Release Date, Prices and Specifications of the GTX 1180: All We Know About Nvidia's Next GPU

Nvidia has long been presenting its latest graphics architecture for game GPUs – more than two years to be precise. This latest architecture was Pascal, and it has powered everything from the top GTX 1080 and GTX 1080 Ti up to the entry-level GTX 1050 and GT 1030. The next generation of Nvidia graphics cards is finally approaching, using the Turing architecture. Here's what we know about the GTX 1180, what we expect in terms of price, specification and release date, and the winding path we have traveled between Pascal and Turing.

The things we "know" about GTX 1180 [19659005ListofThingsWe know – of which we are absolutely certain are correct – can be summed up in one word: nothing. Nvidia has been extremely discreet about her future GPUs this time, and we're not even sure of the name. The rumors of GTX 1180 and GTX 2080 have been turning for months, even though it seems that the 1180 is going to override the official name. We will stay with 1180 for the rest of this room and are pretty confident of the name that it is installed in a cheap photoshop above. (Expect a hasty update if the wind of change begins to make itself felt.) We also do not know what the code name of these new chips will be. The GT104 would be an easy choice, but Nvidia had GT names with Tesla architecture. GTX 280 days (2008-2009). These are all GT200 labels, however, GT100 could still happen.

Although Nvidia has not officially revealed anything, we are 99% sure of three things. First, the new generation architecture is called Turing. Second, it will be manufactured using TSMC's FinFET 12nm process. (We may see later Turing GPUs manufactured by Samsung, as was the case with the GTX 1050/1050 Ti and GT 1030 Pascal parts, but the first pieces will come from TSMC.) Third, the first cards Turing graphics will use GDDR6 memory-not HBM2, due to costs and other factors, but GDDR6 will offer higher performance than the current GDDR5X. Let's hit these last two details a little more in detail

What does the transition to 12nm from 16nm mean in practice? Various sources indicate that the 12nm of TSMC is more of a refinement and tweak to the existing 16nm rather than a true reduction in the size of the entities. In this sense, 12nm is more of a marketing term than a true shrink, but the optimizations made to process technology over the last two years should help improve clock speeds, chip density and the consumption of energy. the execution of chips.

GDDR6 continues in the path of graphics memory traveled from GDDR5 and GDDR5X. During its lifetime, GDDR5 has gone from 3.6 GT / s (ie Giga-transfers per second, but virtually Gbit / s) with the AMD HD 4870 in 2008, at 9 GT / s with the GTX 1060 6GB. GDDR5X has a range of 10-14 GT / s by sending more data per clock rather than higher clock speeds. When the base clock of the GTX 1070 GDDR5 is 2,200 MHz (8,008 MT / s effective), the GTX 1080 has a base clock of 1251 MHz and sends twice as much data per hour (10,008 Effective MT / s). Micron ended up being the only company to produce GDDR5X, Nvidia being the only consumer to manage the GDDR5X at 11 GT / s. The GDDR6 will benefit from much broader support, with the participation of Micron, Samsung and SK-Hynix. GDDR6 has an official target range of 14-16 GT / s, and Micron already shows 18 GT / s modules. The GTX 1180 cards will probably use GDDR6 faster, but the exact clock speeds remain an interrogation point

Expectations for GTX 1180

Turning to what we expect from Turing and GTX 1180, the list is growing considerably. Obviously, performance must be better than existing GPUs and at lower prices for the same level of performance. This is not to say that we will see crazy performance at low prices, but at the very least, we should see the performance levels of the GTX 1080 Ti range between $ 500 and $ 600. Nvidia has multiple paths to provide superior performance to the GTX 1080 Ti, and of which a GTX 1180 is not yet known, so here are the options.

First of all, Nvidia can go with a bigger chip and more hearts. Originally scheduled to arrive last year, Volta has turned into a product that will only see the light in supercomputers, machine learning and professional markets. Volta is incredibly powerful, with the Titan V exceeding the GTX 1080 Ti up to 30%, but it also includes a lot of technology that is of marginal use to the players – in particular, the games do not have the same value. do not need double precision FP64, and they do not need Tensor cores. The simplest solution to consider is that Turing initially comes with a similar design to GV100, but without the TPU or FP64 parts – up to 5,376 CUDA cores would definitely give Turing GPUs a chance in the arm.

to be a deployment similar to Pascal. The first GTX 1180 cards will be launched this year, but they will not be the full version of Turing. Instead, we will get GPUs that look a lot like the current GP102, which means up to 3840 CUDA cores, only with improved efficiency and features and slightly higher clock speeds. Then in 9-12 months we will have Big Turing and GTX 1180 Ti, with more hearts, more memory, and more performance.

But Nvidia is not stuck in a specific core number. If Turing sticks with the 128 CUDA cores by SM, which seems likely, the number of cores present will determine the clock speeds. The Titan V uses 5120 cores up to 1455 MHz, and with a refined 12nm process and changes to the underlying architecture, Turing could run 5120 CUDA cores at 1.5-1.7GHz. Or Nvidia could go with fewer cores and higher clocks, with GTX 1180 being potentially the first Nvidia GPU to ship with stock clocks above 2GHz. But no matter how Nvidia arrives, we expect the performance to be about 25% better than the GTX 1080 Ti FE, just like the GTX 1080 was about 25% faster than the GTX 980 Ti.

What about GDDR6? VRAM will have GTX 1180, and how fast will it work? The safe bet is 8 GB, but 12 GB and 16 GB are also possible. GDDR6 should officially run at 14-16 GT / s, but Micron has already spoken about 18 GT / s. With a 256-bit interface, this would give the GTX 1180 a bandwidth of 448 to 576 GB / s, and improvements in the architecture could allow Turing to better utilize the available bandwidth. My bet would be 16 GT / s GDDR6, with 512 GB / s, as this should be available from several manufacturers

More VRAM is an outdoor possibility, but having 16 GB of VRAM on a graphics card looks a lot like 32 GB system memory: only business applications are likely to use it. Even 8 GB of VRAM is largely exaggerated for games at this time, and with consoles still shipping with 8 GB, it will remain a major target for many years. Plus, going with 8GB on the GTX 1180 leaves the door open for a 12-24 Ti 1180 Ti and / or Titan card in the future. 12 GB with a 384-bit interface could provide 672-864 GB / s of bandwidth, depending on where it is located in the 14-18 GT / s spectrum. 24GB would be almost purely for content creators and professionals, as well as the supercomputer – something that the Tesla V100 already addresses better.

The energy requirements will certainly be higher this turn than the GTX 1070/1080, mainly because the process technology has not changed enough in the last two years. The 250W cards are relatively common these days, with the GTX 1080 using 180W, so the GTX 1180 will likely be in the 200-220W range. The 8-pin plus 6-pin PCIe power connections will likely come on reference models (aka Founders Edition), and the 8-pin dual connectors will be delivered on enthusiastic cards.

And then there is the price, and there is a lot of uncertainty. The shortage of graphics cards caused by the memory offering and cryptocurrency miners' requests is virtually gone, and there are many GPUs that sell at MSRP or less. I've heard rumors ranging from $ 600 to $ 1,000 for the GTX 1180, but the rumors of higher prices came especially when we could not even find a GTX 1080 Ti in stock. Assuming the GTX 1180 meets performance expectations, initial cards should officially be launched at $ 699 to $ 749 for Founders Edition models, and custom cards will arrive a month or two later and cost $ 100 less. It is possible that Nvidia increases initial prices to $ 799 or more, however, it can still reduce prices in the future if necessary.

When is the release date of the GTX 1180?

Earlier rumors have established the date of release of GTX 1180 at the end of July, but according to current indications, this may not happen before mid-August. Nvidia was initially on the chip chip agenda to discuss its next-generation architecture on Aug. 20, but has since been cleared of its schedule. Do not be surprised if the removal was only in name, and that Nvidia comes to Hot Chips for a presentation as originally planned. about the GTX 1170?

Nvidia now has a good path for graphics card launches. It starts with a high-end card like the 780/980/1080, and simultaneously releases or soon after the "responsive" GTX 770/970/1070 alternative. With the 900 series, both pieces were launched at the same time, while the 700 series had a difference of one week and the 10 series had a gap of two weeks between pieces. I think the slightly shifted deployment will also happen with the 11 series, so 2-4 weeks between the 1180 and 1170 launches.

As for the specs, Nvidia's standard practice is to offer a scaled-down version of the same GPU core, essentially harvesting chips that can not run like the full 1180 and selling them as a 1170. The idea is to end up with a performance of about 20-25 percent slower and a price that is 30 to 35 percent lower. If the GTX 1180 has 3584-3840 cores, GTX 1170 will have 2560-2880 cores; If the high end goes with 5120 cores, expect nearly 3840 cores in the GTX 1170. The base and boost clocks are also traditionally slightly lower on the 70 model cards.

In terms of memory, GTX 1170 could use GDDR5 / GDDR5X memory, but GDDR6 seems more likely (so the cards can be the same for the 1170 and 1180). Slightly lower clock speeds for RAM are also typical, so maybe 12-14 GT / s instead of 14-16 GT / s. The new parts will have no less memory than the current generation, so 8GB is the most likely amount for VRAM.

With the reduction in the number of cores and at slightly lower GPU and RAM clock speeds, the GTX 1170 will consume less. A single 8-pin PEG connector may suffice, although 8-pin plus 6-pin seems more likely. The use of power in the range of 180-220W is where this part is likely to land.

Will there be a GTX 1180 Ti or Titan card?

Of course, but not soon. The Titan V exists, and although it is in a different category than previous GTX Titan offers – note the absence of GTX in the name, for example – Nvidia should not withdraw it anytime soon. For business users, the Volta GV100 remains the largest and fastest of the Nvidia GPUs, and Turing relies heavily on players. That means GT104 (or whatever Nvidia calls) as the high-end part of Turing initially, with the potential of GT100 / GT102 in the future.

The first GTX Titan launched before the GTX 780 and with FP64 support upscale. The Titan Black and Xp versions came after the 780 Ti and 1080 Ti, while the Titan X and Titan X (Pascal) came before the 980 Ti and 1080 Ti, respectively. The important distinction is that Nvidia had an existing larger GPU available in the professional space when Titan cards were earlier than Ti cards. The Titan V already uses Nvidia's largest GV100 processor and includes HBM2. Will Nvidia launch GT100 and GT104 at the same time? I would not count on it, but if that happens, expect the prices to be a bit higher than the current 1080 Ti (like $ 999 or more). [ad_2]
Source link