Nvidia brings liquid cooling to A100 PCIe GPU cards for ‘greener’ datacenters Nvidia's GPUs are becoming increasingly more power hungry, so the US giant is hoping to make datacenters using them "greener" with liquid-cooled PCIe cards that contain its highest-performing chips. At this year's Computex event in Taiwan, the computer graphics goliath revealed it will sell a liquid-cooled PCIe card for its …
I have an RTX 3080 and, although you might say that it is air-cooled, there is still a bunch of liquid in there to get the heat from the GPU to the fans.
I wonder how that will work for the datacenter. For the moment, the A100 doesn't seem to be liquid cooled, but it sure is outrageously expensive.
"looking at liquid cooling as a way to cut down on energy consumption"
Assuming that the laws of thermodynamics didn't change whilst I wasn't looking, just how much truth is there to this? In terms of energy expenditure, just how wasteful is air cooling?
The same amount of heat energy has to get moved from A to B, regardless of the medium.
My guess would be that the efficiency increase is due the the fact the the cooling liquid transfers heat more easily between the chip and the coolant. Air is not very good at conducting heat. A lot of energy goes into spinning the numerous fans in a server, where one small pump can move the cooling liquid.
Also, think about how much quieter a completely liquid cooled data center would be to work in.
Exactly this. Water conducts heat much more readily than air, and can hold a lot more of it, requiring significantly less flow of the cooling medium (in this case water vs. air). So instead of using 100W/server or more on spinning fans (which are also contributing to the heat load), you can spend 4.3kW to move warm water for an entire row of cabinets (using something like the CoolIT CHx750 CDU, which claims to handle up to 750kW heat load). So, if you figure 4.3kW to cool 750kW of server load, that's about 0.06% of power going to cooling, vs. 100W of fans to cool a 600W server, being about 16.7% of power going toward cooling. This is only looking at the server side of things; the facility side is similar as they only have to pump water through mostly-passive coolers, vs. moving around a lot of air through active chillers. Due to the thermal characteristics of water it can operate as a sufficient coolant at much higher starting temperatures than air, adding to the efficiency. I believe something like up to 45C is allowed on the *inlet* side of the server, outlet side can go up significantly from there, which means a passive cooling tower in outdoor ambient air can bring it back down to useful temp in most climates (maybe not in Arizona in the summer, but in most northern climates it's perfectly reasonable).
In simpler terms, you need to cool the air much more and move more of it to get the heat transferred from the chips. You then need to compress the air to heat it up to get the heat out of the air into the environment, you then need to cool the air to start again.
Water can pretty much flow in at room temperature, get heated to chip temperature, flow out, get cooled back to outside air temperature and repeat.
Forgive me if I have missed something here.
Compute and storage (chips and all the gubbins that make up this stuff) produces heat, often lots of it, usually to the point that the required cooling is not far off the input power to the equipment.
So liquid cooling may mean that you can possibly get the heat out of the chips more efficiently so you can load the more (HPC is doing this regularly now) but you still have to get rid of the heat and the simple physics of this does not really change the efficiency.
There will be heat exchangers and pipes already with chilled water, you just have another loop for this and the coolant in this case often has nasty chemicals in.
This sounds like a typical greenwashing to make something sound better than it actually is.
The bottom line is that it does not matter how you try to zazz it up, data centres & IT are never ever going to be "green" or environmentally friendly. There is simply too much power, resources and short life-span in the chain with very difficult recycling at the end.
It's still the same amount of heat that you have to get rid of. Whether you extract that heat, at source, with cooling air, or with a liquid coolant, you still have to get that heat into the air. Being the final dump in all but a few locations. So, all you have added is yet another power requirement. Albeit in a slightly remoter location. So, a more complicated system, with higher power requirements.
This is about the efficiency of moving that heat from the server chips to the chiller plant.
Air-cooled means big fans in the A/C cold outlet to cool air and blow it into the cold aisles, big fans in every server to blow lots of air over the heatsinks, and big fans to move that hot air out of the cabinets and through the A/C intakes to transfer the server heat plus all the fan motor heat into the chiller warm water feed.
Liquid cooled replaces all those fans with a far more efficient pump, that has to move far less volume of liquid.
On top of that, you don't need the chiller to cool the liquid to as low temperatures as the air-cooled, because the heat transfer is so much more efficient. That makes the chiller far more efficient as the temperature difference is greatly reduced. In many climates you might not even need a chiller at all, as your "cold water" feed can be 20 or even 30C, so much higher than ambient.
The difference is several orders of magnitude, assuming you've got a lot of servers.
Early details of the specifications for PCIe 7.0 are out, and it's expected to deliver data rates of up to 512 GBps bi-directionally for data-intensive applications such as 800G Ethernet.
The announcement from the The Peripheral Component Interconnect Special Interest Group (PCI SIG) was made to coincide with its Developers Conference 2022, held at the Santa Clara Convention Center in California this week. It also marks the 30th anniversary of the PCI-SIG itself.
While the completed specifications for PCIe 6.0 were only released this January, PCIe 7.0 looks to double the bandwidth of the high-speed interconnect yet again from a raw bit rate of 64 GTps to 128 GTps, and bi-directional speeds of up to 512 GBps in a x16 configuration.
Interview 2023 is shaping up to become a big year for Arm-based server chips, and a significant part of this drive will come from Nvidia, which appears steadfast in its belief in the future of Arm, even if it can't own the company.
Several system vendors are expected to push out servers next year that will use Nvidia's new Arm-based chips. These consist of the Grace Superchip, which combines two of Nvidia's Grace CPUs, and the Grace-Hopper Superchip, which brings together one Grace CPU with one Hopper GPU.
The vendors lining up servers include American companies like Dell Technologies, HPE and Supermicro, as well Lenovo in Hong Kong, Inspur in China, plus ASUS, Foxconn, Gigabyte, and Wiwynn in Taiwan are also on board. The servers will target application areas where high performance is key: AI training and inference, high-performance computing, digital twins, and cloud gaming and graphics.
After taking serious CPU market share from Intel over the last few years, AMD has revealed larger ambitions in AI, datacenters and other areas with an expanded roadmap of CPUs, GPUs and other kinds of chips for the near future.
These ambitions were laid out at AMD's Financial Analyst Day 2022 event on Thursday, where it signaled intentions to become a tougher competitor for Intel, Nvidia and other chip companies with a renewed focus on building better and faster chips for servers and other devices, becoming a bigger player in AI, enabling applications with improved software, and making more custom silicon.
"These are where we think we can win in terms of differentiation," AMD CEO Lisa Su said in opening remarks at the event. "It's about compute technology leadership. It's about expanding datacenter leadership. It's about expanding our AI footprint. It's expanding our software capability. And then it's really bringing together a broader custom solutions effort because we think this is a growth area going forward."
Science fiction is littered with fantastic visions of computing. One of the more pervasive is the idea that one day computers will run on light. After all, what’s faster than the speed of light?
But it turns out Star Trek’s glowing circuit boards might be closer to reality than you think, Ayar Labs CTO Mark Wade tells The Register. While fiber optic communications have been around for half a century, we’ve only recently started applying the technology at the board level. Despite this, Wade expects, within the next decade, optical waveguides will begin supplanting the copper traces on PCBs as shipments of optical I/O products take off.
Driving this transition are a number of factors and emerging technologies that demand ever-higher bandwidths across longer distances without sacrificing on latency or power.
Lenovo has unveiled a small desktop workstation in a new physical format that's smaller than previous compact designs, but which it claims still has the type of performance professional users require.
Available from the end of this month, the ThinkStation P360 Ultra comes in a chassis that is less than 4 liters in total volume, but packs in 12th Gen Intel Core processors – that's the latest Alder Lake generation with up to 16 cores, but not the Xeon chips that we would expect to see in a workstation – and an Nvidia RTX A5000 GPU.
Other specifications include up to 128GB of DDR5 memory, two PCIe 4.0 slots, up to 8TB of storage using plug-in M.2 cards, plus dual Ethernet and Thunderbolt 4 ports, and support for up to eight displays, the latter of which will please many professional users. Pricing is expected to start at $1,299 in the US.
Analysis After re-establishing itself in the datacenter over the past few years, AMD is now hoping to become a big player in the AI compute space with an expanded portfolio of chips that cover everything from the edge to the cloud.
It's quite an ambitious goal, given Nvidia's dominance in the space with its GPUs and the CUDA programming model, plus the increasing competition from Intel and several other companies.
But as executives laid out during AMD's Financial Analyst Day 2022 event last week, the resurgent chip designer believes it has the right silicon and software coming into place to pursue the wider AI space.
Arm is beefing up its role in the rapidly-evolving (yet long-standing) hardware-based real-time ray tracing arena.
The company revealed on Tuesday that it will introduce the feature in its new flagship Immortalis-G715 GPU design for smartphones, promising to deliver graphics in mobile games that realistically recreate the way light interacts with objects.
Arm is promoting the Immortalis-G715 as its best mobile GPU design yet, claiming that it will provide 15 percent faster performance and 15 percent better energy efficiency compared to the currently available Mali-G710.
Power and thermal management equipment essential to building datacenters is in short supply, with delays of months on shipments – a situation that's likely to persist well into 2023, Dell'Oro Group reports.
The analyst firm's latest datacenter physical infrastructure report – which tracks an array of basic but essential components such as uninterruptible power supplies (UPS), thermal management systems, IT racks, and power distribution units – found that manufacturers' shipments accounted for just one to two percent of datacenter physical infrastructure revenue growth during the first quarter.
"Unit shipments, for the most part, were flat to low single-digit growth," Dell'Oro analyst Lucas Beran told The Register.
In the world of fabless chip designers, AMD, Nvidia and Qualcomm usually soak up the most attention since their chips are fueling everything from top-end supercomputers to mobile devices.
This hunger for compute is what has allowed all three companies to grow revenue in the high double digits recently. But there's one fabless chip designer that is growing faster among the largest in the world and it's far from a household name: Marvell Technology.
Silicon Valley-based Marvell grew semiconductor revenue by 72 percent to $1.4 billion in the first quarter, which made it the fastest growing out of the top 10 largest fabless chip designers during that period, according to financials compiled by Taiwanese research firm TrendForce.
Siemens and Nvidia don’t want manufacturers to imagine what the future will hold – they want to build a fancy digital twin that helps them to make predictions about whatever comes next.
During a press conference this week, Siemens CEO Roland Busch painted a picture of a future in which manufacturers are besieged with productivity, labor, and supply chain disruptions.
"The answer to all of these challenges is technology and digitalization," he said. "The point is, we have to make the digital twin as realistic as possible and bring it as close as possible to the real world."
The Register - Independent news and views for the tech community. Part of Situation Publishing
Biting the hand that feeds IT © 1998–2022
