- Mar 3, 2017
- 1,747
- 6,598
- 136
thats intel binning by clock and gpu count. If AMD binned by RAM there should only be 3 SKUs.
It has like 9.It had only 2 SKUs.
That's still nine SKUs for a single part.thats intel binning by clock and gpu count. If AMD binned by RAM there should only be 3 SKUs.
You missed the point. The problem is not the compiler used, but the compiler options used. If one person disables any useful vectorization and the other one enables everything hw can use, comparing scores between what they get is meaningless. What makes sense is comparing AT results with AT past results under the same environment and comparing let's say Huang results with the results he got in the past, but comparing AT SPEC data against Huang's tells you little.But they aren't doing stuff like replacing malloc libraries or using PGO, so I'd argue the results reported by Anandtech (when they had people doing that) or Geekerwan are more useful for comparison than the "official" submissions.
There isn't any effective way to standardize between say macOS and Windows, nor should there be. Pretty much every developer on macOS and iOS uses XCode, so using the latest Xcode release with some basic optimization flags that ordinary developers might use is how Apple Silicon performance should be demonstrated. It doesn't matter if it performs better or worse running Asahi Linux, that's not what 99.9% of Mac buyers are running. Likewise on Windows you'd want to use the MS C Compiler, though there are arguments for using vendor compilers since some developers may do so.
Trying to make them all equal by saying "OK we'll use gcc on everything as the lowest common denominator" might level the playing field, but the information you get doesn't really prove anything. The goal isn't "how does M4 compare against Zen 5" in some sort of abstract sense divorced from the realities of the Mac and PC platforms, the macOS and Windows APIs and development environments, etc. If Zen 5 performs better under Linux or M4 performed better using DDR5 instead of LPDDR5X that's not relevant as far as I'm concerned, because that's not how those CPUs are used (unless you are the 2% or whatever like me and actually do run Linux on your desktop)
The problem is SPEC is a pain to run, so asking people to re-run it just because there's a new compiler rev or something just isn't worth the trouble. If you want something you can run often there's Geekbench. Its just too bad it is so bad as far as benchmark repeatability, but that's going to be true of any benchmark that runs quickly in today's world where you have a bunch of cores and they can all adjust their frequencies moment to moment depending on temperature, load and the phase of the moon.
So since people who so far presented SPEC results are using different environments and different compilers with different compiler options, comparing their results is more like comparing apples to oranges than apples to apples.
You missed the point. The problem is not the compiler used, but the compiler options used. If one person disables any useful vectorization and the other one enables everything hw can use, comparing scores between what they get is meaningless. What makes sense is comparing AT results with AT past results under the same environment and comparing let's say Huang results with the results he got in the past, but comparing AT SPEC data against Huang's tells you little.
Additional problem is the environment, if somebody claims he measures Zen5 core but in reality the environment he is using doesn't allow to enforce this then there is a problem with methodology. I mean it's fine to do system level performance tests but when you want to compare "PPC" like things then let's compare apple to apples...
I completely agree from a user point of view.Trying to make them all equal by saying "OK we'll use gcc on everything as the lowest common denominator" might level the playing field, but the information you get doesn't really prove anything. The goal isn't "how does M4 compare against Zen 5"
What?It seems to have become AMD's MO to announce launch date so far ahead of time to actual launch date that even a dedicated thread to speculation of the chip dies out weeks before the chip actually launches.
Classic AMD launch.
What?
The chips would be in our hands already if it weren't for whatever issue caused them to recall them. They couldn't have forecast that ahead of time.
Considering the situation at Intel, Desktop Zen5 can be postponed until winter just for fun.It seems to have become AMD's MO to announce launch date so far ahead of time to actual launch date that even a dedicated thread to speculation of the chip dies out weeks before the chip actually launches.
Classic AMD launch.
Almost every official CPU announcement from AMD has preceded release by around 1 month. This would have been a bit different with the announcement preceding launch by 2 months, however that doesn't change my point that we would have processors in our hands right now at the very moment you decided to begin to complain if it weren't for an unforeseeable issue.What do you mean, what? Almost the same exact thing happened with Zen 4 launch. They had announcements about announcements about even later launches.
>+15% ST given and blender tease on June 15. August 15 announcement about livestream announcement on August 29. August 29 announcement about September 27 launch.
This thread is exhausted and the desktop chips launch is still a week away, lol.
That reply was for manufacturing cost only.That CPU+dGPU combo is more expensive in every single way that isn't the manufacturing cost of the silicon. More complex PCB, more complex layout, more complex cooling, two different memory pools, etc.
And what do you think I think? Because I didn't say anything about that.It still won't be cheaper, but might not be as much more expensive than you think.
They have to be fake. Didn't we also have Timespy Results above 4k leaked? 3.6k is the best ive seen so far on real units.what about GPD claims?
Comparing ST 7950X and then MT 5950X, I call that <censored> and consider that data point useless.
Well, better not check how many times and how far in advance Intel "previewed/presented/unveiled/talked/preannounced" Meteor Lake and Lunar Lake. I think some of those were like year in advance. I always have problem finding words to describe the nature of the announcement. Not quite launch, not even paperlaunch, but we already used "sneak peek" last time, and they didn't quite "unveil" that much so can't use that word... but you can tell it isn't announcement yet, because they said they are working on it but it will clearly launch only next year and you just know there will be more of these "not announcements" later so you would look silly if you used the word now...It seems to have become AMD's MO to announce launch date so far ahead of time to actual launch date that even a dedicated thread to speculation of the chip dies out weeks before the chip actually launches.
Classic AMD launch.
A made up issue? They announced an official launch date over a month ago, and here we are 4 days post original official launch date and AMD have not even officially announced pricing. Best Buy had to do it, yet its still not official. Its a bit much. And Im making it all up, apparently.That's a made-up issue if I've ever seen one.
The complaint is an odd one. As you correctly point out, the media promo tour surrounding Zen 5 is rather subdued in comparison to many other recent products. Meteor Lake covered a lot of track.Well, better not check how many times and how far in advance Intel "previewed/presented/unveiled/talked/preannounced" Meteor Lake and Lunar Lake. I think some of those were like year in advance. I always have problem finding words to describe the nature of the announcement. Not quite launch, not even paperlaunch, but we already used "sneak peek" last time, and they didn't quite "unveil" that much so can't use that word... but you can tell it isn't announcement yet, because they said they are working on it but it will clearly launch only next year and you just know there will be more of these "not announcements" later so you would look silly if you used the word now...
Or all the Qualcomm announcements spread over 9 months prior to the Snapdragon X (actual) launch. Compared to that, Zen 5 was really kept secret for quite long and not really spammed too much.
Not that I'm criticizing it, it's merely a strategy trying to maintain public and media attention, the more events you make the more coverage. Plus investors see you have not just resigned and quietly awaiting death.
AMD should have delayed it to Sept 1st to ensure all the bugs are ironed out. They risk getting bad reception if some of the above is true on the launch date.Volker Rißka of Computerbase right now:
Beitrag im Thema 'Wochenrück- und Ausblick: AMDs Strix Point überzeugt, Fanatec ist insolvent' https://www.computerbase.de/forum/t...t-fanatec-ist-insolvent.2205209/post-29647520
-Calls ZEN5 "Beta Platform"
-Says consumers should be happy that AMD
pushed it back
-In other Thread he wrote there were 6+ BIOS
Updates during the week
-Says ZEN5 feels rushed
-Says the combination of not ready platform and
CPUs not reaching their Spec would've let to
catastrophic reception, so the Delay was the
right thing to do.
Volker Rißka of Computerbase right now:
Beitrag im Thema 'Wochenrück- und Ausblick: AMDs Strix Point überzeugt, Fanatec ist insolvent' https://www.computerbase.de/forum/t...t-fanatec-ist-insolvent.2205209/post-29647520
-Calls ZEN5 "Beta Platform"
-Says consumers should be happy that AMD
pushed it back
-In other Thread he wrote there were 6+ BIOS
Updates during the week
-Says ZEN5 feels rushed
-Says the combination of not ready platform and
CPUs not reaching their Spec would've let to
catastrophic reception, so the Delay was the
right thing to do.
Sorry, in which test is that? I can't find it.Since you post there you should ask him why in the ST tests he removed CB R15 to replace it with CB R20 while there s already CB R23, and why he s using POVRAY 3.7 wich is known to give an unfair advantage to Intel since it doesnt enable AVX2 for AMD.