They recently said 1.7x the perf AND 1.7x the perf/watt, this mean 1.7x the perf at same TDP, if TDP is increased then it should be more than 1.7x the perf.
They recently said 1.7x the perf AND 1.7x the perf/watt, this mean 1.7x the perf at same TDP, if TDP is increased then it should be more than 1.7x the perf.
Yes.
Page 298 - Question - Zen 6 Speculation Thread
Page 298 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
forums.anandtech.com
adroc_thurston
Diamond Member
- Jul 2, 2023
- 7,563
- 10,333
- 106
Meteor Late
Senior member
- Dec 15, 2023
- 341
- 379
- 96
People always operate under Intel's world for some reason, remember how people were talking about the 5GHz barrier, as if it was some magical number that wasn't possible to surpass or barely surpass.
Let's be honest, the main reason is that Intel was the leader in foundry and also in performance for so long and so we had to operate under the logic that the walls Intel were facing were ultimate walls not possible to overcome.
15% or close to that should be expected under the TSMC and AMD world.
Let's be honest, the main reason is that Intel was the leader in foundry and also in performance for so long and so we had to operate under the logic that the walls Intel were facing were ultimate walls not possible to overcome.
15% or close to that should be expected under the TSMC and AMD world.
adroc_thurston
Diamond Member
- Jul 2, 2023
- 7,563
- 10,333
- 106
DT gets a far better deal (50 > 33 % cc bump and 2 nodes worth of oomph).
That CCD is an nT monster.
adroc_thurston
Diamond Member
- Jul 2, 2023
- 7,563
- 10,333
- 106
You are correct 4.4 was the clock for SDE and i was confusing it with M5
adroc_thurston
Diamond Member
- Jul 2, 2023
- 7,563
- 10,333
- 106
That would increase perf to 1.85x but this would also reduce perf/watt improvement to only 1.55x or so, the 1.7x better perf is at isopower otherwise they couldnt claim 1.7x better perf/watt.
adroc_thurston
Diamond Member
- Jul 2, 2023
- 7,563
- 10,333
- 106
Uh, no.
And it's 20% socket power to accommodate 33% more mem channels and other stuff. It ain't free.
That contradict their slide, they say 1.7x the perf and 1.7x the efficiency, that means 1.7x the perf at same power, the IOD power is included here, if they also use a better process for this part then the 1.33x more RAM channels wont increase the IOD power, also there s not only the memory controlers that sink power in the IOD.
adroc_thurston
Diamond Member
- Jul 2, 2023
- 7,563
- 10,333
- 106
No it just means whatever the peak socket perf at 600W equals to a 1.7x PPW advantage for Venice (relative to a 500W 9965).
It's still 16 channels at much higher signaling rates.
They say 1.7x the perf and 1.7x the efficency, i dont know why you are insisting on the 20% higher power since that contradict those numbers, if they had increased the power by 20% and still retained the 1.7x efficiency then it would mean this latter number along with 1.7 x 1.2 = 2.08x the perf.
adroc_thurston
Diamond Member
- Jul 2, 2023
- 7,563
- 10,333
- 106
It says >1.7x and AMD always measures on sockets at rated power.
I know what it is and what it isn't.
That s a hollow statement, as if we didnt know that a stated Epyc power is measured
at socket level, you are reinventing the wheel, wake up, those CPUs have an officialy rated TDP.
Apparently you have difficulties reading a slide, be it from AMD, i wont get further in this debate, numbers published officialy are supposed to be accurate while you are just doing speculations out of the wild.
This slide explicitely state 1.7x the perf at 1.7x the efficency, methink that you dont understand the notion of efficency of a CPU, namely that it decrease close to linearly with augmented perf as long as you re not too high for frequency.
adroc_thurston
Diamond Member
- Jul 2, 2023
- 7,563
- 10,333
- 106
It's a 500W vs 600W socket joust.
It's vaguepoasted slideware and you should stop.
It's >1.7x perf stuff so vaguepoasting.
You should stop.
With 1.7x better efficiency this would mean 1.85x the perf, period.
Lol, the pot that call the kettle, it s rather you who should get out of basic arithmetics irrationality, sure that if 1.7 x 1.2 = 1.7 was to be right then everything would be possible,
but for sure that i have more confidence on AMD that on someone who s left downplaying their numbers because of his ego and unability to accept their numbers as being way closer to reality than his random speculations.
Edit : it s 1.7x perf at 1.7 efficency, if someone want more perfs he can use 20% more power to get 1.85x the perf but he will have to be content with only 1.55x better efficiency.
Last edited:
adroc_thurston
Diamond Member
- Jul 2, 2023
- 7,563
- 10,333
- 106
It says >1.7x perf.
It can be whatever.
Meds please.
Venice SP7 is a 600W socket with 600W 256c parts.
It say 1.7x the perf along with 1.7x better efficiency, the whatever is just an invention of yours.
I edited my post but i will point it again, it s 1.7x the perf at 1.7x the efficiency, AMD made the comparison at isopower, if power is increased by 20% you ll get 1.85x the perfs but efficiency will be only 1.55x better.
adroc_thurston
Diamond Member
- Jul 2, 2023
- 7,563
- 10,333
- 106
>1.7x. it can be 2.2x perf for all you care.
They bench sockets at rated power, which is 500W vs 600W here.
600W 256c Venice has >1.7x PPW delta relative to 500W 192c 9965.
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 25K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes + WCL Discussion Threads
- Started by Tigerick
- Replies: 23K
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 23K
-
-
Facebook
Twitter