News Nvidia is firing back at AMD, claims Nvidia H100 Is 2X faster than AMD's MI300X

Toggle sidebar Toggle sidebar
Status
Not open for further replies.
Isn't their FP8 and 16 special sauce, non-standard, types?

--
EDIT: https://www.scaleway.com/en/docs/compute/gpu/reference-content/understanding-nvidia-fp8/

Looks like it was ARM, Intel and nVidia pushing the new FP8 type.
--

This feels like nVidia is casually telling people to go out of their way to optimize for a closed technology that will tie them down to nVidia forever... I don't know if this is the underlying message, but it does feel like it for sure.

Oh wait, they have been doing this with CUDA forver. Nevermind.

Regards.
 
Last edited:
  • Like
Reactions: AndrewJacksonZA, helper800, NeoMorpheus and 2 others
hotaru251 said:
rule #1: never trust the results of tests stated by the maker of product....they will ALWAYS cherry pick results.
Click to expand...
Rule #2: Never trust the results from a competitor for a product, as they will ALWAYS (rotten?) cherry pick results.

This is why independent third-party testing is critical for reviews. But I will say, in the case of AI workloads, I would expect properly implemented algorithms to trend toward being highly optimized. Companies are spending millions on the hardware, and they should put forth a real effort to get the software at least reasonably optimized.
 
  • Like
Reactions: AndrewJacksonZA, KyaraM, helper800 and 2 others
-Fran- said:
Isn't their FP8 and 16 special sauce, non-standard, types?

--
EDIT: https://www.scaleway.com/en/docs/compute/gpu/reference-content/understanding-nvidia-fp8/

Looks like it was ARM, Intel and nVidia pushing the new FP8 type.
--

This feels like nVidia is casually telling people to go out of their way to optimize for a closed technology that will tie them down to nVidia forever... I don't know if this is the underlying message, but it does feel like it for sure.

Oh wait, they have been doing this with CUDA forver. Nevermind.

Regards.
Click to expand...
You do understand that this report exposes AMD's evil trickery just to give the ILLUSION, at whatever cost, to gain false momentum? And its not about nothing NVIDIA is doing wrong, huh?
 
-Fran- said:
This feels like nVidia is casually telling people to go out of their way to optimize for a closed technology that will tie them down to nVidia forever... I don't know if this is the underlying message, but it does feel like it for sure.
Click to expand...
The biggest companies are always pushing their proprietary stuff just to make sure you can never leave. The smaller companies tend to use open standards since they don't have the resources to do proprietary stuff. Just look at G-Sync vs Free-Sync. If you have a G-Sync only monitor then you can only use an nVidia GPU to have the variable refresh rate. However, if you have a Free-Sync monitor you can use AMD, nVidia, or Intel GPUs because of the open standard AND the fact that variable refresh rate isn't anything "special." I'm sure that nVidia's FP8 isn't anything special either but they are going to do something with their libraries to lock it to CUDA only.
 
-Fran- said:
Isn't their FP8 and 16 special sauce, non-standard, types?

--
EDIT: https://www.scaleway.com/en/docs/compute/gpu/reference-content/understanding-nvidia-fp8/

Looks like it was ARM, Intel and nVidia pushing the new FP8 type.
--

This feels like nVidia is casually telling people to go out of their way to optimize for a closed technology that will tie them down to nVidia forever... I don't know if this is the underlying message, but it does feel like it for sure.

Oh wait, they have been doing this with CUDA forver. Nevermind.

Regards.
Click to expand...
You forgot the media's favorite lock in tool, DLSS.

That said, You know that Nvidia feels threatened when it responds to competition 😂

Good job AMD
 
  • Like
Reactions: Order 66
DSzymborski said:
Twice as fast is one time faster, not two times faster. Two times faster would be three times as fast.

(Which is one of the reasons why style guides frequently forbid use of the term as it's confusing even when used properly).
Click to expand...
No, that's bad math on your part. 200% faster is three times as fast, but look at what you just said: "Two times faster would be three times as fast." That's basically saying that 2 equals 3. Or to put it another way:

What is anything times (multiplied by) two? Double the value, twice as fast, 100% faster.
 
  • Like
Reactions: Order 66
JarredWaltonGPU said:
No, that's bad math on your part. 200% faster is three times as fast, but look at what you just said: "Two times faster would be three times as fast." That's basically saying that 2 equals 3. Or to put it another way:

What is anything times (multiplied by) two? Double the value, twice as fast, 100% faster.
Click to expand...

He didn't say two times as fast. He said two times *faster*. Twice as fast is 2X. Two times more/faster/larger than X is X + 2X.

If I have 10 apples, two times the apples or twice the 20 apples. But two times *more*, is a quantitative term of addition to the original quantity expressed in terms of the original quantity, so it would be the original 10 apples plus twice the 10 apples, or 30 total apples.

Sorry, this is a particular bugaboo that comes up once in a while professionally.

Otherwise, 200% faster and two times faster would mean different things (and 200% = 2). That's why it's best to avoid the ambiguity.
 
Last edited:
  • Like
Reactions: Order 66 and helper800
DSzymborski said:
He didn't say two times as fast. He said two times *faster*. Twice as fast is 2X. Two times more/faster/larger than X is X + 2X.

If I have 10 apples, two times the apples or twice the 20 apples. But two times *more*, is a quantitative term of addition to the original quantity expressed in terms of the original quantity, so it would be the original 10 apples plus twice the 10 apples, or 30 total apples.

Sorry, this is a particular bugaboo that comes up once in a while professionally.
Click to expand...
I have talked about this and things alike this numerous times in the comments section in different contexts. It is very easy to mix up a word that in math means something completely different from the intent.
 
  • Like
Reactions: KyaraM and Order 66
helper800 said:
I have talked about this and things alike this numerous times in the comments section in different contexts. It is very easy to mix up a word that in math means something completely different from the intent.
Click to expand...

It's definitely a small complaint, but as someone who makes a living in data journalism, it's a particular nitpick of mine. Whether something is twice as fast or three times as fast can be quite important. Ambiguity is much better in Ulysses than reporting!
 
  • Like
Reactions: helper800 and Order 66
DSzymborski said:
It's definitely a small complaint, but as someone who makes a living in data journalism, it's a particular nitpick of mine. Whether something is twice as fast or three times as fast can be quite important. Ambiguity is much better in Ulysses than reporting!
Click to expand...
For me, being an accountant by education, when it comes to financial forecasting and communicating numbers with words, it is extremely important that everyone is on the same page with the numbers and how to express them in words that are disambiguous.
 
  • Like
Reactions: Order 66
helper800 said:
For me, being an accountant by education, when it comes to financial forecasting and communicating numbers with words, it is extremely important that everyone is on the same page with the numbers and how to express them in words that are disambiguous.
Click to expand...

I kinda wonder what the accountant version of stupid journalism fights are.

When I started writing for ESPN in 2010, I basically had to negotiate being allowed to keep Oxford commas (they used AP Style with some sports-specific additions). And I've had an editor want to murder me because I spell out numbers up to ten rather than nine. Despite this, there were lots of writers there that editors hated much more! lol
 
  • Like
Reactions: helper800 and Order 66
DSzymborski said:
I kinda wonder what the accountant version of stupid journalism fights are.

When I started writing for ESPN in 2010, I basically had to negotiate being allowed to keep Oxford commas (they used AP Style with some sports-specific additions). And I've had an editor want to murder me because I spell out numbers up to ten rather than nine. Despite this, there were lots of writers there that editors hated much more! lol
Click to expand...
Its much more mundane on the accounting side. The ways accountants have to make financial statements is usually very specific and leaves no room for choice because you like a certain comma structure. I am most familiar with GAAP, or Generally Accepted Accounting Principles. Most of the actual fighting between accountants is in speculative forecasting because its based a bit on hunches and experience, not that data and analytics are not considered, its actually the opposite. At times accountants get a "feeling" about the data that cannot be quantified conventionally. This is similar to guessing what professional sports team is going to come out on top, there is certainly a lot of data, but how you interpret that data can be one professional opinion or another.
 
  • Like
Reactions: Order 66
Status
Not open for further replies.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom