Today we're taking a look at the Nvidia RTX A4000. That's the full product name, and it probably isn't what some of you think. This isn't a member of the rumored next-gen GeForce RTX 4000 series, but rather an ampere-based GPU designed for workstation PCs.
You may be wondering why, then, since our focus is on gaming hardware, why are we looking at it? We recently received some inquiries about this graphics card, "Would you buy an RTX A4000 for gaming over the RTX 3070, and if so, how much would you also spend on the A4000?"
Interesting point because you can actually buy the RTX A4000 in retail stores. As of this writing, you can buy an RTX 3070 at Newegg and a few other places for around $ 1,000, while the RTX 3070 Ti version starts at $ 1,200. The RTX A4000 is now available for around $ 1,300, or about 30% more than the Vanilla RTX 3070.
Now before we get into testing. What is the RTX A4000?
This is clearly not a GeForce product, we can tell that much from the branding. Earlier this year, Nvidia ditched the "Quadro" nickname for their workstation products and just goes for RTX A2000, A4000, A5000 and A6000 for the various ampere-based workstation GPUs.
In short, the RTX A4000 uses the same GA104 silicon as the RTX 3070 Ti, meaning it has 6144 CUDA cores, 192 TAUs, and 96 ROPs. However, there are a number of important differences between the configuration of the A4000 and the RTX 3070 Ti.
First the kernels. While the RTX 3070 Ti cores work with a base frequency of 1575 MHz with a boost of 1770 MHz, the A4000 cores only clock at 735 MHz with a boost of only 1560 MHz.
Under load, the RTX 3070 Ti should clock at least 13% faster than the A4000. Not much of a difference, but there is one other specification that could have a more serious impact on performance and that is GDDR memory.
|GeForce RTX 3080 Ti||Nvidia RTX A4000||GeForce RTX 3070 Ti||GeForce RTX 3070|
|Price RRP (US $)||$ 1,200||$ 1,000||$ 600||$ 500|
|Release date||June 3, 2021||April 12, 2021||June 10, 2021||October 29, 2020|
|procedure||Samsung 8N||Samsung 8N||Samsung 8N||Samsung 8N|
|Die size (mm2)||628.4||392.5|
|GPU boost clock||1695 MHz||1560 MHz||1770 MHz||1725 MHz|
|Storage capacity||12 GB||16 GB (ECC)||8GB||8GB|
|Storage speed||19 Gbit / s||14 Gbit / s||19 Gbit / s||14 Gbit / s|
|Bus type / bandwidth||384-bit / 912 GB / s||256-bit / 448 GB / s||256-bit / 608 GB / s||256-bit / 448 GB / s|
|Overall performance of the board||350 watts||140 watts||290 watts||220 watts|
The 3070 Ti uses GDDR6X memory with a throughput of 19 Gbit / s and in combination with a 256-bit wide memory bus that offers a bandwidth of 608 Gbit / s.
The A4000, on the other hand, uses more traditional GDDR6 memory, which offers 14 Gbps throughput and using the same 256-bit wide memory bus means memory bandwidth has been reduced by 26% to 448 GB / s.
The RTX A4000 has one advantage, however, and that is storage capacity. With 16 GB of ECC memory as opposed to just 8 GB for the RTX 3070 Ti. We therefore suspect that many of you will be interested in the A4000.
In theory, you should be able to overclock the A4000 to get RTX 3070-like performance, but that all depends on how far you can push the cores and memory, and that's, of course, we're going to investigate.
With that in mind, we should mention that this review focuses solely on gaming performance. We don't care about the advantages of the A4000 for video editing, 3D rendering or CAD work. This might be something we explore in a separate piece, but for this one we have a ton of game data to search through, and of course the whole point of buying the A4000 was to answer the question: is it worth buying over the RTX ? 3070 for games?
The first red flag for the A4000 was immediately the tiny single-slot cooler. Since this is a professional product, Nvidia does not allow partners to design their custom PCBs and coolers. Due to the lower core clocks and the more energy-efficient GDDR6 memory, the A4000 only offers a total board power of 140 watts, almost 40% less than the RTX 3070 and a little more than 50% less than the 3070 Ti.
Since the RTX 3060 has a TDP of 170 watts, it would be pretty amazing if the A4000 could get close to the RTX 3070. But given the tiny single-slot blower fan cooler, it's unlikely to be very quiet, and worse, overclocking for maximum performance could be an issue. Hopefully at some point in the future I can get a water block for the A4000, but for now we'll have to rely on the reference cooler.
The RTX A4000 ran at 1425 MHz when gaming and the fan spun at 2800 RPM which was loud, but it wasn't crazy and really much better than I expected at 47 dBA. When overclocked, it reached a clock rate of 1570 MHz while gaming, which corresponds to a frequency increase of only 10%. However, the GDDR6 memory was overclocked from 14 Gbit / s to 16 Gbit / s, which corresponds to a significant increase of 14%. This overclock was 100% stable and withstood several different gaming loads over a period of 48 hours.
Unfortunately, despite the A4000 overclocking, a stock RTX 3070 Ti clocks about 18% higher for both core and memory, which means that it will be very difficult for the A4000 to keep up, let alone keep up with the RTX 3070 the Ti model. At this point we should probably only dive into the benchmarks with our Ryzen 9 5950X test system. Let's get into that …
Starting with F1 2020 at 1440p, you may be a little disappointed when you learn that the A4000 can only keep up with the RTX 3060 Ti when shipped from the factory and was only 7% faster with our maximum overclock, so it can keep up with the 6700 XT.
Unfortunately, that meant it was 12% slower than the RTX 3070 and 15% slower than the 3070 Ti, not a huge head start, although we're approaching a completely different level of performance with these GeForce models.
Margins are similar at 4K and while the A4000 performs superbly here, it was also a bit slower than the RTX 3060 Ti.Overclocked, the frame rate was increased by 12% this time, but again it only reached the 6700 XT, which means that it lagged far behind the 3070 and 3070 Ti.
Moving on to Cyberpunk 2077, I would say the A4000 is even more disappointing, only matching the RTX 2070 Super averaging 52 fps at 1440p. Overclocking was good for an additional 10%, which allowed for 57 fps on average, but that meant the RTX 3070 was still 12% faster while the 3070 Ti was 18% faster, or nearly 30% faster, than the stock A4000.
The A4000 was more competitive at 4K and was the equivalent of the 6700 XT in production. Overclocked, it was able to keep up with the RTX 3060 Ti, although we are only talking about an average of 28 fps here, which is not playable and certainly not acceptable given the price.
Like most modern graphics cards, the A4000 in Death Stranding is blazingly fast at 1440p with the maximum quality settings in the game and delivers an average of 123 fps or 133 fps with overclocking. That meant stick, it was equivalent to the RTX 3060 Ti and wasn't much slower than the RTX 3070, so here a good result.
At 4K it dropped a bit and now it was 16% slower than the RTX 3070 without overclocking and 22% slower than the 3070 Ti. Although it can deliver very playable performance in Death Stranding at 4K, it's a bit slower than the GeForce -Equivalents.
The results of Horizon Zero Dawn 1440p look a bit more positive, but that's mainly because the faster GeForce GPUs appear artificially limited, with the 3060 Ti only slightly behind the 3070 and the A4000 very close to the RTX 3070 , especially when overclocked.
Margins stayed very similar at 4K, the A4000 was slightly slower in stock than the RTX 3060 Ti, and overclocked it was slightly slower than the RTX 3070, which meant the 3070 Ti was 14% faster than the overclocked A4000. So not a huge margin, but also less than ideal, since the A4000 costs more and really has to be overclocked.
Rainbow Six Siege's margins are rough for the A4000, though the overall performance is pretty extreme at 222 fps out of the box. However, with our maximum overclock it could only keep up with the RTX 3060 Ti, making the RTX 3070 18% faster and the 3070 Ti 25% faster.
The 4K results were more or less the same. Comparing the stock configurations, the RTX 3070 was 30% faster than the A4000, while the 3070 Ti was almost 40% faster and the A4000 was only increased 12% by our overclocking.
The last game we'll be looking at is Watch Dogs Legion, and that's where we're seeing more of it, although the overclocked A4000 does reasonably well compared to the RTX 3070 and is only 4% behind it at 1440p.
It's a similar story at 4K, so I'd say overclocking is one of the better results for the A4000, although the standard still only matches the RTX 3060 Ti.
Where the A4000 really shines is the power consumption and this allows Nvidia to get away with the tiny single-slot cooler. The total consumption of the system with the A4000 was only 312 watts or 320 watts with overclocking. Overclocking actually improved efficiency, but of course there is no guarantee that all models at the voltage the A4000 is set to will reach the same frequency as ours.
In this regard, the A4000 was only fed 770 millivolts under load, while the RTX 3070 Ti was fed with 1040 millivolts in power consumption, as was the use of GDDR6X memory. Unfortunately, Nvidia doesn't allow tinkering with the A4000's voltage, or at least I haven't figured out how to do that yet.
Overclocking aside, it's pretty remarkable how efficient the A4000 is as it uses slightly less power than the Radeon RX 6600 XT in Doom at 1440p while delivering 20% more power at the same time. Or compared to the RTX 3060, we see a reduction in overall system usage of 13% with the same level of performance.
At this point we looked at benchmarks for 6 of the 12 games we tested, but we think by this point you have a good idea of where the RTX A4000 stands. So let's take a look at the average data from 12 games.
Based on the 1440p results, we can see that the A4000 is, on average, a bit slower than the RTX 3060 Ti, so not that great given the price. The best I could get with the fixed voltage was an 11% increase in performance and that allowed the A4000 to beat the RTX 3060 Ti, albeit a stock 3060 Ti. That means, at best, an RTX 3070 series 6% faster and a 3070 Ti 14% faster.
That's actually not bad considering how much more efficient the A4000 is and the performance of the Radeon 6700 XT is obviously very good, but again it comes back to the price. But with pricing and availability horrific for all parts right now, I can see how the A4000 could make sense for some of you.
The margins compared to the faster RTX 3070 and 3070 Ti remain about the same at 4K, so the larger memory buffer of the A4000 is not yet used here, but that is not surprising, it will probably take a few years until then, which gives the A4000 a performance advantage .
What we learned
So there you have it, in short, the Nvidia RTX A4000 works much like a GeForce RTX 3060 Ti, which could make it a viable option depending on price and availability in your region. In Australia, however, the A4000 doesn't make sense for players with a $ 2,000 price tag on stock lists. That's because the RTX 3060 Tis can easily be had for $ 1,250, while an in-stock RTX 3070 Ti costs around $ 2,000.
For gamers, the A4000's only potential benefit is the much larger memory buffer, but right now it will do very little for you, while the 3070 Ti's much faster memory will get you better across the board, as will the higher operating frequency of the cores .
The other advantage of the A4000 is its efficiency, but that gaming advantage is lost due to the single-slot cooler, making the A4000 a louder and hotter product. Gamers also typically don't care that much about efficiency, and if you do, you can still underutilize a GeForce graphics card.
The A4000's limited overclocking headroom was disappointing, but your hands are tied with no way to unlock the voltage regulation. Right now I don't think there is a water block for the A4000, but if there was and someone can find a way to remove the voltage lock that would be a really interesting test. It's worth noting, however, that the VRM was designed for a 140W configuration, so it may not need much more power to get 3070-like performance. Anyway, at the moment it looks like you can only get about 10% more out of the A4000.
It is worth mentioning that all GeForce RTX features such as DLSS, Reflex, Raytracing, DSR, Shadowplay and Nvenc are supported and work the same with the A4000 as with a GeForce GPU. Nvidia has a couple of different driver branches. We installed the Production Branch / Studio driver and it worked fine in the games tested for this article and the performance was unchanged, and so did the new Feature Branch driver. To access the GeForce features, you can install GeForce Experience manually and everything will work fine with these drivers.
So, would we buy an RTX A4000 over the RTX 3070 for gaming? And if so, how much would you also spend on the A4000? We would certainly consider buying it, but we wouldn't do it at an additional cost. So the A4000 would likely have to cost somewhere between the 3060 Ti and 3070 Ti, and likely a lot closer to the 3060 Ti.