- Mar 3, 2017
- 1,747
- 6,598
- 136
Yes, on some workloads. But the base clock is the frequency that is guaranteed to be sustained on all workloads (i.e. the worst case which outputs max heat too), assuming cooling as specified by the TDP is used.No, they will definitely clock above that if the workload allows even if you set PPT to 65W too.
Sure but it doesn't apply to almost anything. Unless you're handwriting HPC code. So the 'effective' clock rate of the 7700 is far closer to the 7700X than the base clocks suggest.Yes, on some workloads. But the base clock is the frequency that is guaranteed to be sustained on all workloads (i.e. the worst case which outputs max heat too).
Zen 5 is the first major core designed from Ryzen/EPYC money. Its development took a while. It features rather notable changes to the core. Yet, the IPC gain is not really notably higher than that of its predecessors'.
Sure. But that applies for both 9700X and 7700X.Sure but it doesn't apply to almost anything. Unless you're handwriting HPC code. So the 'effective' clock rates of the 7700 are far closer to the 7700X than the base clocks suggest.
Nope, base clocks are the same as last generation 7700 at the same power limit despite adding double vector bandwidth to make the worst case much, much worse.Sure. But that applies for both 9700X and 7700X.
So the original point I made should still stand, i.e. the reason 9700X (65W) has lower TDP than 7700X (105W) is because the former has lower base clock. 3.8 GHz vs 4.5 GHz.
Nope, base clocks are the same as last generation despite adding double vector bandwidth to make the worst case much much worse.
I amended my post to make it clear for the contextually inept.
Says 3.8 GHz base clock for 9700X, and 4.5 GHz for 7700X.
Yeah, it sounds in the same vein as "There is not enough space in the BIOS for Zen3 support on 3xx/4xx boards, here buy a new x570 board", which turned into a PR disaster for them later.I'm curious as to how well Zen 5 will scale with increasing power limits. Knowing PR/corporate speak, I suspect there are other reasons.
Lower base clocks are mostly due to 2x 512-bit FMA.Sure. But that applies for both 9700X and 7700X.
So the original point I made should still stand, i.e. the (main) reason 9700X (65W) has lower TDP than 7700X (105W) is because the former has lower base clock. 3.8 GHz vs 4.5 GHz.
It is what the TDP is based on, so it matters.But we all know base clocks don't matter.
Nope.It is what the TDP is based on, so it matters
It's arbitrary.Otherwise, why do you think AMD specified lower base clock for 9700X than for 7700X?
Then why specify 3.8 GHz base clock if it is capable of 4.5 GHz or whatever? Are you suggesting they are using a random number generator to select it?It's arbitrary.
TDP is about thermals. Simply put the reference cooler on a CPU in a worst case scenario and this will be the highest power consumed/dissipated.It is what the TDP is based on, so it matters.
Yes, that could be an explanation.Lower base clocks are mostly due to 2x 512-bit FMA.
I'd be very surprised if the clocks were any lower compared to 7700X in general use.
Unless you're running lots of 512 bit avx code this whole conversation is moot.Yes, that could be an explanation.
Looking at 9950X vs 7950X, the difference in base clock is not as great. 4.3 GHz vs 4.5 GHz. Same TDP in that case though.
So are we saying that the difference between base clock for worst case vs base clock in general use is greater with Zen5, due to 2x 512-bit FMA on Zen5 (assuming this is the worst case, with max heat output)? In that case, should we also expect the base clock for general use to actually be higher on 9950X than for 7950X, despite that the base clock according to the tech specs is lower on the former?
Either it's a binning issue, or, if bins are real good, just market segmentation.Then why specify 3.8 GHz base clock if it is capable of 4.5 GHz or whatever? Are you suggesting they are using a random number generator to select it?
No. You missed the whole point of the last paragraph in my post that you quoted.Unless you're running lots of 512 bit avx code this whole conversation is moot.
Need the mythical +40% IPC in my veins to hit F5 and get new posts here 40% faster talking about IPC increase
Why can't amd cpu juice 250watts to get astronomical performance over struggling 250watt intel equivalent
total benchmark domination
total market domination
servers.
2 years for 16%? It's very very mediocre. This is the lowest uplift in zen history (Zen+ doesn't count), with the highest amount of budget too.EDIT: Find me one vendor that routinely drops 15+% IPC or performance increases for CPUs without driving power consumption through the roof. I'll wait.
7950X base clocks 4.5ghz at 170WNope, base clocks are the same as last generation 7700 at the same power limit despite adding double vector bandwidth to make the worst case much, much worse.
But we all know base clocks don't matter. And the boost clock is 200MHz higher at the same TDP.
No. You missed the whole point of the last paragraph in my post that you quoted.
I.e. that 9950X might actually have higher base clock in general use than 7950X, despite that the base clock according to the tech specs is lower on the former. This due to that 2x 512-bit FMA (assumed worst case for heat and clocks) is pulling down the base clock according to the tech specs on 9950X.