Categories
Computing

Why the RTX 4080 12GB feels a lot like a rebranded RTX 4070

Nvidia announced two versions of its RTX 4080 at its GTC keynote — a 12GB model and a 16GB model. On the surface, this seems simple. Two configurations of the same graphics cards, except with different amount of memory.

This is, after all, what Nvidia did with its RTX 3080 last year. There was the original 8GB RTX 3080, and the 12GB RTX 3080 that got released earlier this year.

But the situation with the two “versions” of the RTX 4080 couldn’t be more different. Not only is there a $300 gulf in price between these two products, but Nvidia confirmed to the media today that they do, in fact, use two different GPUs. The RTX 4080 16GB uses AD104, and the RTX 4080 12GB uses AD103. To call these two products different “versions” of the same graphics card is a pretty serious misnomer.

Nvidia GeForce RTX 4080 16GB Nvidia GeForce RTX 4080 12GB
GPU AD104 AD103
CUDA cores 9,728 7,680
Shader / RT cores 49 / 113 40 / 82
Tensor TFLOPS 780 641
Base clock 2,210MHz 2,310MHz
Maximum clock 2,510MHz 2,610MHz
Memory size 16GB GDDR6X 12GB GDDR6X
Memory bus 256-bit 192-bit
TDP 320 watts 285 watts
Price $1,199 $899

Looking at the other specs we now have, you can see how that plays out. The RTX 4080 16GB has 21% more CUDA cores, 27% more RT cores, and is capable of 18% more Tensor TFLOPS (trillion floating-point operations per second) than the 12GB model. Of course, it also has a wider memory bus and consumes more power too. All in all, the 16GB model is a much more powerful graphics card.

So, what then is going on with the naming of this 12GB RTX 4080? Well, just look at what Nvidia did with its initial launch of the first RTX 30-series cards. At launch, the company announced the RTX 3090, 3080, and 3070. Three GPUs down the line. What it’s doing with the RTX 40-series line is nearly identical, meaning the 12GB 4080, which retails for $899, feels a lot more like a proper RTX 4070 than anything else. That’s a problem, considering the RTX 3070 retailed for just $499.

When asked, of course, Nvidia sees the 16GB model as an “enhanced” RTX 4080, not the other way around. And maybe the company has a point, at least with how these cards are priced. The 16GB model is certainly priced as if it were an RTX 4080 Ti — or something along those lines. Nvidia has also confirmed that there will be no first-party Founders Edition of the 12GB RTX 4080.

Still, the whole thing has left a sour taste in the mouths of PC enthusiasts, who are looking at this 12GB RTX 4080 as a repackaged 4070 as a way to secretly raise prices. Nvidia hasn’t been shy about commenting on the rising cost of GPUs in the future, confirming that falling prices are a thing of the past.

We’ll have to wait and see what Nvidia eventually does with the rest of the lineup to get the full picture, but at the very least, it’s obvious that GPU pricing is continuing to rise, even if some of the costs are buried in the specs.

Editors’ Choice




Repost: Original Source and Author Link

Categories
Computing

Nvidia GeForce RTX 4080 16GB vs RTX 4080 12GB

Nvidia has surprised us all by announcing two versions of the GeForce RTX 4080 instead of an RTX 4080 and an RTX 4070. Following the RTX 4090, the two 4080s will likely be some of the more popular GPUs in the brand-new RTX 40 “Ada Lovelace” lineup.

While the RTX 4070 didn’t make an appearance yet, the two versions of the RTX 4080 give us plenty to get hyped for. Let’s see how they compare to one another.

Specs

Nvidia

While the two GPUs are both called RTX 4080, they differ quite a lot in terms of their specifications. Many leaks suspected that Nvidia would be launching the RTX 4090, RTX 4080, and the RTX 4070 initially. Now, it seems that the RTX 4080 12GB may have inherited some of the specs that were initially leaked as RTX 4070.

The RTX 4080 16GB obviously sports more memory, but interestingly, it’s the RTX 4080 12GB that has slightly higher clock speeds. However, the extra memory and CUDA cores on the RTX 4080 16GB will both have an impact on performance.

Nvidia GeForce RTX 4080 16GB Nvidia GeForce RTX 4080 12GB
CUDA cores 9,728 7,680
Base clock 2,210MHz 2,310MHz
Maximum clock 2,510MHz 2,610MHz
Memory size 16GB GDDR6X 12GB GDDR6X
Memory bus 256-bit 192-bit
TDP 320 watts 285 watts

Expected performance

Comparison of the RTX 4080 16GB and 12GB versions.
Nvidia

Nvidia hasn’t said much about the expected performance of the RTX 4080, so it’s hard to predict how powerful the two GPUs are going to be. We can guess based on their specs, but the real knowledge will come from benchmarks. Fortunately, we’re likely to start seeing them begin to leak out soon, and once the cards are fully out, we should be able to test them ourselves.

The CUDA core volume of the RTX 4080 12GB puts it between the RTX 3070 Ti and the RTX 3080. However, it sports more memory than the RTX 3070 Ti, and also utilizes Nvidia’s latest tech such as DLSS 3 and Shader Execution Reordering (SER). Suffice it to say that we will be seeing an improvement in terms of performance from both cards, but it’s too early to gauge just how they compare to each other.

Nvidia has teased that the RTX 4080 will be two to four times faster than the RTX 3080 Ti, but these numbers may change. It did give us one thing, though — the benchmark linked above that implies the RTX 4080 16GB outperforms the RTX 4080 12GB in each of the three titles, but it’s not a massive difference. However, they both dwarf the RTX 3080 Ti.

Pricing and availability

Nvidia's Ada Lovelace chip.
Nvidia

We don’t have an exact release date for the two RTX 4080 GPUs just yet, but we do know that they will be launching in November this year, so a little later than the flagship RTX 4090.

Once they arrive, the GPUs will be priced at $899 for the RTX 4080 12GB and $1,119 for the RTX 4080 16GB. Custom models from Nvidia’s board partners, such as Gigabyte, Asus, Zotac, MSI, and others, will also be available soon enough, and those might be priced higher depending on their specifications.

It’s a close call

A comparison between the graphics quality without DLSS 3 and with it.
Nvidia

Choosing between the RTX 4080 16GB and the RTX 4080 12GB is going to be a pretty close call once these GPUs are available for sale.

On the one hand, the RTX 4080 12GB is $300 cheaper, and that’s nothing to sneeze at. On the other hand, the 16GB version will, of course, offer better performance, but it’s hard to say whether that difference will be worth $300.

Based on specifications alone, the RTX 4080 16GB will be the better choice, no contest — but if you’re looking for a mix of affordable and powerful, the 12GB option might be the better pick. The RTX 3080 and RTX 3080 Ti will also retain their good value if you’ll be focusing on the price rather than pushing for the latest technologies.

Editors’ Choice




Repost: Original Source and Author Link

Categories
Computing

Gigabyte Inadvertently Confirms 12GB Nvidia RTX 2060 Rumors

When rumors of a 12GB Nvidia RTX 2060 Super refresh started making the rounds, we said that they probably weren’t true. But it looks like we may have been wrong. Graphics card maker Gigabyte filed a new listing with the Eurasian Economic Commission (EEC) that inadvertently confirms this card’s existence..

Twitter user @momomo_us uncovered the listing, which lists four Gigabyte graphics cards. Although the listing doesn’t call out the 12GB RTX 2060 Super by name, the model numbers all line up with previous Gigabyte RTX 2060 cards, with one notable change — 12GB of RAM. The GV-N2060WF2OC-6GD (Gigabyte’s Windforce RTX 2060), for example, is listed as GV-N2060WF2OC-12GD.

Dan Baker/Digital Trends

The listing comes amid mounting evidence for a refresh to Nvidia’s last-gen card. On November 14, a day before the listing went live, YouTube channel Gamers Nexus published a video saying that a 12GB RTX 2060 Super was on the way. This isn’t a channel that normally leaks new releases, but that, combined with the ECC filing and murmurs from around the community, has an 12GB RTX 2060 Super looking likely.

A dedicated leaking channel, Moore’s Law is Dead, revealed in October that the card would arrive in 2022 to take on low-end AMD RDNA 2 graphics cards. Rumors of Nvidia reintroducing the RTX 2060 in some form date back to January 2021, and they haven’t stopped since.

The question: Why? Nvidia released the RTX 3080 more than a year ago, so it’s a strange move to resurrect a GPU that’s more than two years old. There could be a good reason to bring it back, though. It’s no secret that graphics cards are tough to find right now, and Nvidia could be splitting its manufacturing efforts to get more cards out in the wild.

Evidence of the GPU shortage emerged when it was revealed that Nvidia was having manufacturing yield issues with its RTX 30-series graphics cards. Nvidia chose Samsung as its manufacturing partner, and reports circulating shortly after the launch showed that the manufacturer produced fewer usable chips than expected.

Samsung didn’t build the RTX 2060 Super — chipmaker TSMC did. TSMC is the semiconductor company behind AMD’s Ryzen 5000 processors and Radeon RX 6000 graphics cards, as well as a longtime partner for Nvidia. It looks like Nvidia could be splitting its manufacturing to bypass supply chain issues.

That’s something the company did with its GTX 10-series GPUs. The range started on TSMC’s 16nm manufacturing process, but Nvidia eventually moved to Samsung’s 14nm process. Reintroducing the RTX 2060 Super allows Nvidia to quickly produce new cards on a node the company is already familiar with.

The strange bit is the 12GB of video memory. The RTX 2060 Super originally launched with 6GB, and doubling that to 12GB probably won’t do much for gaming performance. That’s something Nvidia’s RTX 3060 proved — even with 12GB of video memory, which is more than the RTX 3080, it performs below other cards in the range.

Unfortunately, an RTX 2060 Super refresh may not be enough to alleviate supply chain issues. Nvidia has been clear that it expects the GPU shortage to continue throughout 2022, so hunting down a graphics card will continue to be a practice in patience.

It’s also possible that the 12GB RTX 2060 Super won’t ever see the light of day. Although multiple sources have confirmed the existence of the card, it’s possible that Nvidia has shelved the idea. That’s something Nvidia already did with the 20GB RTX 3080 Ti, which was reportedly canned earlier this year.

Editors’ Choice






Repost: Original Source and Author Link

Categories
Computing

AMD Claims 12GB of VRAM Is Essential for Gaming GPU

As the new entry-level model in AMD’s Radeon RX 6000 graphics lineup, the Radeon RX 6700 XT comes with more modest specifications — but don’t let the GPU’s humble hardware fool you. The key to the Radeon RX 6700 XT‘s performance in 1440p gaming with maximum game settings is that the card ships with 12GB of VRAM — more than the amount of video memory that ships on most of rival Nvidia’s latest RTX 3000 family.

With 12GB of GDDR6 memory, AMD claims that its GPU can even handle titles that challenged last year’s high-end Nvidia GeForce RTX 2080 Super.

“One of the reasons RX 6700 XT is so great with modern titles is because we’ve equipped the RX 6700 XT with 12GB of GDDR6 to handle increasing texture loads and greater visual demands … at higher resolutions and Max settings,” said AMD corporate vice president Scott Herkelman as he unveiled the graphics card during a streamed presentation.

In a no-so-subtle jab at rival Nvidia’s decision to equip its flagship GeForce RTX 3080 with a paltry 10GB of video memory, Herkelman went on to show that a number of titles are already pushing up against the 10GB threshold and a few already passed that mark. For reference, while the the RTX 3080 has a more limited amount of VRAM compared to AMD’s latest GPU, it does ship with a faster G6X memory standard compared to the G6 format on the RX 6700 XT.

“Games such as Call of Duty: Black Ops Cold War in DX12 Ultra with high texture packs consume approximately 11.1GB [of VRAM] and Dirt 5 in DX12 Ultra high settings consumes 9.5GB,” Herkelman explained. “Horizon Zero Dawn in DX12 Ultimate settings can easily use 10.4GB, and my favorite, Red Dead Redemption 2 — Vulcan, Ultra 9.4GB.”

Calling the Radeon RX 6700 XT the “sweet spot for 1440p gaming,” AMD said equipping the card with 12GB of video memory will make it future-proof to handle today’s high-end titles and the AAA games that will launch in the coming months.

When compared to the modern GPU competition, AMD’s claims appear to be true, at least in the company’s own benchmarks for frame rates on titles like Assassin’s Creed Valhalla, Borderlands 3, Call of Duty: Black Ops  Cold War, Cyberpunk 2077, Dirt 5, Gears 5, Hitman 3, and Watch Dogs: Legion.

When compared against the competing Nvidia GeForce RTX 3060 Ti with 8GB of VRAM and the RTX 3070 with 8GB of VRAM, the Radeon RX 6700 XT delivered better or similar gaming performance at 1440p.

Announcing the AMD Radeon RX 6700 XT. Built for the gaming enthusiast looking for maximum settings at 1440p and high-refresh competitive gameplay. #GameOnAMD pic.twitter.com/rdzqNGWQEj

— Radeon RX (@Radeon) March 3, 2021

In Assassin’s Creed Valhalla, both RTX GPUs delivered frame rates in the range of 60 frames per second (fps), while the RX 6700 XT performed at a better 82 fps. In the titles listed, the RX 6700 XT outperformed its GeForce competitors in all but three games.

The RX 6700 XT outperformed the RTX 3060 Ti but fell short of the RTX 3070’s performance  in Cyberpunk 2077 and Gears 5, but it underperformed both Nvidia cards in Watch Dogs Legion.

The company also went on to highlight how some of its technologies — like Smart Access Memory, Fidelity FX, and FreeSync — will make its ecosystem very competitive against its rival when combined.

Editors’ Choice




Repost: Original Source and Author Link