This seems to be an attempt to compete with people running local models on Apple hardware—even though those local Mac Mini setups aren't really powerful.
I expect we'll get there in a few years, so perhaps this is Nvidia taking an early step in that direction.
In that case, this goes against Anthropic and OpenAI's business models. Which is a double whammy after Jensen Huang's recent comment about how agentic coding will only increase demand for software engineers, not reduced it.
So it also feels like a part of a budding shift in the competitive tension between the various parts of the AI supply chain.
I think this is the first time an ARM windows device gets marketed for gaming. Would be interesting to see what kind of performance hit games have on the x86 to ARM translation layer.
Rosetta on Mac was obviously impressive. There was also impressive Arm->Intel translation in the mobile ecosystem at one time.
One reason it works surprisingly well on modern systems is how much is offloaded to the GPU. You aren't going to get great power optimization or anything without it being truly native though.
There are games which are CPU limited though, and it will be interesting how those do. Curiously those also tend to be in engines with Arm support already.
For Apple use of Rosetta 2 was only temporary as they moved whole lineup to ARM. MS would not abandon x64 anytime soon. So I'm guessing they will try hard to convince developers to release for both architectures.
Honestly this looks like Microsoft must have thrown a pile of money at them to not mention it, as it's just too obviously the main question.
No one seriously cares about this running Windows. We want Steam and CUDA/Ollama, and Windows just gets in the way. nVidia are simply not that oblivious, but I have to admit in their position I'd have considered the Microsoft involvement more trouble than it's worth, which is among the many reasons I'm not a billionaire.
Maybe they think the RAM market is so terrible it will kill the whole initiative regardless.
I don't think there's any incentive for Nvidia to make this a Windows-only device, so most likely it will be fully supported on Linux, just like their GPUs are.
Doesn't it come with Nvidia's blend of Ubuntu with a custom kernel? Do other distros work as well as "DGX OS" or are nvidia's kernel changes pretty important to have?
There are two new things being announced here: the GB10 chip being put into laptops, and GB10 running Windows. GB10 running Linux is not news, it's a product that's been shipping since last fall.
Some competition for Apple in this space and competition for Intel and AMD is great.
But I really do question how well Windows on Arm is really going to work out long term.
For Apple it worked because they were able to force the issue. If you wanted a new Mac it was going to be Arm and we all knew eventually (this year or is it next year?) Intel support would drop. Over time we have seen M series exclusive features.
Developers were forced to update or abandon Mac which gave users a great experience (with some early growing pains).
This is something that Windows will never be able too do. They will always be stuck maintaining an emulator and a likely large subset of apps only supporting one over the other. (also does this work the other way around with an Arm only app working on x86?)
This seems like a repeat of when it was not uncommon for games to only support Intel or AMD or NVIDIA or AMD. But worse since they are not both x86. Sure at least we have emulation but just like with Rosetta2 it shouldn't ever be the long term solution.
For Apple it worked because they waited until they had a really, really good ARM ISA CPU (combined with arguably sandbagging their x86 offering for a few years prior but I digress).
Qualcomm is also working on a really good ARM ISA CPU with their acquisition of NuVia and subsequent Oryon architecture.
Meanwhile this is just using off-the-shelf ARM CPUs in a MediaTek SoC with blackwell bolted to the side of it. ARM's CPUs so far have been subpar for laptop-class chips. Hence why neither Apple nor Qualcomm are using them.
That's surely one thing, Apple went all-in on ARM, for Microsoft it's still a kinda "reduced experience".
But the bigger problem in my opinion: How much of the Windows userbase actually sticks to Windows because of its backwards-compatibility?
--> What would happen if they break this model and the OS is only judged based on its user experience and available applications...?
I'm not sure it would stand any chance to compete in the B2C space. If I think about it, there's not a single new feature in Windows of the last ~20 years I particularly care about.
Without backwards compatibility, there's barely any ecosystem. MacOS on the other hand is full of ecosystem features, improving collaboration, connectivity, handoff across devices, etc.
No thunderbolt is a big no for me. Its one of the greatest feature of MacbookPro that makes it dockable and expandable as a desktop with a good thunderbolt dock.
I am wary of those ARM-based Windows machines because I am unsure how good the ongoing driver support for those SoCs will be. Will they even outlive the Windows version they currently ship with?
Looking at devices like the NVIDIA Shield gives me some hope that NVIDIA will be better than Qualcomm here. I just hope this is not a case where the OEM has to purchase X years of driver support from the chip vendor beforehand, and that NVIDIA will provide support directly itself.
For anyone curious to know how this will fare against Macbooks, at least in CPU perf: DGX Spark has the exact same GPU and CPU as the top RTX Spark laptops will, so you can just directly compare from that.
Of course, DGX Spark is a miniPC, so laptops will likely be slower due to power limits/throttling.
We'll need to wait for the benchmarks, but this looks great! Windows 11 ARM64 is already amazing, and if these really are an upgrade from the Qualcomm chips we're going to have even better laptops on the market.
What is this product anyway? Is it a general purpose CPU or is it specifically designed for MS Windows? Nvidia stepping back from the open source?
"Introducing the NVIDIA RTX Spark™ Superchip. The fusion of NVIDIA AI and RTX graphics in a single chip redefines Windows PCs and delivers amazing creating, AI development, and gaming—on the slimmest, most beautiful RTX laptops ever and small, ultra-efficient desktops."
Its nvidia attempt to gain additional market share and expected as well. If the whole ecosystem is around nvidia and its the easiest way of running stuff, Nvidia offering more enterprise infrastrcuture allows companies to just buy directly nvidia.
Nvidia is also very very rich and pushes the boundaries of stuff. They stoped waiting for industry standards. You can see this in there network stuff. All nvidia.
Next logical step (at least now, not something i thought about) was there CPU for their GPU racks/clusters/systems.
Now they have everything anyway, RTX Spark is just logical.
I don't think its specificly targeted at Apple at all.
Apple has like 10-15% market share and just because some IT nerds buy themselves a mac mini doesn't mean much.
Plenty of them actually just run openclaw without local models. Something which surprised me quite a lot.
But i have two 4090 at home. They consume a lot of power and i had to research the proper Mainboardmodel and had to mod one 4090 to use water cooling because they run too hot.
There Spark setup was at 3k, way to expensive for normal people. If they can get this down and sell more, great for their ecosystem (strengthening it) and getting more money from people.
It does surprise me though that they have enough capacity for this chip and not just putting everyting in Rubin but perhaps the build out has slowed down a little or they start to diverse already for economic savety
All the news articles in my feed mentioned Nvidia reinventing personal computing which is laughable given the specs are worse than the m series. I’m guessing they saw how well Apple devices were selling and rushed to get something similar out so they can ride the hype train and have something to fall back on if ai DC spend slows down.
I wouldn't say it's Nvidia stepping back from open source... if anything this is doubling down on it, as one of the selling points of this is the 128GB of unified memory which will allow for hosting local models (i.e, nvidia's new open model they just released). I guess it's pretty cool, I'm a big supporter of local LLMs/open weight models so seems enticing to me, although I'm not sure this will be super applicable to a lot of regular consumers. Seems like a pretty niche product.
+ battery too. I've wondered if a mini pc with battery would make for a good form factor. I often move between places where I have a desk with a screen but still use a laptop because I want to just suspend and resume. If a mini pc had a small battery just to hold its RAM while suspended I could move between places and just plug in a single USB-C cable and have my full workstation up and running. The thermals could be better than in a laptop and having a built-in UPS better than with a desktop. But last time I checked no one packaged things like that.
But probably worth clarifying it's not a typical "MediaTek CPU" some might assume by that. It has Nvidia's customized ARM CPU implementation + their GPU.
I didn't see this in the article but elsewhere I've seen the memory bandwidth quoted as 600GB/s [1]. For comparison:
- 5090/6000 Pro: 1792GB/s
- 5080:: 960GB/s
- 5070Ti: 892GB/s
- M3 Ultra: 819GB/s
- DGX Spark: 273GB/s (less than an M5 Pro at 307GB/s)
Memory bandwidth isn't everything but it will cap inference rate pretty heavily. Also, the M3 Ultra is for an almost 2 year old Mac Studio. It's widely expected that it'll be refreshed in Q3 with a likely M5 or M4 Ultra with >1000GB/s. I really hope Apple realizes what a market opportunity Apple has here.
The above shows just how good value the 5090 really is. It basically is a stripped down rTX 6000 Pro, which is a ~$10k card, for 20-30% of the price. This also demonstrates how NVidia uses VRAM for market segmentation. As an aside, the true data center cards (eg B100, H100) use HBM memory at ~3.2TB/s.
Unified RAM means its soldered to the mainboard, right?
I'm not sure if I like this. Sure for a laptop this might be not a big problem but if this ARM ecosystem is a success it will spread to desktop computers and I fear we could lose the existing modularity.
ARM64+GPU sure seems like the future. I'm still using my M1 and even that can handle models well, has decent graphics, M5 is a beast, and M6 must surely go even bigger on LLM compute. Now Microsoft has a compelling ARM64+GPU future too.
I have no idea how powerful or power efficient these guys are, but this seems to be the first step in a bigger push towards Windows on ARM (without loosing gaming).
I think more announcements will follow soon from other companies.
It's worth noting that Nvidia power management on Linux has been absymal. There also aren't any of the usual power management options to see how much power things are using, which is quite atypical for a modern system.
Nvidia really threw stuff over the wall with the DGX Spark release. They don't seem to really care. I sort of think they'll spend a little more time on Windows, where there's no pesky upstreaming to do and they can just do whatever, but man, it's such typical hubris from Nvidia to build such an expensive box with good chips but make it basically unsupportable and roasty hot all the time.
You also generally have to run an ever more stale two year old Ubuntu derived DGX OS to get anywhere, with bespoke kernel and drivers all. None of it is well supported, none of it just works like a comparable PC or even well behaved arm system would.
As for other ARM, there were rumors AMD Sound Wave is/was going to be a ~10W arm APU, but there hasn't been much said about it lately. Honestly given the ram crunch, it's maybe just not worth trying to build a system with a cheap core, if the rest of your costs are going to stay so stratospheric.
https://www.techpowerup.com/341848/amd-sound-wave-arm-powere...
Looks like the MSI one might be a 2-in-1, if it has good stylus support I might have a good candidate for an upgrade, thought my ~3-4 year old Galaxy Book is holding up alright for now.
I really like this, but I think the reason Apple Silicon took off was that Apple sort of forced devs to support ARM. Not sure if Microsoft can do the same for Windows…
Developers weren’t really “forced” to support ARM. They simply recognized that all future Macs would be ARM, whereas most new PCs would continue to run on x86. So the incentive to adopt ARM was much weaker on the PC side.
Strix halo's 8060S gpu is very weak, and is roughly equivalent to a 4060 laptop GPU, whereas GB10's gpu is equivalent to a desktop 5070. For LLM throughput, tok/s is similar due to bottleneck by memory bandwidth, but the GB10 has 3x faster prefill. People have also been able to squeeze out much better performance on GB10 using NVFP4 and other improvements in the months after the DGX Spark launch, so don't be misled by early lackluster benchmarks. For the RTX Spark, which also targets gaming and creative applications, the 3x faster GPU is quite nice.
The only Question is is it worth suffering hip and x86? I suspect a lot of folks might like a machine that mimics their GB300 But costs less than a dgx.
Also I heard the tensor core instructions on the dgx are gimped and you’re better off with a rtx pro x000. Is that the same with these machines?
Is CUDA really a lead for long? Aren’t all the latest competitive approaches avoiding all the standard software stacks and writing deeply customized software that is very directly tied to whatever hardware they use?
And is it really a way to lock in people? With AI coding tools, isn’t it trivial to write software on top of CUDA and rewrite it to target some other hardware?
It all sounds good on paper. But I have trouble believing Windows can be a good platform for this. Microsoft has lost all trust after inserting ads into windows, slowly removing power user features, and exploiting every dark pattern they can. And for years, the ARM based Windows laptops have been useless due to app compatibility issues. Why would this change now? Is it priced to be a lot cheaper than Apple’s laptops? Or is this a niche product for AI developers basically?
The "gaming" take is a strange one indeed for an ARM platform. Hopefully they (Microsoft or Nvidia?) put some real effort into the translation layer. They claim modern AAA games, but it is possible they strongarmed the developers to make them an ARM build for a few select titles...
It's clear gaming was not a major concern, it's just "good enough" for someone running AI models and occasionally wants to play some games, not made to primarily play games.
Yep. I noticed the press releases talk about all the partners they have. It seems like a desperate attempt to manufacture a consensus to invest in this new hardware instead of leaving it sort of abandoned like the other Windows ARM stuff. But the problem is that these attempts end up having a few very visible apps working on the architecture and others not actually doing anything substantial.
Sure the graphics capabilities are probably very good. But if you’re a game developer who has traditionally built on Windows on x86 chips, would you want to invest in this new chip or invest in making games for the Apple ecosystem? Aren’t there more new customers to reach in the Apple world than this new Nvidia world?
Who cares about Windows, the goal is to run local AI models similar to AMD Strix Halo and Apple Silicon machines. The OS is honestly a distant last concern as long as the models work well, as you could put Linux on these too, but not sure how well wake lock works.
Anecdotally Windows ARM works fine for me, although to be honest most of my work is command line + browser anyway. WSL works like a treat. Steam installs and most lower end games also play fine on my ARM laptop too. Games that require kernel anticheat don't work.
I think they make a great "second device" where you have something meatier to fall back to if something doesn't quite work right. I'm not sure if it's ready to take on the "main device" role just yet. But it's a far far better experience than the Surface RT days.
This may finally be the chip family ARM on Windows has always needed. Qualcomm's chips have always been dogs with slow off-the-shelf ARM CPU cores that have pathetic single-threaded performance compared to x86 AMD/Intel or ARM Apple Silicon designs.
Qualcomm Snapdragon x1 and upcoming x2 use their Oryon core and have much faster single-thread performance than Intel/Amd and this nvidia soc that uses off-the-shelf arm cores
That wasn't true of the X1, but apparently the X2 (which is only in a single device so far) does appear to finally be fast. The first Windows ARM CPU to be faster than any of its x86 rivals. Competitive with Apple Silicon single-thread performance even.
I was disappointed to see that the RTX Spark has the ARM cores from the DGX Spark. I was hoping it had their new in-house developed cores that Nvidia is starting to use on their latest gen server parts. They look really fast. That said, if RTX Spark has CPU performance like the DGX Spark, it will be almost as fast as the top AMD/Intel parts.
After nvidia's many years of neglecting Linux, paired with direct Microsoft's involvement? Are we going to trust them, to allow installing Linux in these easily?
It won't, the top tier RTX Spark has the same exact CPU and GPU as DGX Spark, so you can check DGX Spark CPU benchmarks to see how it fares. Spoiler: it's about M3 Max level. And they're only coming this fall.
This seems to be an attempt to compete with people running local models on Apple hardware—even though those local Mac Mini setups aren't really powerful.
I expect we'll get there in a few years, so perhaps this is Nvidia taking an early step in that direction.
In that case, this goes against Anthropic and OpenAI's business models. Which is a double whammy after Jensen Huang's recent comment about how agentic coding will only increase demand for software engineers, not reduced it.
So it also feels like a part of a budding shift in the competitive tension between the various parts of the AI supply chain.
I think this is the first time an ARM windows device gets marketed for gaming. Would be interesting to see what kind of performance hit games have on the x86 to ARM translation layer.
Rosetta on Mac was obviously impressive. There was also impressive Arm->Intel translation in the mobile ecosystem at one time.
One reason it works surprisingly well on modern systems is how much is offloaded to the GPU. You aren't going to get great power optimization or anything without it being truly native though.
There are games which are CPU limited though, and it will be interesting how those do. Curiously those also tend to be in engines with Arm support already.
Apple Silicon has a special mode that modified how the ARM chip handles memory transactions to be like x86. Does this nvidia ARM have the same?
What would be interesting to me would be how quickly developers start targeting ARM64 directly.
For Apple use of Rosetta 2 was only temporary as they moved whole lineup to ARM. MS would not abandon x64 anytime soon. So I'm guessing they will try hard to convince developers to release for both architectures.
Can it work with Linux? That's all I care about.
Sort of. It's the same chipset as in the DGX Spark & DGX Station, which run Ubuntu (NVIDIA's flavor).
Honestly this looks like Microsoft must have thrown a pile of money at them to not mention it, as it's just too obviously the main question.
No one seriously cares about this running Windows. We want Steam and CUDA/Ollama, and Windows just gets in the way. nVidia are simply not that oblivious, but I have to admit in their position I'd have considered the Microsoft involvement more trouble than it's worth, which is among the many reasons I'm not a billionaire.
Maybe they think the RAM market is so terrible it will kill the whole initiative regardless.
WSL is the answer in what most folks are concerned.
Has Steam finally started to push for native Linux games instead of translating Windows ones?
1 reply →
I don't think there's any incentive for Nvidia to make this a Windows-only device, so most likely it will be fully supported on Linux, just like their GPUs are.
> just like their GPUs are
So with proprietary blobs that give you more trouble that they're worth?
1 reply →
DGX Spark comes with linux out of the box, it would be hard to imagine this device is not also compatible
Doesn't it come with Nvidia's blend of Ubuntu with a custom kernel? Do other distros work as well as "DGX OS" or are nvidia's kernel changes pretty important to have?
This is strangely absent from the news.
There are two new things being announced here: the GB10 chip being put into laptops, and GB10 running Windows. GB10 running Linux is not news, it's a product that's been shipping since last fall.
It's a collaboration with Microsoft so going to say no, probably not.
Some competition for Apple in this space and competition for Intel and AMD is great.
But I really do question how well Windows on Arm is really going to work out long term.
For Apple it worked because they were able to force the issue. If you wanted a new Mac it was going to be Arm and we all knew eventually (this year or is it next year?) Intel support would drop. Over time we have seen M series exclusive features.
Developers were forced to update or abandon Mac which gave users a great experience (with some early growing pains).
This is something that Windows will never be able too do. They will always be stuck maintaining an emulator and a likely large subset of apps only supporting one over the other. (also does this work the other way around with an Arm only app working on x86?)
This seems like a repeat of when it was not uncommon for games to only support Intel or AMD or NVIDIA or AMD. But worse since they are not both x86. Sure at least we have emulation but just like with Rosetta2 it shouldn't ever be the long term solution.
For Apple it worked because they waited until they had a really, really good ARM ISA CPU (combined with arguably sandbagging their x86 offering for a few years prior but I digress).
Qualcomm is also working on a really good ARM ISA CPU with their acquisition of NuVia and subsequent Oryon architecture.
Meanwhile this is just using off-the-shelf ARM CPUs in a MediaTek SoC with blackwell bolted to the side of it. ARM's CPUs so far have been subpar for laptop-class chips. Hence why neither Apple nor Qualcomm are using them.
That's surely one thing, Apple went all-in on ARM, for Microsoft it's still a kinda "reduced experience".
But the bigger problem in my opinion: How much of the Windows userbase actually sticks to Windows because of its backwards-compatibility?
--> What would happen if they break this model and the OS is only judged based on its user experience and available applications...?
I'm not sure it would stand any chance to compete in the B2C space. If I think about it, there's not a single new feature in Windows of the last ~20 years I particularly care about.
Without backwards compatibility, there's barely any ecosystem. MacOS on the other hand is full of ecosystem features, improving collaboration, connectivity, handoff across devices, etc.
I feel like making universal binaries a thing, and pushing for it to be standard is one viable path.
They already kind of are with ARM64EC, however Windows ecosystem isn't macOS, unless there is market pressure, most shops will keep doing x86/x64.
Microslop doesn’t want people to be able to run their binaries elsewhere, it’s the only reason people buy their product.
1 reply →
No thunderbolt is a big no for me. Its one of the greatest feature of MacbookPro that makes it dockable and expandable as a desktop with a good thunderbolt dock.
Thats also possible with usb-c.
With some caveats, you wouldn't be able to connect two 4k monitors to a dock without TB5.
I am wary of those ARM-based Windows machines because I am unsure how good the ongoing driver support for those SoCs will be. Will they even outlive the Windows version they currently ship with?
Looking at devices like the NVIDIA Shield gives me some hope that NVIDIA will be better than Qualcomm here. I just hope this is not a case where the OEM has to purchase X years of driver support from the chip vendor beforehand, and that NVIDIA will provide support directly itself.
Oh, btw, we are only making 10 of these, the rest of our capacity has been sold off to the large AI firms.
For anyone curious to know how this will fare against Macbooks, at least in CPU perf: DGX Spark has the exact same GPU and CPU as the top RTX Spark laptops will, so you can just directly compare from that.
Of course, DGX Spark is a miniPC, so laptops will likely be slower due to power limits/throttling.
DGX Spark is also $4700
It's been almost 30 years, and a single letter changed. When will we get the Sparkstation, the UltraSpark and the SuperSpark?
SuperSpark and then UltraSpark. And then we can get SparkCube, Sparkii, and SparkiiU.
I'm personally waiting for the OpenSpark.
3 replies →
We'll need to wait for the benchmarks, but this looks great! Windows 11 ARM64 is already amazing, and if these really are an upgrade from the Qualcomm chips we're going to have even better laptops on the market.
"Unified Memory" still means divided address space right? You have to pre-allocate system vs gpu and copy from one to the other?
Awesome, won't be buying it all at current prices but once they calm down, I will very much like to get one.
Around 2-3K USD something with a good GPU + CPU + 128GB of integrated RAM is just going to be an awesome experience.
Considering Mac options are north of 5K+ even on a regular day.
DGX Spark is $4700, so I kind of doubt that RTX Spark's top configs will be cheaper than that.
The DGX also contains the 200 GbE networking and linux support.
6 replies →
isn't dgx ai first and rtx prosumer first. I think it will be cheaper longer term not atm with component inflation
What is this product anyway? Is it a general purpose CPU or is it specifically designed for MS Windows? Nvidia stepping back from the open source?
"Introducing the NVIDIA RTX Spark™ Superchip. The fusion of NVIDIA AI and RTX graphics in a single chip redefines Windows PCs and delivers amazing creating, AI development, and gaming—on the slimmest, most beautiful RTX laptops ever and small, ultra-efficient desktops."
It’s nivdia attempting to compete with Apple’s M-series
Its nvidia attempt to gain additional market share and expected as well. If the whole ecosystem is around nvidia and its the easiest way of running stuff, Nvidia offering more enterprise infrastrcuture allows companies to just buy directly nvidia.
Nvidia is also very very rich and pushes the boundaries of stuff. They stoped waiting for industry standards. You can see this in there network stuff. All nvidia.
Next logical step (at least now, not something i thought about) was there CPU for their GPU racks/clusters/systems.
Now they have everything anyway, RTX Spark is just logical.
I don't think its specificly targeted at Apple at all.
Apple has like 10-15% market share and just because some IT nerds buy themselves a mac mini doesn't mean much.
Plenty of them actually just run openclaw without local models. Something which surprised me quite a lot.
But i have two 4090 at home. They consume a lot of power and i had to research the proper Mainboardmodel and had to mod one 4090 to use water cooling because they run too hot.
There Spark setup was at 3k, way to expensive for normal people. If they can get this down and sell more, great for their ecosystem (strengthening it) and getting more money from people.
It does surprise me though that they have enough capacity for this chip and not just putting everyting in Rubin but perhaps the build out has slowed down a little or they start to diverse already for economic savety
Their target competition is the AMD Strix Halo which is eating the Sparks lunch right now.
Also sounds like they are ditching the discrete GPU altogether.
All the news articles in my feed mentioned Nvidia reinventing personal computing which is laughable given the specs are worse than the m series. I’m guessing they saw how well Apple devices were selling and rushed to get something similar out so they can ride the hype train and have something to fall back on if ai DC spend slows down.
1 reply →
Linux works but MS is just paying them not to mention it.
I wouldn't say it's Nvidia stepping back from open source... if anything this is doubling down on it, as one of the selling points of this is the 128GB of unified memory which will allow for hosting local models (i.e, nvidia's new open model they just released). I guess it's pretty cool, I'm a big supporter of local LLMs/open weight models so seems enticing to me, although I'm not sure this will be super applicable to a lot of regular consumers. Seems like a pretty niche product.
Question is: "Can it run Doom?"
Is this just dgx spark, but a laptop?
yes, same chip
+ Windows
+ Screen
- ConnectX-7 Smart NIC
> - ConnectX-7 Smart NIC
Can the link type be toggled between Ethernet and Infiniband? (Don't think I've ever heard of a laptop with IB.)
+ battery too. I've wondered if a mini pc with battery would make for a good form factor. I often move between places where I have a desk with a screen but still use a laptop because I want to just suspend and resume. If a mini pc had a small battery just to hold its RAM while suspended I could move between places and just plug in a single USB-C cable and have my full workstation up and running. The thermals could be better than in a laptop and having a built-in UPS better than with a desktop. But last time I checked no one packaged things like that.
1 reply →
What about the desktop version? It seemed like it is not a dgx since it has the CPUs cores done by mediatek
7 replies →
Will NVIDIA get a monopoly on providing laptops and desktops with a lot of RAM going forward?
No. You can get a PowerBook today with 128 GB ram.
https://www.bhphotovideo.com/c/product/1957120-REG/apple_mbp...
I'm sorry to announce this to you, but the last PowerBook was released 21 years ago
Or get an AMD 395 laptop or mini PC for half the price of an equivalent mac device
3 replies →
It was wintel (windows + intel) before. This will be what? Windia? Wintek?
Nvideous
Winvidia
Nvidiows
Nvindows
Yeaaaah . But at what Cost though.
They made their own x86 CPU? Or was that part outsourced? Ok ARM MediaTek.
ARM cpu made by MediaTek.
But probably worth clarifying it's not a typical "MediaTek CPU" some might assume by that. It has Nvidia's customized ARM CPU implementation + their GPU.
1 reply →
I think that Nvidia made GPU and CPU, and Mediatek made other parts of SoC necessary for a notebook. Grace is Nvidia's own CPU ARM core
1 reply →
I didn't see this in the article but elsewhere I've seen the memory bandwidth quoted as 600GB/s [1]. For comparison:
- 5090/6000 Pro: 1792GB/s
- 5080:: 960GB/s
- 5070Ti: 892GB/s
- M3 Ultra: 819GB/s
- DGX Spark: 273GB/s (less than an M5 Pro at 307GB/s)
Memory bandwidth isn't everything but it will cap inference rate pretty heavily. Also, the M3 Ultra is for an almost 2 year old Mac Studio. It's widely expected that it'll be refreshed in Q3 with a likely M5 or M4 Ultra with >1000GB/s. I really hope Apple realizes what a market opportunity Apple has here.
The above shows just how good value the 5090 really is. It basically is a stripped down rTX 6000 Pro, which is a ~$10k card, for 20-30% of the price. This also demonstrates how NVidia uses VRAM for market segmentation. As an aside, the true data center cards (eg B100, H100) use HBM memory at ~3.2TB/s.
[1]: https://wccftech.com/nvidia-enters-pc-space-with-rtx-spark/
Spark memory bandwidth is ~300 GB/s. Internal bandwidth is 600 GB/s but that doesn't matter.
Yeah and also the quoted 1 PF is only for sparse models (only half that for dense, if that), and the DGX had serious hardware issues: https://x.com/ID_AA_Carmack/status/1982831774850748825
hope nvidia support driver better than qualcomm. also hope they support linux soon.
Is this finally Macbook Chip Efficiency coming to Windows or will it just be shittier compatibility for slightly better battery life?
I heard leaked geekbench putting it behind the m3, which is couple years old now.
All I care about is if I can get one of these for significantly less than a dgx and get Linux on it for some cuda Blackwell kerneling.
Imagine a Beowulf cluster of these. /slashdot
Unified RAM means its soldered to the mainboard, right?
I'm not sure if I like this. Sure for a laptop this might be not a big problem but if this ARM ecosystem is a success it will spread to desktop computers and I fear we could lose the existing modularity.
I think unified RAM means soldered to the SoC, which is in turn soldered to the mainboard
"Unified" means that it's shared between CPU and GPU, I believe.
But yes, it tends to be soldered on.
No, but LPDDR means soldered, there are no LPDDR dimms
There's LPCAMM2, but it's very recent. The Framework Pro laptop supports it, for example, although only on the Intel variant.
[dead]
ARM64+GPU sure seems like the future. I'm still using my M1 and even that can handle models well, has decent graphics, M5 is a beast, and M6 must surely go even bigger on LLM compute. Now Microsoft has a compelling ARM64+GPU future too.
What does AMD or Intel have here?
I have no idea how powerful or power efficient these guys are, but this seems to be the first step in a bigger push towards Windows on ARM (without loosing gaming).
I think more announcements will follow soon from other companies.
My DGX Sparks are the first and only devices I have with 200W USB-C PD. Low power by AI workstation standards, but intolerable in a laptop.
Intolerable? Why?
2 replies →
It's worth noting that Nvidia power management on Linux has been absymal. There also aren't any of the usual power management options to see how much power things are using, which is quite atypical for a modern system.
Nvidia really threw stuff over the wall with the DGX Spark release. They don't seem to really care. I sort of think they'll spend a little more time on Windows, where there's no pesky upstreaming to do and they can just do whatever, but man, it's such typical hubris from Nvidia to build such an expensive box with good chips but make it basically unsupportable and roasty hot all the time.
You also generally have to run an ever more stale two year old Ubuntu derived DGX OS to get anywhere, with bespoke kernel and drivers all. None of it is well supported, none of it just works like a comparable PC or even well behaved arm system would.
As for other ARM, there were rumors AMD Sound Wave is/was going to be a ~10W arm APU, but there hasn't been much said about it lately. Honestly given the ram crunch, it's maybe just not worth trying to build a system with a cheap core, if the rest of your costs are going to stay so stratospheric. https://www.techpowerup.com/341848/amd-sound-wave-arm-powere...
Looks like the MSI one might be a 2-in-1, if it has good stylus support I might have a good candidate for an upgrade, thought my ~3-4 year old Galaxy Book is holding up alright for now.
I really like this, but I think the reason Apple Silicon took off was that Apple sort of forced devs to support ARM. Not sure if Microsoft can do the same for Windows…
Developers weren’t really “forced” to support ARM. They simply recognized that all future Macs would be ARM, whereas most new PCs would continue to run on x86. So the incentive to adopt ARM was much weaker on the PC side.
They didn’t though. Rosetta 2.
rosetta is a relatively short term solution. will be supported up to macOS 28
Microsoft can do the same for windows - they need to address the fat bundle solution that Apple came up with, but for Windows, though ..
Related:
A powerful new chapter for Windows PCs, accelerated by Nvidia RTX Spark
https://news.ycombinator.com/item?id=48352627
competitor is already on the market and is x86: AMD AI 395+
bechmarks with DGX arnt spectacular for NVIDIAs software and CUDA lead.
wouldnt count on this being a price/compute challenger. especially with overpriced VRAM.
Strix halo's 8060S gpu is very weak, and is roughly equivalent to a 4060 laptop GPU, whereas GB10's gpu is equivalent to a desktop 5070. For LLM throughput, tok/s is similar due to bottleneck by memory bandwidth, but the GB10 has 3x faster prefill. People have also been able to squeeze out much better performance on GB10 using NVFP4 and other improvements in the months after the DGX Spark launch, so don't be misled by early lackluster benchmarks. For the RTX Spark, which also targets gaming and creative applications, the 3x faster GPU is quite nice.
Or like a m4 max? This thing has <300GB/s vs the max with 550GB/s
All those CUDA cores in the sparks but they're starved for memory bandwidth.
I am still waiting for NVidia to release a system that legit beats 3090 maxxing for the home gamer...
1 reply →
The only Question is is it worth suffering hip and x86? I suspect a lot of folks might like a machine that mimics their GB300 But costs less than a dgx.
Also I heard the tensor core instructions on the dgx are gimped and you’re better off with a rtx pro x000. Is that the same with these machines?
Is CUDA really a lead for long? Aren’t all the latest competitive approaches avoiding all the standard software stacks and writing deeply customized software that is very directly tied to whatever hardware they use?
And is it really a way to lock in people? With AI coding tools, isn’t it trivial to write software on top of CUDA and rewrite it to target some other hardware?
yes.
no.
Some other relevant discussions and sources …
NVIDIA and Microsoft Reinvent Windows PCs for the Age of Personal AI
https://news.ycombinator.com/item?id=48352693
2 comments in total there
It all sounds good on paper. But I have trouble believing Windows can be a good platform for this. Microsoft has lost all trust after inserting ads into windows, slowly removing power user features, and exploiting every dark pattern they can. And for years, the ARM based Windows laptops have been useless due to app compatibility issues. Why would this change now? Is it priced to be a lot cheaper than Apple’s laptops? Or is this a niche product for AI developers basically?
A lot of the app compatibility issues on current machines are down to Qualcomm's poor drivers - the actual core bits are mostly okay.
The "gaming" take is a strange one indeed for an ARM platform. Hopefully they (Microsoft or Nvidia?) put some real effort into the translation layer. They claim modern AAA games, but it is possible they strongarmed the developers to make them an ARM build for a few select titles...
It's clear gaming was not a major concern, it's just "good enough" for someone running AI models and occasionally wants to play some games, not made to primarily play games.
Yep. I noticed the press releases talk about all the partners they have. It seems like a desperate attempt to manufacture a consensus to invest in this new hardware instead of leaving it sort of abandoned like the other Windows ARM stuff. But the problem is that these attempts end up having a few very visible apps working on the architecture and others not actually doing anything substantial.
Sure the graphics capabilities are probably very good. But if you’re a game developer who has traditionally built on Windows on x86 chips, would you want to invest in this new chip or invest in making games for the Apple ecosystem? Aren’t there more new customers to reach in the Apple world than this new Nvidia world?
1 reply →
Who cares about Windows, the goal is to run local AI models similar to AMD Strix Halo and Apple Silicon machines. The OS is honestly a distant last concern as long as the models work well, as you could put Linux on these too, but not sure how well wake lock works.
Anecdotally Windows ARM works fine for me, although to be honest most of my work is command line + browser anyway. WSL works like a treat. Steam installs and most lower end games also play fine on my ARM laptop too. Games that require kernel anticheat don't work.
I think they make a great "second device" where you have something meatier to fall back to if something doesn't quite work right. I'm not sure if it's ready to take on the "main device" role just yet. But it's a far far better experience than the Surface RT days.
Hopefully MSFT would look at this as a do or die system, and go all in on improving the user and ownership experience. Will they? Not so sure.
Microsoft sees windows purely as a platform to sell AI products these days.
That's what they're working on, in theory, with Windows K2.
2 replies →
[dead]
So basically Cerebras style?
Not at all. This is a more like what Apple has been doing the past few years. A bunch of decent arm cores paired with a beefy integrated GPU.
No.
This may finally be the chip family ARM on Windows has always needed. Qualcomm's chips have always been dogs with slow off-the-shelf ARM CPU cores that have pathetic single-threaded performance compared to x86 AMD/Intel or ARM Apple Silicon designs.
This will likely have worse single threaded performance than recent Qualcomm CPUs.
These chips also appear to be using off-the-shelf ARM cores.
Qualcomm Snapdragon x1 and upcoming x2 use their Oryon core and have much faster single-thread performance than Intel/Amd and this nvidia soc that uses off-the-shelf arm cores
That wasn't true of the X1, but apparently the X2 (which is only in a single device so far) does appear to finally be fast. The first Windows ARM CPU to be faster than any of its x86 rivals. Competitive with Apple Silicon single-thread performance even.
I was disappointed to see that the RTX Spark has the ARM cores from the DGX Spark. I was hoping it had their new in-house developed cores that Nvidia is starting to use on their latest gen server parts. They look really fast. That said, if RTX Spark has CPU performance like the DGX Spark, it will be almost as fast as the top AMD/Intel parts.
After nvidia's many years of neglecting Linux, paired with direct Microsoft's involvement? Are we going to trust them, to allow installing Linux in these easily?
I don't think so.
This most likely be a winmodem situation, again
DGX Spark has the same soc and ships with Ubuntu
Okay, but still it's highly skeptical trusting MS, and NVIDIA.
[dead]
This will crush the M5 Max going by the numbers. I'm curious to see how much they end up costing
It won't, the top tier RTX Spark has the same exact CPU and GPU as DGX Spark, so you can check DGX Spark CPU benchmarks to see how it fares. Spoiler: it's about M3 Max level. And they're only coming this fall.
Nah, still ~300GB/s memory bandwidth. That will be slower than the M5 max, by a wide margin for LLM inference.
M5 max is 3x stronger and 50% more power efficient. nice try though.
... but you'll be rewriting inference for any model that isn't a well-known LLM. Yourself.
1 reply →