- Mar 3, 2017
- 1,747
- 6,598
- 136
With the experience they now have with V-Cache, I wonder if it might have been beneficial to launch X3D from the beginning. Combine some increased engineering resources with a 2 to 3 month delay in launch. Especially if they are able to make another cut in the max frequency gap.I think most would be ok with this if they add a lower end X3D, like a 9600X3D or 9700X3D, ala 5000 series.
That power consumption gain is largely due to switching from N5 to N4P. What design or engineering performance improvements exist beyond the benefits of TSMC silicon?First Im not cherry picking anything, thats a (bad) assumption.
I literally said - "Thats the only thing I have had time to look at" and you still feel the need to insult.
Second 5% more performance for much less power consumption still supports my original statement of more performance for less power.
Based on the reviews, AMD may have to ditch the X3D v-cache and just make it the standard on all Zen 5 chips. They need to be able to push 3D v-cache to the maximum clock limits, identical to standard Zen 5 max clocks.With the experience they now have with V-Cache, I wonder if it might have been beneficial to launch X3D from the beginning. Combine some increased engineering resources with a 2 to 3 month delay in launch. Especially if they are able to make another cut in the max frequency gap.
I’m also very curious to see the the suite of benchmark results with 16 cores on a 256-bit bus (Halo.)
Based on the reviews, AMD may have to ditch the X3D v-cache and just make it the standard on all Zen 5 chips. They need to be able to push 3D v-cache to the maximum clock limits, identical to standard Zen 5 max clocks.
lolThat power consumption gain is largely due to switching from N5 to N4P. What design or engineering performance improvements exist beyond the benefits of TSMC silicon?
So they will rely on X3D variants and XT versions alone until 2027? Or should we hope for a "Zen 5.5" with N3?But they are
There's only a mobile refresh AFAIK.So they will rely on X3D variants and XT versions alone until 2027? Or should we hope for a "Zen 5.5" with N3?
You do realize Apple is a different architecture entirely? Nobody compares X86 architecture to an entirely different platform. I am comparing Zen 4 to Zen 5 efficiency. When Zen was originally released in 2017, they had it mapped out all the way to Zen 4. Everything beyond Zen 4 is uncharted territory for AMD.lol
You do realise that Apple chips exist? They use the same TSMC nodes as everyone else, but M2 on N5 is massively more efficient in 1T compared to Zen 4 on N5, as well as their ARM competition.
lol
You do realise that Apple chips exist? They use the same TSMC nodes as everyone else, but M2 on N5 is massively more efficient in 1T compared to Zen 4 on N5, as well as their ARM competition.
If you want 1) efficiency andAt first I was a bit disappointed but after checking more reviews (especially the ones with PBO and EXPO tuning) I think the Zen 5 is a nice gen on gen upgrade. It is not revolutionary (on desktop at least), but it brings ECO mode Ryzens to the regular SKU list. I think that 9950X will top all of the charts and will be a tough matchup for 285K ARL part.
X3D models, if they are allowed to boost better than 7800X3D, will likely be top dogs in gaming for years to come.
If there is no efficiency improvement coming from the architecture, then how come HX 370 pulls less power 3-4W less power in 1T while being within ~100MHz compared to Phoenix, despite N4P having no power savings compared to N4? Fairy dust?You do realize Apple is a different architecture entirely? Nobody compares X86 architecture to an entirely different platform. I am comparing Zen 4 to Zen 5 efficiency. When Zen was originally released in 2017, they had it mapped out all the way to Zen 4. Everything beyond Zen 4 is uncharted territory for AMD.
TSMC's own numbers for performance uplift and efficiency (power consumption) are inline with the increased efficiency of N4P.
Edit: nvm got it.Computerbase has 5.5Ghz Zen 5 at 14% better than 7950x @ 5.7Ghz. At least in their ST mix (new suite), Zen 5 shows >16% IPC increase.
View attachment 104702
Bald Eagle PointThere's only a mobile refresh AFAIK.
But they are
If the rumour is true that Zen6 converges both desktop and laptop; then Desktop may ditch standard DDR entirely and adopt LPDDR same as mobile. LPDDR6 LPCAMM should be available in 2027.Because Zen 6 is on AM6 w/ DDR6?
He's not the only one that's said that's what's on AMD's roadmap...I'm afraid it might be true.Sorry, but I find it extremely hard to believe you have a source with that level of information, who simultaneously didn't have enough information to steer you away from your 40% IPC insanity.
If it's anything like the last time, you are maybe at best privy to some kernel of truth that you are inappropriately extrapolating from in order to reach an unlikely conclusion.
Council was talking about 1T. The Z1 Extreme is a 8C/16T part and your comparing MT to a 4P+4E/8T M2. Of course Z1 is going to score higher in MT, not to mention the disadvantage of using R23 for ARM based products.Better to look at actual numbers than doing such statements that are contradicted by measurements, 8700 pts in CB R23 at something like 20W, Z1 Extreme does 10700 pts at 20W and 8600 pts at 15W, so the latter is much more efficient, you just got the numbers backward.
To score the same as the Z1EX@20W the M2 would require roughly 32W.
From your chart. N4P provides 22% better power efficiency over N5 and 11% performance uplift over N5. You are citing N4 vs N4P in your response which is not accurate. Zen 4 was based on N5 silicon. How AMD utilizes that performance boost is up to them. If they add a AGESA bios update with 65w/95w/105w/120w, we would see significant differences/improvements in performance for the 9700x. The 7700x was originally 105w TDP vs 65w for the 9700x.If there is no efficiency improvement coming from the architecture, then how come HX 370 pulls less power 3-4W less power in 1T while being within ~100MHz compared to Phoenix, despite N4P having no power savings compared to N4? Fairy dust?
View attachment 104724
View attachment 104725
View attachment 104726
It's better to say single thread rather than 1T. It will confuse people with 1t memory timings on DDR5 memory.Council was talking about 1T. The Z1 Extreme is a 8C/16T part and your comparing MT to a 4P+4E/8T M2. Of course Z1 is going to score higher in MT, not to mention the disadvantage of using R23 for ARM based products.
—-
Anyway better in the meantime AMD, starts work on custom ARM silicon. Get a seperate team and kill Qualcomm push before Nvidia makes a move.
From your chart. N4P provides 22% better power efficiency over N5 and 11% performance uplift over N5. You are citing N4 vs N4P in your response which is not accurate. Zen 4 was based on N5 silicon. How AMD utilizes that performance boost is up to them. If they add a AGESA bios update with 65w/95w/105w/120w, we would see significant differences/improvements in performance for the 9700x. The 7700x was originally 105w TDP vs 65w for the 9700x.
Very much hoping it's not and I have to agree with the others that said that AMD may as well give up if they take that long.
Wrong. First, it provides either 22% power reduction OR 11% perf. Not at the same time. If you ever saw a single node comparison chart, you'd know that.From your chart. N4P provides 22% better power efficiency over N5 and 11% performance uplift over N5. You are citing N4 vs N4P in your response which is not accurate.
Wrong again, only desktop parts used N5. Phoenix and Hawk Point both use N4.Zen 4 was based on N5 silicon.
That change nothing, from the 20W Z1EX MT score we can deduct that it use 2W per core for 1020 pts per core excluding SMT.Council was talking about 1T. The Z1 Extreme is a 8C/16T part and your comparing MT to a 4P+4E/8T M2. Of course Z1 is going to score higher in MT, not to mention the disadvantage of using R23 for ARM based products.
—-
Anyway better in the meantime AMD, starts work on custom ARM silicon. Get a seperate team and kill Qualcomm push before Nvidia makes a move.