Here at Puget Systems, our customers range from those needing relatively modest workstations, all the way up to those that need the most powerful system possible. These farthermost workstations used to be dual or quad CPU systems, but recently, the shift has been to quad GPU setups. Having this much GPU power can provide incredible performance for a number of industries including rendering, scientific computing, and fifty-fifty has some apply for video editing applications like DaVinci Resolve.

Editor's Note:
Matt Bach is the head of Puget Labs, a division of Puget Systems that does research and benchmarking to test workstation-level software packages. Puget Systems is a specialized builder of gaming and workstation PCs. This commodity was originally published on the Puget blog. Republished with permission.

With Nvidia's latest GPUs, the GeForce RTX 3080 10GB and 3090 24GB, the amount of performance yous tin can get from a single GPU has dramatically increased. Unfortunately, the amount of power these cards require (and conversely the corporeality of oestrus they generate) has also increased. In add-on, well-nigh all the GPU models available from Nvidia and third party manufacturers are not designed for use in multi-GPU configurations which in many cases limits you to merely 1 or two cards.

Notwithstanding, Gigabyte has recently launched a blower-fashion RTX 3090 that should give u.s. our best chance of using 3 or four RTX 3090's in a workstations: the GeForce RTX 3090 TURBO 24G.

This blazon of blower-style cooling system is much better for multi-GPU configurations as it exhausts the majority of the heat directly out the back of the chassis. And when we are dealing with iv 350 watt video cards, that is i,400 watts of heat that we certainly want out of the arrangement as rapidly every bit possible.

While the cooler pattern may exist able to help with the heat output, we too have the problem of total power draw. We should be able to power 1,400 watts with a unmarried one,600 watt ability supply, only that doesn't leave much room for any voltage spikes, non to mention having enough to ability the CPU, motherboard, RAM, storage, and other devices inside the reckoner.

This begs the question: is having 4 RTX 3090s inside a desktop workstation actually feasible? Or is the heat and ability describe too much to handle?

Examination Setup

To see if quad RTX 3090 is something we even want to consider offering, nosotros wanted to put a number of configurations to the examination to look at performance, temperatures, and power depict. The main system we will exist using has the following specs:

Exam Platform
CPU Intel Xeon W-2255 10 Core
CPU Libation Noctua NH-U12DX i4
Motherboard Asus WS C422 SAGE/10G
RAM 8x DDR4-3200 16GB REG ECC(128GB total)
Video Card 1-4x Gigabyte RTX 3090 TURBO 24G
Hard Drive Samsung 970 Pro 512GB
PSU 1-2x EVGA SuperNOVA 1600W P2
Software Windows ten Pro 64-scrap (Ver. 2004)

While our testing with upwardly to three RTX 3090s is fairly straightforward, we have some serious concerns about power describe when nosotros get upward to 4 GPUs. We are going to attempt using just a single 1600W ability supply with stock settings, merely also try setting the power limits for each card to 300W (which should bring us comfortably below the 1600W max), as well as using a dual PSU setup to ensure none of the cards are starved for power.

Power draw is a bit of a business with quad RTX 3090

For our testing, we volition await at functioning, power describe, and GPU temperature with OctaneBench, RedShift, Five-Ray Adjacent, and the GPU Effects portion of our PugetBench for DaVinci Resolve benchmark.

Benchmarks: OctaneBench

OctaneBench is often one of our go-to benchmarks for GPUs considering it runs extremely well with multiple video cards. As you lot tin can see in the charts to a higher place, the scaling from i to iv RTX 3090 cards is about perfect with iv cards scoring exactly four times college than a single RTX 3090.

In fact, the biggest surprise hither was that limiting each of the GPUs to 300W only dropped performance by a little more than 1%. While we are going to talk almost ability depict in more than detail afterward, we will tease that limiting the GPUs reduced the overall system power by ~16%, which is a great render for such a tiny drop in performance.

Benchmarks: V-Ray Next

V-Ray Next is quickly becoming some other staple for our GPU testing because it not merely scales just too equally OctaneRender, but it really causes slightly higher overall system power draw which makes it a great benchmark for stressing GPUs.

Scaling up to four RTX 3090 cards is perfect, and limiting the GPU power reduced the criterion consequence by less than 1%. Nosotros also aren't seeing any increase in performance with dual power supplies, which means that and so far, a single 1600W power supply appears to be doing OK.

Benchmarks: RedShift

RedShift is interesting considering it does not calibration as well as OctaneRender or Five-Ray, but its recent acquisition by Maxon (makers of Cinema4D) means that nosotros are likely to run across more people using it in the near future.

1 thing to notation is that this benchmark returns the results in seconds, so a lower upshot is meliorate (the opposite of our other tests).

In RedShift, we didn't see quite as good of scaling, but four RTX 3090 cards is still 3.half-dozen times faster than a single carte du jour. Again, power limiting the cards and using dual power supplies didn't impact the functioning to a significant degree.

Benchmarks: DaVinci Resolve

To round out our testing, we wanted to expect at something that wasn't rendering. We really were going to include a few other tests such as NeatBench and a CUDA NBody simulation, but either the scaling wasn't very good with multiple GPUs, or we had issues running information technology due to how new the RTX 3090 is.

Our DaVinci Resolve benchmark, however, has back up for these cards and the "GPU Effects" portion of the benchmark scales adequately well up to iii GPUs.

We didn't meet a significant increase in functioning with four RTX 3090 cards, which may in function be due to our choice of CPU. This test is going to load the processor more than any of the others, and while that shouldn't explain the performance wall entirely, information technology may be a contributing factor.

Ability Draw, Thermals, and Noise

Performance is slap-up to wait at, but 1 of the main reasons nosotros wanted to do this testing was to discover if putting four RTX 3090 cards into a desktop workstation was even feasible. The higher power draw and estrus output means that just cards similar the Gigabyte RTX 3090 TURBO 24G with a blower-style cooler will even have a gamble, only these cards are however rated for 350W each.

Power describe was one of our biggest concerns, then we decided to get-go at that place:

We measured the power depict from the wall during each benchmark, which showed us some very interesting results. Starting time, the benchmark that pulled the most power was actually our DaVinci Resolve GPU Effects test. This is likely because it not simply uses the GPUs, merely puts a decent load on the CPU equally well. Because of this, this test is likely a bit more than accurate for the kind of maximum load you lot might put on a system in an "everyday" situation.

Overall, what we found was that while Quad RTX 3090s was able to run on a single 1600W power supply, it is cut is extremely close. Remember that this is power draw from the wall, and going past the rated 92% efficiency of the EVGA unit of measurement nosotros are using, our peak power draw of 1,717 watts should only interpret to 1,580 watts of power internally. That leaves a whole twenty watts to spare!

We didn't have the arrangement close down on us during our testing, but this is way likewise close for long-term use. Not to mention that if nosotros used a more than power-hungry CPU, or even just added a few more than storage drives, we probable would have been pushed over the edge. So, while nosotros technically succeeded with four RTX 3090s on a single 1600W ability supply, that is definitely not something nosotros would recommend.

There are larger power supplies you tin go that go all the style upward to 2400W, but at that point yous are going to desire to bring in an electrician to make sure your outlets tin handle the power draw. The 1717 watts we saw translates to 14.3 amps of power since we are using 120V circuits, and well-nigh household and office outlets are going to be wired to a fifteen amp breaker. That leaves almost no room for your monitors, speakers, and other peripherals. If you lot do go the route of a 2400W PSU, you are going to demand to ensure that you lot are using a 20 amp excursion.

But ability draw bated, how did the Gigabyte cards handle the heat generated by these cards? 1700 watts of ability is more rut than nearly electrical space heaters put out, which can be very hard to handle within a computer chassis.

In the two charts to a higher place, we are looking at the tiptop GPU temperature for each card, every bit well as the peak GPU fan speed for each configuration in OctaneBench. We recorded results for the other tests as well, only OctaneBench was the simply test that ran long plenty to truly get the system upwardly to temperature.

Surprisingly, the temperatures were smashing at all. Even with all iv GPUs running at total speed, the temperatures ranged from 73C on the bottom GPU, to just 80C on the top card However, something to keep in listen is that the temperature of the GPU is only half the picture. GPU coolers - and CPU coolers for that thing - are tuned to increase the fan speed gradually, which means that the temperatures non only have to be acceptable, but there must be adequate fan speed head room to account for higher ambient temperature, longer load times, and the heat generated by other components that might be installed in the future.

In this instance, quad RTX 3090s peaked at 88% of the maximum fan speed. To the states, that is starting to cutting it close, only technically information technology should be enough headroom - specially if yous beefed up the chassis cooling with additional front or side fans.

The last affair nosotros wanted to consider was noise. The temperature of these cards was really fairly decent, but if the system sounds like a jet engine, that may not be acceptable for many users. Noise is very difficult to get across, just we decided the best way would be to record a curt video of each configuration and then you can at least hear the relative deviation.

All things considered, the noise level of four RTX 3090 cards is not likewise bad. It certainly isn't what anyone would call quiet, but for the amount of compute power these cards are able to provide, about terminate users would probable deem it adequate.

Quad RTX 3090: Viable or Fantasy?

Interestingly plenty, we really had very few issues getting four RTX 3090 cards to work in a desktop workstation. Using a blower-way carte like the RTX 3090 TURBO 24G from Gigabyte is certainly a must, but even under load the GPU temperatures stayed below 80C without going above 90% of the maximum GPU fan speed.

The merely true problem we ran into was power describe. We measured a maximum power draw of 1717 watts from the wall, which not only exceeds what we would be comfy with from a 1600W power supply, simply besides ways that you should run your system from a xx amp breaker if possible. Most house and office outlets will be on 15 amp circuits in the The states, which may hateful hiring an electrician to practice some electrical piece of work if you make up one's mind to apply one of the very few 2400W power supplies that are available.

So, will we be offering quad RTX 3090 workstations? Outside of some very custom configurations, we likely are not going to offer this kind of setup to the general public due to the power depict concerns. On the other mitt, triple RTX 3090 is something we are likely to pursue, although that has not notwithstanding passed our full qualification process quite still. Fifty-fifty three RTX 3090 cards is going to give a very healthy performance heave over a quad RTX 2080 Ti setup, all the same, which is slap-up news for users that need faster return times, or those working in AI/ML development.