It has change into a long-standing custom that with virtually each GPU era, Nvidia does one thing deceptive, anti-consumer, or borderline misleading. In the event you thought that just about two years into the GeForce 40 sequence, they’d break this custom, you had been clearly mistaken.
However what will we imply by this? Let’s return and take the much-loved GeForce 10 sequence for instance. Many are fairly keen on elements just like the GTX 1080 and 1080 Ti. Nevertheless, nearly as good as these high-end fashions had been, Nvidia discovered a solution to undercut finances players with the notorious DDR4 model of the GT 1030.
The GeForce GT 1030 was launched in Might 2017 with 2GB of GDDR5 reminiscence. It was a modest product by all accounts, and with simply 48 GB/s of reminiscence bandwidth, it was no powerhouse. Nonetheless, it was usable, and at simply $70, it wasn’t precisely costly.
However a yr after its launch, Nvidia pulled some of the egregious bait-and-switches the GPU market had ever seen. They started arming some GT 1030s with desktop DDR4 reminiscence, lowering the reminiscence bandwidth to a mere 16.8 GB/s. It turned extraordinarily tough for consumers to find out which mannequin they had been getting, and worse nonetheless, most did not even know there have been two fashions, as each had been merely known as the GT 1030. Briefly, the DDR4 model was, on common, 40% slower for a similar value – it was insane.
There was an affordable quantity of backlash, however that did not cease Nvidia. Whereas they saved issues comparatively professional for the Turing era (RTX 20 / GTX 16 sequence), we nonetheless noticed fashions with a number of variants underneath the identical title.
It wasn’t till Ampere, the GeForce 30 sequence, that Nvidia absolutely reverted to its previous techniques, releasing a number of fashions underneath the identical title. The quietly launched follow-up fashions had been typically worse than the unique variations.
For example, Nvidia launched the 8GB model of the RTX 3060 a yr and a half after the unique 12GB model. The 12GB mannequin got here with GDDR6 utilizing a 192-bit vast reminiscence bus for a bandwidth of 360GB/s. The brand new 8GB version, nonetheless, was compelled to scale back its bus width to 128-bit, slashing the bandwidth by 33% to simply 240GB/s, and in consequence, it was typically a lot slower.
Each fashions had been known as the “GeForce RTX 3060,” however the 8GB model was, on common, 15% slower, with cases the place it was as much as 26% slower. To make issues worse, the 8GB mannequin was usually solely about 5% cheaper, regardless of providing, on common, 15% worse efficiency. So, by way of worth, the 8GB mannequin was a poor deal.
The true challenge, nonetheless, was that most individuals did not even know there have been two variations of the identical product. Once they researched RTX 3060 efficiency, they had been virtually actually benchmarks for the 12GB mannequin, however when it got here time to buy, they may have ended up with the 8GB model, unknowingly making a poor selection.
Now, we arrive on the GeForce 40 sequence. Regardless of a number of lackluster merchandise, such because the $400 RTX 4060 Ti 8GB and the $500 16GB model (although it will definitely noticed a slight low cost), this sequence had been comparatively free from shady practices – till now. Nvidia has quietly launched a brand new model of the GeForce RTX 4070, and whereas the downgrade is minor, the value stays unchanged.
For these unaware, Nvidia launched the GeForce RTX 4070 for $600 a few yr and a half in the past. It featured 12GB of GDDR6X 21Gbps reminiscence, leading to a reminiscence bandwidth of 504GB/s. At launch, it was thought-about one of many higher merchandise within the GeForce 40 sequence, and whereas we weren’t blown away, it supplied respectable worth at $600.
Since then, the RTX 4070 Super has taken its place on the $600 value level, with the unique 4070 dropping to $550 – a $50 low cost. Nevertheless, potential RTX 4070 consumers now face a brand new downside: the mannequin they buy won’t include GDDR6X reminiscence.
Quietly introduced late final month, GDDR6 fashions have began showing on cabinets, so we bought one for $900 Australian – the identical value because the GDDR6X mannequin. The excellent news is that this model has solely been downgraded barely, with 20Gbps reminiscence, lowering the reminiscence bandwidth to 480GB/s – a 5% discount. This would possibly not drastically impression efficiency, however it’s nonetheless a downgrade, and the product title stays the identical.
The reality is, the overwhelming majority of shoppers do not know whether or not the RTX 4070 is meant to come back with GDDR6 or GDDR6X reminiscence. And with out trying it up, most individuals would not know both. In any case, why ought to they? Shoppers belief that they’re getting what Nvidia marketed a yr and a half in the past, however sadly, that is not the case.
Take a look at Setup
| CPU | AMD Ryzen 7 7800X3D |
| Cooling | be quiet! Pure Loop 2 FX 360mm |
| Motherboard | Gigabyte X670E Aorus Grasp [BIOS F32b] |
| Reminiscence | G.Talent Trident Z5 RGB 32GB DDR5-6000 [CL30-38-38-96] |
| ATX Case | MSI Prospect 700R |
| Energy Provide | Kolink Regulator Gold ATX 3.0 1200W |
| Storage | TeamGroup T-Drive Cardea A440 M.2 PCle 4 NVMe SSD 4TB |
| Working System | Home windows 11 |
| Show Driver | Nvidia Recreation Prepared Driver 560.94 |
| AMD Adrenalin Version 24.10.37.10 |
So let’s have a look at how the GDDR6 model compares to the GDDR6X mannequin with some gaming benchmarks. As common, we will likely be testing with our AMD Ryzen 7 7800X3D system, and we have up to date the outcomes for each RTX 4070 fashions.
The unique RTX 4070 outcomes are primarily based on the Founders Edition mannequin, which is clocked simply barely decrease than the Gigabyte Windforce OC GDDR6 model we bought, so core clock speeds are primarily the identical. The important thing distinction right here is the reminiscence velocity, which drops from 21Gbps to 20Gbps.
Let’s try the outcomes…
Benchmarks
Cyberpunk 2077
We can’t spend an excessive amount of time on this information, because it’s primarily to validate the efficiency adjustments. We’ll go over a number of particular person video games at 1440p after which try the 12-game common information at 1080p, 1440p, and 4K.
In Cyberpunk 2077 Phantom Liberty at 1440p, we’re seeing a 5% efficiency discount – that is the most important efficiency drop in any of the benchmarks, spoiler alert. In consequence, the RTX 4070 is now simply 3% quicker than the less expensive 7700 XT.
Dying Mild 2
We noticed a 4% discount in Dying Mild 2, dropping from 89 fps on common to 85 fps. Whereas it is a small discount, it does imply that the RTX 4070 is now slower than the 7800 XT.
A Plague Story: Requiem
Lastly, in A Plague Story: Requiem, we see one other 4% discount in efficiency, dropping from 81 fps to 78 fps – a lack of just some frames. As soon as once more, this strikes the RTX 4070 nearer in efficiency to the less expensive 7700 XT.
Common Efficiency
At 1080p, we noticed a mean 4% efficiency discount, making the GDDR6 model of the RTX 4070 simply 6% quicker than the 7700 XT, regardless of costing over 40% extra. It is a regarding hole in worth.
At 1440p, the GDDR6 model noticed only a 3% efficiency drop in comparison with the unique, however once more, this brings it nearer to the efficiency of the 7700 XT.
At 4K, we see solely a 2% discount in efficiency, as reminiscence bandwidth is not the first bottleneck at this decision.
Energy Consumption
For these excited by energy utilization, when whole system energy consumption, the Gigabyte OC mannequin consumed a number of watts lower than the Founders Version mannequin, however it additionally produced a number of fewer frames. Briefly, energy consumption stays very comparable between the 2.
What We Realized
The excellent news is that the RTX 4070 GDDR6 is not a lot slower, as anticipated, given we’re solely seeing a 5% discount in reminiscence bandwidth. The unhealthy information, nonetheless, is that it’s nonetheless slower. Whereas those that purchase the mistaken mannequin aren’t going through a dramatic efficiency drop, they’re nonetheless dropping out on some efficiency.
In any case, many are prepared to pay a small premium for OC fashions that supply solely 2-5% extra efficiency out of the field, so paying the identical quantity for 2-5% much less efficiency is unlikely to take a seat properly with many.
In truth, we weren’t positive how upset we ought to be about this transfer from Nvidia. Our preliminary response to conditions like that is often fairly adverse – it simply rubs us the mistaken method. To place it bluntly, it is pisses me off. Nvidia marketed the RTX 4070 with 21Gbps GDDR6X reminiscence, and that is what it ought to have. If availability turns into a problem, Nvidia ought to both discontinue the product or re-release it underneath a brand new title.
So, on behalf of shoppers in every single place, we had been annoyed. That is yet one more anti-consumer transfer by Nvidia. However earlier than occurring a full-blown rant about how unhealthy the GDDR6 model of the RTX 4070 is, we thought we would ask the neighborhood what they assume. In any case, if nobody cares, ought to we even trouble?
Just a few days in the past, we ran a ballot on the HUB neighborhood tab, asking how aggravated we ought to be with this case. About 70% of you stated we ought to be ‘extraordinarily aggravated,’ with an additional 20% saying ‘extra aggravated than you’d wish to be.’ In truth, solely 8% stated they had been ‘a little bit aggravated, however it’s not an enormous deal,’ and a mere 2% claimed they had been ‘not aggravated in any respect.’
We made it clear that efficiency had solely regressed by 2-4%. So, with all of the info on the desk, it is evident that the overwhelming majority of you might be annoyed with Nvidia’s transfer – and rightly so. The truth that each fashions carry the identical title, and it is unclear which model is which, is a significant challenge.
Most individuals do not know whether or not the RTX 4070 is meant to come back with GDDR6 or GDDR6X, or that two variations even exist. To make issues worse, the GDDR6 model prices the identical because the GDDR6X model.
It is not like Nvidia is providing a reduction to those that find yourself with the GDDR6 model – they’re merely offering a cheaper-to-produce product and pocketing the distinction. The one silver lining, in our opinion, is that Nvidia is not deceptive shoppers as severely as they did with the GT 1030.
We’re pretty assured that what Nvidia is doing right here may very well be thought-about unlawful in some areas, however they have been utilizing these bait-and-switch techniques for years with out repercussions. So, we count on them to proceed doing so sooner or later. And to be honest, it isn’t simply Nvidia – although the worst examples are likely to contain GeForce merchandise. AMD has additionally misled prospects with dodgy product names, as we have seen not too long ago with the Ryzen 7 5700. So, we’re not simply singling out Nvidia right here.
Purchasing Shortcuts:
