Overclock.net banner

1 - 8 of 8 Posts

·
Just Lift Bro
Joined
·
3,234 Posts
Discussion Starter #1
In November 2019, the company NUVIA broke out of stealth mode. Founded by former senior Apple and Google processor architects, John Bruno, Manu Gulati and Gerard Williams III, the company came crashing out of the gate with quite considerable goals to revamp the server market with an SoC that would provide ‘A step-function increase in compute performance and power efficiency’. Today NUVIA is putting more data behind those goals.
Source
 

Attachments

·
sudo apt install sl
Joined
·
7,334 Posts
Confused why they're comparing their chips to mobile x86 chips while their actual competitors are x86 server chips and Amazon, Ampere Altra and HiSilicon.
 

·
Laptop Enthusiast
Joined
·
9,238 Posts
So the article is literally 100% speculation and blank statements with nothing to back the claim?

AMD shot for a 40% IPC improvement with Zen over Excavator. They surpassed that with a 52% IPC increase, but that was at the same power envelope, roughly. Saying you're going to net a 50% increase at a third the power is where I'm skeptical.
 

·
Hey I get one of these!
Joined
·
7,510 Posts
They're actually using Geekbench... Why?

And they think a phone CPU is 67% the speed of a 4700U. Uh huh.

And then they fail entirely to target multi-thread performance at all and focus on single thread against CPUs that are designed to boost hard for quick time to complete work. Right.

And they're trying to target server land?

Whoever conducted the testing and wrote this up should be fired.

------------------

So, why is it that basically every testing environment like this is literal garbage? I don't understand, people who are pouring thousands of dollars (or hundreds of thousands/millions) sit there with the worlds smallest testing base. They have no variety, they take one kinda-OEM config and treat it like the average.

Security researchers somehow can not come up with at least one CPU from each generation. Testers like this prop up laptops at odd angles and remove the base plate, with their exhaust pointed into the desk because why not, right? They have a fully loaded server right there but they can't just go buy an actual PC to test against? And of course the server doesn't have the lid on it.

Oh but don't worry, they are perfectly willing to attach leads to the laptop's motherboard and inside $1000 cellphones.

Then, to follow that up, they're running not only one of the worst "tests" of performance in general, but their chosen software and hardware don't even simulate the workload they are trying to build for!

------------------

EDIT: These guys do understand that core perf/watt is not the end all of TCO... right? Just converting your software from x86 to ARM adds to the TCO massively. Hardware is never really the cost bottleneck.

EDIT2: And their testing is flawed anyway. I'm not going to inline it because that would be mean, but; https://images.anandtech.com/doci/15483/Power.png
https://www.anandtech.com/show/15483/amd-threadripper-3990x-review/2

AMD is fully capable of doing ~3.5Ghz at 3w/core. 4.3w/core if you count the IO die. This is what they are up against.
 

·
Registered
Joined
·
393 Posts
They're actually using Geekbench... Why?

And they think a phone CPU is 67% the speed of a 4700U. Uh huh.

And then they fail entirely to target multi-thread performance at all and focus on single thread against CPUs that are designed to boost hard for quick time to complete work. Right.

And they're trying to target server land?

Whoever conducted the testing and wrote this up should be fired.

------------------

So, why is it that basically every testing environment like this is literal garbage? I don't understand, people who are pouring thousands of dollars (or hundreds of thousands/millions) sit there with the worlds smallest testing base. They have no variety, they take one kinda-OEM config and treat it like the average.

Security researchers somehow can not come up with at least one CPU from each generation. Testers like this prop up laptops at odd angles and remove the base plate, with their exhaust pointed into the desk because why not, right? They have a fully loaded server right there but they can't just go buy an actual PC to test against? And of course the server doesn't have the lid on it.

Oh but don't worry, they are perfectly willing to attach leads to the laptop's motherboard and inside $1000 cellphones.

Then, to follow that up, they're running not only one of the worst "tests" of performance in general, but their chosen software and hardware don't even simulate the workload they are trying to build for!

------------------

EDIT: These guys do understand that core perf/watt is not the end all of TCO... right? Just converting your software from x86 to ARM adds to the TCO massively. Hardware is never really the cost bottleneck.

EDIT2: And their testing is flawed anyway. I'm not going to inline it because that would be mean, but; https://images.anandtech.com/doci/15483/Power.png
https://www.anandtech.com/show/15483/amd-threadripper-3990x-review/2

AMD is fully capable of doing ~3.5Ghz at 3w/core. 4.3w/core if you count the IO die. This is what they are up against.
I found their article to be rather light on details. For example, what about posting the actual Geekbench results as well as power usage during the tests? It's full of indirection and nebulous claims. It's obvious the chip actually is NOT competitive, but they want to spin it as such, possibly to get funding or something.

Zen 2 desktop CPUs use at most 11 watts per core. As you've stated, you can cut the power constraints down and they still perform great.
 

·
Overclocker
Joined
·
11,399 Posts
Believe it when they actually have developers not only marketing and a final product has been bought in retail by 3rd party testers.

Last time I checked this new Via it was all marketing and they were hiring. Maybe in 2 years they will eventually release something or try run/milk some funding even more.
 

·
Politically incorrect
Joined
·
9,224 Posts
+40-50% marketing for only 33% of the slides. Pretty good performance, I'm extremely excited to see this news about NUVIA being put in the rumour section because it's just PR (propaganda).
 
1 - 8 of 8 Posts
Top