifically, it might be GP106 or GP107.
Nvidia took some heat earlier in the year when CEO Jen-Hsun Huang took to the stage and waved a Drive PX 2 around that was Maxwell. This time around though, it looks like Pascal has come far enough along to make it into engineering samples. Based off of size comparisons to GM206, the Drive PX 2 GPU is too small in comparison. This means it is either the budget GP106 or the smaller mobile GP107.
This stems from the fact that the die is really similar to GM206 and given what we know of GP104 and GP100, the GP106 and GP107 should be of a similar size to GM207 and GM206. I believe that given the MXM format, we may be seeing the GP107 as that chip is meant for the mobile market. Using a chip that hits both the lucrative laptop and car markets would make a lot of sense. Either way, budget Pascal may be coming sooner than expected.
So far all of the rumours around the GP104 and GTX 1000 series have mostly been about release date and specifications. The closest we’ve gotten to physical evidence have been the shrouds for the GTX 1080 and GTX 1070. For the first time, we’re getting a picture of the physical die and parts of the GPU board around it. According to ChipHell, the die shot you see below belongs to the GP104, the mainstream Pascal GPU.
From the die shot picture, GP104 appears to be about 15.35mm x 19.18mm for a total of about 290 to 300mm². This is the same as GK104 which was also a die shrink and came in at 294mm² and much smaller than GM204 which was a relatively massive 398mm². This shows that Nvidia is starting out with small dies first with the GTX 1070 and 1080 and releasing a GP100/102 Titan and 980Ti later on.
For now, we still don’t know what GP104 will look like, but it seems that most of the FP64 units in GP100 will likely be stripped out and replaced by the more ‘useful’ FP32 ones. The leak also suggests that total FP32 CUDA core count will be around the same as the Titan X but the TMU and ROP count seems closer to the GTX 980. I expect that clock for clock, GP104 won’t be much faster than the Titan X but it will be ahead and much more efficient.
Finally, we can see what appears to be Samsung 2Ghz 1GB GDDR5 DRAM modules for 8GB total. This suggests that that GDDR5X isn’t ready in time or will be reserved for the GP100/102 consumer release. This follows the same trend set by the GTX 680 which was more powerful than the 580 but featured lower memory bus width but faster VRAM and more memory overall.
While the leak is promising, it is a leak after all and I would make sure to take all of this with a shipload of salt. Given the information we know though, this leak may very well reflect reality.
From the many leaks and rumours that have come out, the expected release of Pascal will come later this year at Computex. During the Taiwanese event, Nvidia will finally unveil the GTX 1000 lineup to the public. Today, we’re getting yet another report confirming this. In addition, Nvidia AiB partners like ASUS, Gigabyte and MSI will showcase their reference cards then as well. As revealed yesterday, mass shipments won’t begin till July.
As we’ve covered earlier, the sources appear to suggest that Nvidia will have a head start over AMD in launching Pascal ahead of Polaris. However, the lead might not amount to much. The report suggests that Nvidia will only ship large amounts till July. AMD on the other hand is also launching Polaris in June, the same month as Pascal. Given AMD’s previous history, we will probably see Polaris cards out by July as well. If Nvidia does have a lead, it won’t be for very long.
Right now, there is no word yet if GDDR5X will be utilized for top end Pascal chips. While there are some reports out that suggest GDDR5X, the timeline is very tight as GDDR5X probably won’t reach enough capacity till May/June at the earliest. Perhaps this is why we won’t be seeing Pascal or Polaris in numbers till July.
Yesterday, we reported on AMD’s plans to supposedly launch their next graphics architecture, codenamed ‘Polaris’ in June. The details surrounding NVIDIA’s response with Pascal aimed at consumers was unknown but it seemed likely the range would arrive at a similar date. According to new information sourced by Digitimes, Pascal will be unveiled during Computex and enter mass shipments in July:
“The graphics card players will begin mass shipping their Pascal graphics cards in July and they expect the new-generation graphics card to increase their shipments and profits in the third quarter.”
Interestingly, the report claims that AMD might only unveil the Polaris range in June, and shipments could occur at a later date. Apparently, NVIDIA will have the advantage and be the first company to release their new line-up to the public:
“Meanwhile, AMD has prepared Polaris-based GPUs to compete against Nvidia’s Pascal; however, the GPUs will be released later than Nvidia’s Pascal and therefore the graphics card players’ third-quarter performance will mainly be driven by demand for their Nvidia products.”
Thankfully, both graphics card manufacturers look set to release brand new products and I cannot wait to see the performance improvements and pricing across various tiers. AMD appears to be focussing on performance per watt on Polaris and the demonstrations thus far have been impressive. Sadly, we haven’t really seen anything from NVIDIA’s new consumer line-up, so it will be fascinating when the samples finally arrive. It’s still unknown which products will opt for HBM2, if any. It’s clear that the majority of models from both companies are set to utilize GDDR5X. While this isn’t a patch on HBM2, it’s a good improvement from the older GDDR5 standard.
Recently, there were some murmurings about NVIDIA delaying mainstream Pascal under 2017. This doesn’t look like the case at all, and if anything, reports suggest they will be the first to market.
With the reveal of the Tesla P100, Nvidia has taken the wraps off of their new Pascal architecture. Originally set to debut last year, delays with 16nm kept Pascal from being a reality, leading to Maxwell on 28nm. Now that Pascal is finally here, we are getting an architecture that combines the gaming abilities of Maxwell with much improved compute performance. The new Unified Memory and Compute Pre-Emption are the main highlights.
First off, Pascal changes the SM (Stream Multiprocessor) configuration yet again. Kepler featured 192 CUDA cores per SM, Maxwell had 128 and Pascal will now have 64. By reducing the number of CUDA cores per SM, it increases the fine grain control over compute tasks and ensure higher efficiency. Interestingly, 64 is also the same amount of cores GCN has in each CU, AMD’s equivalent to SM. The TMU to CUDA core ratio remains the same as Maxwell with 4 per SM instead of 8, in line with the drop in cores/SM.
For compute, the gains mostly come from increasing the number of FP64 or Dual Precision CUDA cores. DP is important for scientific and compute workloads though game rarely make use of them. Kepler started cutting out some FP64 units and Maxwell went even further, with virtually no FP64 even in the Tesla’s. This was one reason why Maxwell cards were so efficient and Nvidia only managed to hold onto their leadership in compute due to CUDA and their Single Precision performance.
With Pascal, the ratio of SP to DP units goes to 2:1, significantly higher than the 32:1 of Maxwell and 3:1 of Kepler. GP100 in particular has about 50% of its die space dedicated to FP32, about 25% to DP and the last 25% split between LD/ST and SFUs. This suggests that Pascal won’t be changing much in terms of gaming performance. The only gains will be from a slight increase in efficiency due to the smaller SMs and the die shrinking from 16nmFF+. GeForce variants of Pascal may have their FP64 units trimmed to cram in more FP32 resources but again, most of the gains will be due to increased density.
Lastly, Pascal brings forward unified memory to allow threads to better share information. This comes along with improved L2 cache sizes and the more than double register file sizes. P100, the first Pascal chip, also uses HBM2, with 16GB of VRAM over a 4096bit bus for a peak bandwidth of 720 GB/s. For CUDA compute tasks, a new Unified Memory model allows Pascal GPUs to utilize the entire system memory pool with global coherency. This is one way to tackle AMD’s advancement with HSA and GCN and Intel’s Xeon Phi’s.
Overall, Pascal looks to be an evolutionary update for Nvidia. Perhaps, Nvidia has reached the point that Intel has, making incremental progress. In other ways though, the reduction in SM size has great potential and provides a more flexible framework to build GPUs. Now all we are waiting for is for the chips to finally drop.
After revealing their next flagship Telsa earlier, Nvidia has let loose with a few more details and specifications. Based on the new Pascal architecture, the P100 will be utilizing TSMC’s latest 16nmFF+ process. As we know from the keynote, the chip will feature 15.3 billion transistors and the latest HBM2 memory. The P100 also features what Nvidia is calling the “5 miracles”.
First off, the P100 will run at an impressive 1328 MHz base clock and 1480 MHz boost. This is high for a professional Tesla card though well in line with GeForce clocks. The card won’t be using the full GP100 die with 60 SMs and 3840 CUDA cores, rather it will use a cut-down version with 56 SMs with 3584 cores. This mirrors Kepler’s launch where the cut-down Titan came before the Titan Black. In addition to the usual FP32 CUDA cores, there are also 1792 FP64 CUDA cores for Dual Precision Work. This gives a 2SP/1DP ratio, higher than anything from Kepler or Maxwell. The P100 also has 224 TMUs and massive amounts of cache and register files.
Next, we have the massive 610 mm² die on 16nmFF+. About 50% of that is FP32 CUDA cores, 25% is FP64 and rest on other parts. This means despite the massive die size, the P100 and GP100 derivatives won’t be great gamers, as games generally only use FP32 CUDA cores. There may be a GP100 variant though that swaps out the FP64 cores for FP32 ones. Even saddled with compute though, GP100 will still beat the Titan X by a good margin. TDP is a relatively tame 300W, as expected from the use of 16nm and 16GB of HBM2.
Finally, most marketing statements are hyperbole and the “5 miracles” are no exception. They are the Pascal Architecture, 16nm FinFET, CoWoS with HBM2, NVLink, and New AI Algorithms. Honestly, none of these are really that amazing on their own and have been expecting. Combining all of them in one go on such a massive chip though is pretty amazing though. While the P100 will be shipping soon, don’t expect many till Q1 2017.
One of the inevitable signs of an imminent release of new products is when the old model starts becoming hard to find. A seamless transition to the new version is a mark of good logistics and something Nvidia is known for. In line with the expectations for Pascal, Nvidia has reportedly stopped shipping GTX 980Ti’s to their AiB partners, which indicates that Nvidia is winding down the supply chain for the high-end card.
A stop in GTX 980Ti production points to a Pascal chip to replace it coming soon down the line. Usually, shipments to stores are ahead by a month and production a month or so before that. If Nvidia stops supplies now, there will still be about 2-3months before supplies run low. This puts the timeframe smack dab during Computex where Pascal is expected to be launched. It seems like perfect timing for GTX 980Ti supply to dry up just as Pascal launches and becomes available.
Given the movement to 16nmFF+, we can expect the GTX 1080 to at the very least match the GTX 980Ti. With a replacement product, it makes sense for the GTX 980Ti to cease production now. For now, it seems that Nvidia hasn’t started supplying their partners with Pascal just yet but that should happen shortly if Pascal is to arrive at Computex. The leaked shrouds suggest that the AiB partners have already tooled up in expectation of Pascal. Of course, this is still an unconfirmed report, released on April 1st to boot, so take this with a fist full of salt.
NVIDIA’s new Pascal GPU micro-architecture – billed as 10x faster than the previous Maxwell iteration, and set for release in retail graphics cards later this year – is rumoured to be having problems when dealing with Asynchronous Compute code in video games.
“Broadly speaking, Pascal will be an improved version of Maxwell, especially about FP64 performances, but not about Asynchronous Compute performances,” according to Bits and Chips. “NVIDIA will bet on raw power, instead of Asynchronous Compute abilities.”
“This means that Pascal cards will be highly dependent on driver optimizations and games developers kindness,” Bits and Chips adds. “So, GamesWorks optimizations will play a fundamental role in company strategy. Is it for this reason that NVIDIA has made publicly available some GamesWorks codes?”
This report has not been independently verified, but if it is true, it could spell bad news for NVIDIA, especially since, despite fears to the contrary, its Maxwell architecture was capable of processing Async Compute, and AMD’s Radeon graphics cards currently leading all DirectX 12 benchmarks
We shouldn’t have to wait too long before we find out, though, with WCCFTech reporting that NVIDIA’s flagship Pascal graphics card, the “GTX 1080”, will be unveiled at GTC 2016, which takes place in Silicon Valley, California, from 4th-7th April.
With both the Pascal announcement and GeForce launch coming ins the next 2 months, more information is being leaked about the upcoming Nvidia cards. According to the latest rumour, the first GeForce Pascal card to launch, the GTX 1080, will not be as impressive as many had hoped. As expected from Nvidia, they are keeping with their tradition to launch first with the mainstream GP104 die first in order to maximise yields and profits.
Utilizing the GP104 based on the 16nmFF+ process from TSMC, the GTX 1080 may yet be the fastest Nvidia card yet on the market till the bigger GP100 GeForces launches later. Despite the boost in performance, it appears that Nvidia will be sticking to 8GB of plain old GDDR5X, and not using HBM2 as some have suggested. While GDDD5X does have some disadvantages, it is a decent upgrade over GDDR5 and allows for an earlier launch than using HBM2 as production for those chips are still ramping up.
Furthermore, the leak specifies the display outputs as DisplayPort x2, HDMI x1, DVI x1 and the use of only 1 PCIe 8 pin power connector. This limits power to 225W but with the new architecture and use of 16nmFF+, this may still allow the card to dance with the 980Ti. The launch date is reported as May 27th, just before Computex. Big Pascal GP100 is set to launch before that date though so stay tuned!
After the recent Taiwanese earthquake, many Nvidia and AMD watchers may have worried about their upcoming Pascal and Polaris GPUs. While TSMC did eventually reveal that there would be a hit to their chip production, especially 16nm, it seems like things should be fine. According to the latest reports, TSMC is planning to double their 16nm wafer production from 40,000 per month up to 80,000 per month.
While this number may still be slightly depressed due to the earthquake, it does mean TSMC is taking in more 16nm orders and is able to supply them. Nvidia is relying on TSMC to supply them with 16nmFF+ GPUs for use with Pascal which is set to launch later this year. A ramp up now would mean the a mid-2016 launch for the earliest Pascal chips, right in line with rumours. For AMD, TSMC will play a lesser role as Polaris may be using GlobalFoundries 14nmLPP exclusively.
One snag in the above analysis is that these maybe Apple A10 SoCs. Apple has been moving away from Samsung as their main chip supplier and Apple may be starting to ramp up iPhone SoC orders. Either way, the fact that 16nmFF+ is doing well means the earthquake likely won’t affect chip supply and prices.
NVIDIA’s upcoming architecture codenamed ‘Pascal’ is based on the TSMC 16nm manufacturing process and utilizes the latest iteration of high bandwidth memory. HBM2 will feature bandwidth exceeding 1 Terabyte per second and ship in larger forms compared to HBM1. Some of you might remember, HBM1 was a major talking point on AMD’s Fiji XT line-up but the memory has a limitation of 4GB. Thankfully, HBM2 resolves this and could potentially allow for GPUs sporting a whopping 32GB. While this isn’t confirmed, rumours suggest that NVIDIA could launch a flagship with a huge amount of video memory. However, I personally think a 32GB graphics card will target professionals requiring huge compete power.
It’s important to remember that HBM is up to 9 times faster than the current GDDR5 standard and a huge revolution in memory technology. Pascal will also be the first architecture to utilize NV-Link. This is an energy-efficient, high-bandwidth communications channel that uses up to three times less energy to move data on the node at speeds 5-12 times conventional PCIe Gen3 x16. In theory, this enables faster communication between the CPU and GPU as well as multiple graphics cards. This is only a very brief insight into the potential of Pascal and it could be the biggest step forward in graphics technology we’ve seen in a long time.
Recent information from Zauba.com shows a shipment of Pascal graphics cards which suggests the launch date isn’t too far off. However, according to the Tech Times, Pascal might not launch until June at the earliest! The general consensus is the launch will occur sometime in Q2 this year. With AMD and NVIDIA preparing new architectures, it’s a very exciting time and I cannot wait to see the kind of performance gains compared to the previous generation. Hopefully, the new range offers superb performance at reasonable prices, and AMD keeps NVIDIA in check to drive costs down.
Last month we received word that the first Pascal chips would be launching a bit sooner than expected. At that time, GP100 was expected to drop in April and GP104 in June. According to the latest rumors, it looks like that timetable was accurate, with the GP100 based Tesla chip coming in April around GTC. What’s more, we’re getting more details about when the rest of Nvidia’s Pascal lineup will launch.
As with the previous report, GP104 will arrive in June and it looks like the GTX lineup will be based off that, with both the GTX 1080 and 1070 being GP104 chips. Near the end of the year in Q4, we can expect GP106 and GP107. These will be longer end chips and likely power the GTX 1060 and 1050. Finally, we have the Titan which will use GP100 and a GP108 in early 2017. This follow’s Nvidia’s new trend of releasing a GTX x80 first, followed by the Titan, then finally a GTX x80Ti. While it’s good for Nvidia’s sales, it moves high-end users into a quicker upgrade schedule than if all the cards launched at the same time.
The Tesla launching first makes a whole lot of sense as enterprise users can pay the high premiums for early HBM2 and 16nm. A June launch for GP104 may point to them using HBM2 as the timeline is a bit tight for GDDR5X though it is doable. The biggest question is how well Pascal will perform as it is a stopgap architecture between Maxwell and Volta, like a Maxwellv2 though the die shrink to 16nm should make some great gains in and of itself.
For those waiting on Nvidia’s next-generation GPUs, the wait may not be as long as expected. Last week, we found out that Pascal would be arriving a bit sooner than expected, in the earlier parts of 2H 2016 rather than late in 2016. Now, it looks like Nvidia may be moving even faster than those rumours, with GP100 to arrive in April with GP104 to follow 2 months later in June. What’s more, the GTX 1080 will also debut in June and reportedly be based off GP104; perhaps we’ll see it at Computex 2016?
While GP100 or Big Pascal will launch first in April, that is only for the Titan and various enterprise models. This is in line with what Nvidia has done in the past by launching models with higher margins in order to reduce their risk and grab as much of the early adopter crowd cash. Later on, the more mainstream GP104 will follow up with gaming oriented GeForce models with the compute units cut out. The biggest change is that the Titan will be launching before the GeForce this time.
If the GTX 1080 is based off GP104 as rumoured, this would suggest a GTX 1080Ti based off of GP100 would arrive later on, just like what happened with the 9xx and 6xx series. For those looking to get the very best gaming card for the next generation, waiting may be a smart move. AMD is also set to launch their own Polaris GPUs around the same time though it looks like Nvidia may beat them to the punch with GP100.
Even as this generation’s GPUs are continuing to fly off the shelves, Nvidia is already gearing up for their Pascal launch. Despite being quieter than AMD, it looks like Nvidia will launch their Pascal cards around the same time, in 2H 2016 as AMD’s Polaris will. What’s more, 2H 2016 will see Nvidia’s flagship Pascal GPU based on TSMC’s 16nmFF+ process and utilizing HBM2. This is still a rumour right now but it does fit the time frame since 1H 2016 would be too soon and 2017 too late.
The biggest question is what does”flagship” mean exactly. Ever since GTX 680 was launched, Nvidia has been playing around with the word flagship. Traditionally, the big dies like GF110 would launch first with the smaller mainstream GF104 launching after. Kepler and Maxwell saw that switch with GK104 and GM104 launching ahead of GK110 and GM200 respectively. This suggests that the so-called “flagship” may only be GP104 and not GP100. Even if it is GP100, it may well be a cut-down version, similar to how the GTX 780 was the cut-down variant of the later GTX 780Ti. This strategy does maximize sales for Nvidia but isn’t that great for consumers.
Whatever the card is, be it GP104 or GP100, it is going to use HBM2, giving it at least 512GB/s with 8GB of VRAM but potentially much higher at 16-32GB with 1TB/s+ of bandwidth. With AMD set to launch Polaris around the same time, Q3 2016 should make for exciting times as a slew of new GPUs arrive.
Nvidia has been mum about their upcoming Pascal architecture, even as rival AMD has been showing off their new Polaris chips at CES. Set to launch in 2016 as well, we’re now getting a hint that Pascal may soon be ready. According to shipment tracking site Zabua, several chips likely to be GP104 were sent out on December 29th last year. With engineering samples out, Nvidia may have GP104 ready in about 6 months.
Being a 37.5×37.5mm BGA package, the chip is most likely the GP104, replacing the current GM204 which comes in at about 40x40mm. Since GP104 is “small Pascal”, it will likely come as a GTX 970/980 replacement for the GM204. As with the Kepler and Maxwell releases, Nvidia probably plans to lead with their smaller dies first and later launch their top end GP110 chips in order to maximise sales.
With Pascal engineering samples just shipped, it looks like AMD will have several months lead on Nvidia for the upcoming process node. Given the gains Polaris has shown off, Pascal may do the same. With Maxwell as efficient as it is though, Pascal will probably find most of its efficiency gains from the new 16nm node. It will be interesting to see which direction Pascal will take Nvidia.