• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.
  • The forums have been upgraded with support for dark mode. By default it will follow the setting on your system/browser. You may override it by scrolling to the end of the page and clicking the gears icon.

NVIDIA's GB10 Arm Superchip Looks Promising in Leaked Benchmark Results

Nomad76

News Editor
Staff member
Joined
May 21, 2024
Messages
1,369 (3.66/day)
Recent benchmark leaks from Geekbench have revealed that NVIDIA's first Arm-based "superchip," the GB10 Grace Blackwell, is on the verge of its market launch as reported by Notebookcheck. This processor is expected to be showcased at Computex 2025 later this month, where NVIDIA may also roll out the N1 and N1X (MediaTek confirmed in April that their CEO—Dr. Rick Tsai—will be delivering a big keynote speech at Computex 2025 trade show) alternatives tailored for desktop and laptop use. ASUS and Dell have already put the GB10 in their upcoming products while NVIDIA has also used it in its Project DIGITS AI supercomputer. The company announced this machine at CES 2025 saying it would cost around $2,999 and be ready to buy this month.

The benchmark listings show some inconsistencies, like identifying the chipset as Armv8 instead of Armv9. However, they point out that the GB10's Cortex-X925 cores can reach speeds up to 3.9 GHz. The performance results show that the GB10 can compete with high-end Arm and x86 processors in single-core metrics. Yet, Apple's M4 Max processors still leads in this area. The GB10 marks NVIDIA's move into the workstation-grade Arm processor market and could shake up the established players in the high-performance computing field.



View at TechPowerUp Main Site | Source
 
Sounds good enough. Weird thing is that it has been tested on windows instead of Linux, given that Linux support for it should be already in place and I believe most people will use it with such OS.

There's no real reason to expect this to have the fastest CPU out there. Grace on the GH200/GB200 wasn't beating Epyc or Xeons either. The idea of such CPU is to have unified memory with the GPU, and be GREAT and moving data, which it manages to with its NVLink built straight into the CPU, eliminating PCIe bottlenecks.
 
Yeah, I get a GPU article instead of the subject at Tom’s. Maybe it’s a new way for GPU makers to advertise.
 

Attachments

Last edited:
Are you sure that's the right link? Or is Tom's Hardware having yet another server linking problem?

Yeah, I get a GPU article instead of the subject at Tom’s. Maybe it’s a new way for GPU makers to advertise.
No, they simply deleted the whole thing.

To be honest, the 50 series launch has done one big positive thing, for the ones willing to look into it.

The whole thing has been a big FU to everyone but you can see who is deeper than whom in Ngreedia pockets.

How? Actions like this (deleting articles) or anyone writing articles in positive lights that dont benefit us the consumers.
 
Your link redirects me to a list of "Best graphics cards for gaming in 2025" o_O
Huh? Dang, really. :(

Confirmed.


Are you sure that's the right link? Or is Tom's Hardware having yet another server linking problem?
It worked perfectly when I posted. It looks like the article has been removed. It was called "Nvidia GB10 superchip tails behind Apple's M3". It even had benchmark results showing that it was last in the pack compared to competitors.

No, they simply deleted the whole thing.

To be honest, the 50 series launch has done one big positive thing, for the ones willing to look into it.

The whole thing has been a big FU to everyone but you can see who is deeper than whom in Ngreedia pockets.

How? Actions like this (deleting articles) or anyone writing articles in positive lights that dont benefit us the consumers.
Unfortunately, the whole hardware review scene has become a clusterfuck of paid advertising recently. You can't trust any single website these days. :(
 
Huh? Dang, really. :(


It worked perfectly when I posted. It looks like the article has been removed. It was called "Nvidia GB10 superchip tails behind Apple's M3". It even had benchmark results showing that it was last in the pack compared to competitors.


Unfortunately, the whole hardware review scene has become a clusterfuck of paid advertising recently. You can't trust any single website these days. :(
I stopped going to thg, I’m kinda happy the holding killed off anand and didn’t make it go down that path all the way.

The article is pointless, geekbench is an irrelevant test for the product
It might become that when mediatek releases their consumer level chip for windows on arm with optimised schedulers
But this is like putting a van in a lineup of suvs and complaining it accelerate the slowest
 
I stopped going to thg
You lost a hilarious change of heart from THG. Coincidently and jmo, after 20+ years being Intel friendly, they started posting much more AMD friendly articles and reviews, when Intel had to take strict financial measures and cut free fruits, free coffee and some other money streams.
 
I stopped going to thg, I’m kinda happy the holding killed off anand and didn’t make it go down that path all the way.

The article is pointless, geekbench is an irrelevant test for the product
It might become that when mediatek releases their consumer level chip for windows on arm with optimised schedulers
But this is like putting a van in a lineup of suvs and complaining it accelerate the slowest
Synthetics benchmarks are...funny.

But they do bring SOME info to the table.

Now, your comment does fall into the "my favorite corporation does no wrong hence I must blindly defend it."

I would say about that, I am biased against Ngreedia and intel, after disclaiming that, I would not ignore everything that makes AMD looks bad, like every favorites bribed influencer option, the "RT", loves to push.

My suggestion is....look somewhere in between.

That said, my personal and very BIASED opinion is, question everything positive pro-Ngreedia post/articles that you see from all these places because...money is definitely exchanging hands and they dont have your best interest at hand....
 
You lost a hilarious change of heart from THG. Coincidently and jmo, after 20+ years being Intel friendly, they started posting much more AMD friendly articles and reviews, when Intel had to take strict financial measures and cut free fruits, free coffee and some other money streams.
Problem with THG wasnt bias in articles but rather the website itself thats overly bloated with articles and adverts. At one point over a decade back anti viruses simply used to block THG as a malware risk. Look at Techpowerup or Techgage(now shut down) or even Dpreview they do show adverts but they are not in readers face and sparingly used. Overall Toms hardware feels like any other crappy "news" outlets with google advert thrown after every second paragraph or so. As Louis Rossman states these days you really need a condom(adblockers and other tools) to surf internet or to do something as simple as read articles.
 
Now, your comment does fall into the "my favorite corporation does no wrong hence I must blindly defend it."
We have performance benchmarks of the dimensity 9400, which is also powered by the same X925 but at 3630mhz
it scores 2874 geekbench 6 single threaded where the GB10 did 2,960 in the article at 3900mhz without the thermal constraints of being in a mobile device.
So I have doubts on how well things were optimized during that test.

not much fanboying about it.
 
I believe most people will use it with such OS.

wouldent be so sure. They will prolly release this as a windows laptop/nuc. They prolly wouldent invest any money in this if they couldent. So what it will run better, native etc, makes sense on linux? windows users do not care - people still running windows in 2025 are morons by definition. It already makes more sense to run x86 on linux

rest what you said is probably correct

Sounds good enough. Weird thing is that it has been tested on windows instead of Linux, given that Linux support for it should be already in place and I believe most people will use it with such OS.

There's no real reason to expect this to have the fastest CPU out there. Grace on the GH200/GB200 wasn't beating Epyc or Xeons either. The idea of such CPU is to have unified memory with the GPU, and be GREAT and moving data, which it manages to with its NVLink built straight into the CPU, eliminating PCIe bottlenecks.
 
Sounds good enough. Weird thing is that it has been tested on windows instead of Linux, given that Linux support for it should be already in place and I believe most people will use it with such OS.
It’s not a server CPU, it’s a potential competitor to the Apple M chips, so theoretically high powered laptop and AIOs, Mac/MacBook Pro analogues. Makes sense to test Windows ARM for it primarily, that’s what it will run unless NV actually unironically goes for their own OS, which is extremely unlikely.

windows users do not care - people still running windows in 2025 are morons by definition. It already makes more sense to run x86 on linux
Morons, yeah, also known as “people who have software they need running that does not run on Linux”. Christ, the level of discourse on TPU sometimes reaches new lows.
 
Morons, yeah, also known as “people who have software they need running that does not run on Linux”. Christ, the level of discourse on TPU sometimes reaches new lows.
no worries, i can run any windows software i want or need the way i like,

how are you going to run all those x86 apps on your ARM machine? using a "transition layer"? exactly what i do now?

If the cap fits, wear it.

1746878492081.png
 
@fevgatos
I am not actually going to use ARM, I have no idea where you even took that from. And no, there is plenty of software, like scientific tool controls and data analytics that I use in the lab or CNC control software that I know of, that have a direct need for OS to hardware link and they will not run in a VM or a translation layer. You are SOL if you run Linux, full stop.

You don’t know even what you don’t know, as usual. Stop being a clown and don’t insult people for their choices of a fucking OS of all things. I swear, Linux advocates are one of the main hurdles to its adoption - people just don’t want to be associated with such obnoxious behavior.
 
@fevgatos
I am not actually going to use ARM,

Maybe i'll use ARM PC - why not? - i'll just use better. I think it will be seamless transition to me.

Maybe you already use ARM too in your smartphone and maybe its even linux port.

I have no idea where you even took that from. And no, there is plenty of software, like scientific tool controls and data analytics that I use in the lab or CNC control software that I know of, that have a direct need for OS to hardware link and they will not run in a VM or a translation layer. You are SOL if you run Linux, full stop.

You quickly went into finding some niches where - supposedly (im not sure) - windows is still needed.

Well, i attached a picture that i do igpu passthrough and obv can do passthrough of other hardware. I have those VMs mostly because I use legacy PLC, 3D design apps. Working with embedded controllers is more fun in linux.


You don’t know even what you don’t know, as usual. Stop being a clown and don’t insult people for their choices of a fucking OS of all things. I swear, Linux advocates are one of the main hurdles to its adoption - people just don’t want to be associated with such obnoxious behavior.

why should i care? maybe like 15,10 years ago more people using linux would mean, idk, better linux. Now i feel everything is there. They are rewriting GNU to Rust. Crazy timeline, things are getting reversed. Lets not forget AI - ill prolly get better open-source AI agents and windows users will have to use ms ones OR ill run ms AI agents in a container.

Im sorry the truth offends you.
 
wouldent be so sure. They will prolly release this as a windows laptop/nuc. They prolly wouldent invest any money in this if they couldent. So what it will run better, native etc, makes sense on linux? windows users do not care - people still running windows in 2025 are morons by definition. It already makes more sense to run x86 on linux
This specific soc is the one going into the dgx spark product (and the 3rd party offerings), which has always been marketed as a mini AI workstation, so windows definitely doesn't make sense.
If we were talking about the mediatek-based product that is meant for the consumer market, then I wouldn't have made such comment.

It’s not a server CPU, it’s a potential competitor to the Apple M chips, so theoretically high powered laptop and AIOs, Mac/MacBook Pro analogues. Makes sense to test Windows ARM for it primarily, that’s what it will run unless NV actually unironically goes for their own OS, which is extremely unlikely.
See above. You're thinking of the N1(X) chip in partnership with mediatek.
The GB10 is for a mini workstation, it won't be found in laptops.
Nvidia also has released their Ubuntu-based OS for those devices and the GB300 workstation.
 
Lets be honest, people wont buy GB10 for CPU performance. Also most AI/LLM software is Linux based, so Windows performance might also not be the main focus of this platform.
 
wouldent be so sure. They will prolly release this as a windows laptop/nuc. They prolly wouldent invest any money in this if they couldent. So what it will run better, native etc, makes sense on linux? windows users do not care - people still running windows in 2025 are morons by definition. It already makes more sense to run x86 on linux

rest what you said is probably correct

Guy using “wouldent”, “couldent”, failing to capitalise letters calling Windows users “morons”? I guess spellcheckers don’t work on Linux… :p
 
Back
Top