Question Zen 6 Speculation Thread

Page 190 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
Toggle sidebar Toggle sidebar
poke01

poke01

Diamond Member
Mar 8, 2022
3,723
5,054
106
511 said:
He is just spewing nonsense how can AMD make N2X when it is not in HVM from TSMC same for N2P the silicon they are receiving is Vanilla N2 rn it just happens that N2/N2P share the PDK with everything same N2P is H2 26 HVM how can we get products before HVM lol

View attachment 127018
Click to expand...
Yeah, N2X isn’t happening. Maybe a refresh in late 2027 but definitely not in 2026.
 
L

LightningZ71

Platinum Member
Mar 10, 2017
2,305
2,885
136
Saylick said:
It just shows you that the tech rumor mill doesn’t actually understand tech. Their main motivation is to be first to the press and generate impressions, not to be accurate and uphold high journalistic integrity by doing due diligence before publishing.
Click to expand...
They are businesses here to make money and NOTHING MORE. They will act in the interest of making money over providing accurate information every time.
 
  • Like
Reactions: Tlh97, Saylick, yuri69 and 1 other person
Josh128

Josh128

Golden Member
Oct 14, 2022
1,050
1,583
106
He actually claimed they are targeting MORE than 7 GHz. Beware the hype train, THIS IS the new ">40% IPC". There is no indication that even N2X will allow for such speeds, speeds that are not just +5% or +10%, but over 20% of where N4 and N3 (note N3 had ZERO max freq gains over N5/N4) were.

I'll repeat. Expect more than 6.0 to 6.2 GHz, expect to be disappointed.
 
  • Like
Reactions: Tlh97, Saylick, Thunder 57 and 5 others
Markfw

Markfw

Moderator Emeritus, Elite Member
May 16, 2002
27,094
16,014
136
Josh128 said:
He actually claimed they are targeting MORE than 7 GHz. Beware the hype train, THIS IS the new ">40% IPC". There is no indication that even N2X will allow for such speeds, speeds that are not just +5% or +10%, but over 20% of where N4 and N3 (note N3 had ZERO max freq gains over N5/N4) were.

I'll repeat. Expect more than 6.0 to 6.2 GHz, expect to be disappointed.
Click to expand...
I will add that the ~40% for Zen5 was true for pure heavy AVX-512 only, but it had some truth to it.
 
5

511

Platinum Member
Jul 12, 2024
2,864
2,867
106
  • Like
Reactions: Tlh97, Makaveli, Josh128 and 1 other person
Io Magnesso

Io Magnesso

Senior member
Jun 12, 2025
487
135
71
Markfw said:
I will add that the ~40% for Zen5 was true for pure heavy AVX-512 only, but it had some truth to it.
Click to expand...
However, the case is too local
Although the performance improvements were quite good for server workloads…
Performance improvements are not visible in the consumer environment…
I wouldn't say it's not there, but...
 
  • Like
Reactions: Tlh97 and derfred
Josh128

Josh128

Golden Member
Oct 14, 2022
1,050
1,583
106
511 said:
Yeah but that's not general IPC also browsing SPEC CPU seems like the 16% figure from AMD was spot on I always wonder why they put game into IPC Improvements
Click to expand...

They almost certainly derive their final target figure from SPEC, then intentionally show a range of workloads that when geomean averaged, hit exactly that number. Its actually a decent and truthful way to do it, IMO.

As far as AVX-512 40%, thats some very few, very specific workloads, nothing even close to approximating IPC gains average over a wide swath of workloads. Its so specific that you could probably find such specific use case increases every gen from Zen 1 to Zen 5 if you really tried. Remember CPU-z had to be reworked because there was a specific instruction that Zen 1 was blistering fast at and thus made its results better than Intels, despite losing in most other workloads? It happens. Its fun, but we know thats not a valid average IPC claim.
 
Last edited:
  • Like
Reactions: Io Magnesso
Markfw

Markfw

Moderator Emeritus, Elite Member
May 16, 2002
27,094
16,014
136
Josh128 said:
They almost certainly derive their final target figure from SPEC, then intentionally show a range of workloads that when geomean averaged, hit exactly that number. Its actually a decent and truthful way to do it, IMO.

As far as AVX-512 40%, thats some very few, very specific workloads, nothing even close to approximating IPC gains average over a wide swath of workloads. Its so specific that you could probably find such specific use case increases every gen from Zen 1 to Zen 5 if you really tried. Remember CPU-z had to be reworked because there was a specific instruction that Zen 1 was blistering fast at and thus made its results better than Intels, despite losing in most other workloads? It happens. Its fun, but we know thats not a valid average IPC claim.
Click to expand...
In primegrid, it really was close to 40%. Most of primegrid is VERY heavy on avx-512. Thats why I said it was not a lie, but very specific. There are not even any primegrid benchmarks that I know of.

I was only stating it was not a total lie.
 
Last edited:
  • Like
Reactions: Io Magnesso
Josh128

Josh128

Golden Member
Oct 14, 2022
1,050
1,583
106
Markfw said:
In primegrid, it really was close to 40%. Most of primegrid is VERY heary on avx-512. Thats why I said it was not a lie, but very specific. There are not even any primegrid benchmarks that I know of.

I was only stating it was not a total lie.
Click to expand...
It was a lie in regards to general IPC gains, which was the context it was claimed for by the click bait "leakers".
 
  • Like
Reactions: Io Magnesso and okoroezenwa
I

igor_kavinski

Lifer
Jul 27, 2020
25,994
17,937
146
Joe NYC said:
https://twitter.com/x/status/1943928924704047123
Click to expand...
Seriously that guy is so annoying I think that forum policy should be implemented to put his rumors in spoiler tags clearly mentioning that it's MLID crap.

Did they hire P4 die sniffers or something? Don't they know that only Intel should be doing something crazy like that? Are we going to get special PC cases with a special compartment for the CPU for better airflow AGAIN?

Can anyone imagine the hotspots that thing could create at those speeds?
 
  • Like
Reactions: Makaveli, Tlh97, Racan and 3 others
I

igor_kavinski

Lifer
Jul 27, 2020
25,994
17,937
146
511 said:
N2 is denser than Zen4 as well so you would require extreme cooling to manage that
Click to expand...
Yeah unless AMD developed some sort of nanochannels cooler inside the heatspreader...

I just would hate something like the 14900KS that can only realize its full potential with delidding and custom cooling. Not every enthusiast wants to worry about possible leaks.
 
J

Joe NYC

Diamond Member
Jun 26, 2021
3,238
4,736
136
igor_kavinski said:
Seriously that guy is so annoying I think that forum policy should be implemented to put his rumors in spoiler tags clearly mentioning that it's MLID crap.

Did they hire P4 die sniffers or something? Don't they know that only Intel should be doing something crazy like that? Are we going to get special PC cases with a special compartment for the CPU for better airflow AGAIN?

Can anyone imagine the hotspots that thing could create at those speeds?
Click to expand...

In the same video, he didn't say AMD is expecting extra ordinary cooling challenges.

Whatever the clock is, we are all talking here about single core boost or couple of cores, not sustained 12 / 24 cores 100% loaded with AVS-512 code.
 
J

Joe NYC

Diamond Member
Jun 26, 2021
3,238
4,736
136
igor_kavinski said:
In the interest of preventing possible negative drama at Zen 6 launch, AMD should release a statement discrediting MLID by name. He's already done more than enough damage by suggesting 7 GHz. If Zen 6 doesn't hit that frequency, people will go, oh well, it's AMD. Their CPUs are never as they are hyped up to.
Click to expand...

You have to keep in mind that AMD leaks to MLID could be AMD's psychological warfare against Intel.

If there is no chance in hell that Intel can compete with even super expensive N2 Nova Lake with bLLC, that will also drain resources, why wouldn't LBT just cancel is? And instead concentrate on in-house 18A produced lower performance dies?
 
You must log in or register to reply here.

TRENDING THREADS

COMPANY
RESOURCES
FOLLOW
Top Bottom