If it uses 35 watts, then all one would have to do is change my math to add +5 watts to the IOD of the power consumption and re-look at the graph lol. It doesn't change much.The IOD take 35W in those Blender runs that s displayed on the pics, at 80W this amount to 2.81W/core, if a 7950X IOD use 20W then at 80W this will amount to 3.75W left per core.
You can see that the footprint of the IOD can result at 80W to roughly 33% more power per core for the 7950X, that s basically what you are doing, that is comparing two cores at 33% power difference and then doing wrong conclusions out of this flawed basis
No it won't be. We have no idea how anything scales. A wider core generally has higher perf/watt at the higher end of the curve than the lower end vs a smaller core. That's not anything new....Yes but the scaling will be the same for both Zen 4 and 5, why should those elements scale differently with Zen 5.?
And to take account to the fact that the 7950X IOD can take 15W less, wich is 1W/core, and wich you did completely ignore in your random estimation.If it uses 35 watts, then all one would have to do is change my math to add +5 watts to the IOD of the power consumption and re-look at the graph lol.
Also, I just want to point out, even if we do claim that Zen 5 uses 35 watts on it's IOD, and Zen 4 only uses 30, it won't change the results much either.If it uses 35 watts, then all one would have to do is change my math to add +5 watts to the IOD of the power consumption and re-look at the graph lol. It doesn't change much.
Considering Zen 4 and Zen5 use the exact same IOD, it is unlikely the power difference between the two parts are very large at all.
And as I said in my original, comment... assuming 30 watts per core. I didn't actually base the 30 watts on any specific test, but just generally what I've heard of Zen 4 IOD power consumption and idle behavior.
I'm not comparing the cores with different IOD power consumptions at all.
I used the 30 watt figure for both Zen 5 and Zen 4. If it uses 35 watts, just use 35 watts for both Zen 4 and Zen 5.
No it won't be. We have no idea how anything scales. A wider core generally has higher perf/watt at the higher end of the curve than the lower end vs a smaller core. That's not anything new....
Also, I just want to point out, even if we do claim that Zen 5 uses 35 watts on it's IOD, and Zen 4 only uses 30,
Sure, if you want to reduce the amount of power the IOD takes, then realistically the situation for Zen 5 vs Zen 4's perf/watt gains become even worse. The perf/watt advantage in the graph is lower the lower your total power goes, meaning that lowering IOD power draw in that calculation means that your perf/watt advantage shrinks as well.And to take account to the fact that the 7950X IOD can take 15W less, wich is 1W/core, and wich you did completely ignore in your random estimation.
Nothing here is a trick lol. I've been extremely open with my math. It's not complicated.A graph in isolation means nothing, what should be displayed is the power/core of both Zen 5 and 4, rest is number trickings of yours using incomplete datas where Zen 5 power is oversestimated and Zen 4 one underestimated.
The Zen 4 IOD is the exact same IOD as Zen 5 lol. If there were any large differences, it would be due to it being an ES sample, which would be covered in the *take with a grain of salt* section.Zen 4 IOD use 20W, see my previous post rather than engaging in even more senseless
assumptionSS forever based on wrong numbers.
Dunno about exact same unless we can see them side by side. I think Zen 5's IOD may be more refined where needed to support the newer cores.The Zen 4 IOD is the exact same IOD as Zen 5 lol.
It might be another stepping or something, but I don't think there's any rumors or info claiming Zen 5's IOD will be much better.Dunno about exact same unless we can see them side by side. I think Zen 5's IOD may be more refined where needed to support the newer cores.
In regards to that power/performance curve, keep in mind the 9950X ran memory @ 2:1 8000MT/s (?) and FCLK @ 2133mhz.. Which means it was pretty much tuned for the high end from the get-goView attachment 103000
This graph is not exactly super optimistic for Zen 5 perf/watt...
Edit: sorry, I should give credit... I forgot to link the post I got it from. Here it is.
I didn't know the fabric clock was OC'd, that's fair.In regards to that power/performance curve, keep in mind the 9950X ran memory @ 2:1 8000MT/s (?) and FCLK @ 2133mhz.. Which means it was pretty much tuned for the high end from the get-go
While the 7950X ran at 1:1 6000MT/s and FCLK @ 2000mhz
I'm sure that also very much affected how much power the cores got at lower PPT targets.
So lets take a closer look at the 60w PPT target
9950X @ 60w PPT = 25w core pow
View attachment 103024
7950X @ 60w PPT = 30w core power
View attachment 103025
30/25 = 20% difference
There is your effiency difference at the lower end of the power scale
my2cent
In most cases wouldn’t the real life differences be most significant in the non 3D models? And besides benchmarking and fun and giggles, be less than 5% for the 3D models in real life scenarios?Will copy my post over from from a other forum as a response:
So over the last few weeks ive been working on a memory performance comparison between the following maxed out daily memory profiles
The CPU used for this purpose is my newly acquired SP99 7800X3D @ daily settings, running on the ASUS GENE motherboard
- Profile1 = SR 2x16gigs adie @ 6600MT/s CL26-37-32-30-62 + 2200mhz FCLK 1:1 mode
- Profile2 = SR 2x16gigs adie @ 8000MT/s CL32-45-40-44-84 + 2200mhz FCLK 2:1 mode
- Profile3 = DR 2x32gigs adie @ 6600MT/s CL28-38-36-36-72 + 2200mhz FCLK 1:1 mode
- Profile4 = DR 2x32gigs adie @ 8000MT/s CL34-46-44-60-104 + 2200mhz FCLK 2:1 mode
View attachment 102953View attachment 102954
My main performance metric for this comparison have been Clam cache/memory benchmark and/or Karhu ramtest, but i have also included AIDA64 and hwinfo in my screenshots as i know people in this thread mostly like to look at them. My criteria for being a fully stable daily memory profile and being added to this performance comparison is being able to survive atleast 6 hours in karhu and over 1 hours Y-cruncher all memtests only.
With all that out of the way, we can start looking at some numbers
SR 2x16gigs adie @ 6600MT/s CL26-37-32-30-62 + 2200mhz FCLK 1:1 mode
View attachment 102955View attachment 102956
SR 2x16gigs adie @ 8000MT/s CL32-45-40-44-84 + 2200mhz FCLK 2:1 mode
View attachment 102957View attachment 102958
DR 2x32gigs adie @ 6600MT/s CL28-38-36-36-72 + 2200mhz FCLK 1:1 mode
View attachment 102959View attachment 102960
DR 2x32gigs adie @ 8000MT/s CL34-46-44-60-104 + 2200mhz FCLK 2:1 mode
View attachment 102965View attachment 102966
Results in Clam cache/mem benchmark:
Latency ranking:
- SR 2x16gigs @ 6600MT/s 1:1 mode= 68.75ns
- DR 2x32gigs @ 6600MT/s 1:1 mode =70.17ns
- SR 2x16gigs @ 8000MT/s 2:1 mode = 70.24ns
- DR 2x32gigs @ 8000MT/s 2:1 mode = 71.84ns
Bandwidth read-modify-write (ADD) ranking:
A few comments in random order to my findings above
- SR 2x16gigs @ 8000MT/s 2:1 mode= 97.11GB/s
- DR 2x32gigs @ 8000MT/s 2:1 mode = 92.87GB/s
- SR 2x16gigs @ 6600MT/s 1:1 mode = 91.23GB/s
- DR 2x32gigs @ 6600MT/s 1:1 mode = 87.34GB/s
A single 8core Zen4 CCD can take advantage of the higher bandwidth afforded by 2:1 mode vs 1:1 mode, even if the common misconception on many forums is that there is no benefit because they can hardly see any difference in gimmicky AIDA64 memory bench. (its also easy to double check this in other benchmarks such as y-cruncher / GB3 membench which will show the same)
The next question would naturally be what's the "best memory setup", 1:1 mode with its lower latency or 2:1 with its higher bandwidth. There is no easy answer for this as it all depends on what benchmark/game you comparing the numbers in.. Some will prefer latency while others bandwidth, so you just have to check on an individual basis.
But what i can say is that i pretty much always think higher memoryspeed is better, be it in 1:1 mode or 2:1 mode... From time to time i see some limit themself to something like 6000/6200MT/s because they think its faster in games than say 6400MT/s for some reason (?)
My next observation is that i did not find any bandwidth benefit from the "dual rank" (quad) in Clam cache/mem benchmark, but karhu is seemingly showing higher mb/s. But i suspect this is because the higher memory size tested, not increased bandwidth from DR. I will do some more DR karhu runs where i limit used memorysize to same as SR and check if the numbers change. (y) edit Its also possible the forced GDM enabled with DR is eating up the bandwidth benefit compared to SR
Have also seen some complains about some ppl having a hardtime tuning memory on the 1.1.7.0 PatchA FireRangeP AGESA, i can only say that is working pretty good for me on the ASUS GENE, even if i'm using a beta bios. But be warned, stabilizing DR 64gigs @ 8000MT/s is still insanely hard, think i spent like 5x the time on this profile compared to all others combined... Its really on a razors edge, +-5 mv on some rails and you can forget about 10k karhu.
I will say, if the difference really is still only ~5 watts between the two IODs, the picture doesn't change that much overall.I didn't know the fabric clock was OC'd, that's fair.
The Zen 4 IOD is the exact same IOD as Zen 5 lol. If there were any large differences, it
Someone literally just posted a picture of the Zen 5 IOD consuming 35 watts (as it was tuned and OC'd) and the Zen 4 one using 30 watts.It s the same and it use up to 21W only when RAM is set at 6000, so it shouldnt take 35W with the 9950X, you just aknowledged that you are cluless about the IOD actual TDP.
I ve no time to waste if you are unaware of the numbers, here are the numbers, do your homework and get here later...
13.7-16.3W at stock and depending of the MB and 19.4-21.3W with the same MB and RAM at 6000.
AMD Ryzen 7000 im Test: So schnell sind 7950X und 7700X: Leistungsaufnahme und Effizienz
Ryzen 9 7950X & Ryzen 7 7700X im Test: Leistungsaufnahme und Effizienz / TDP und PPT steigen deutlichwww.computerbase.de
Someone literally just posted a picture of the Zen 5 IOD consuming 35 watts (as it was tuned and OC'd) and the Zen 4 one using 30 watts.
Maybe we're talking around each other (?)I will say, if the difference really is still only ~5 watts between the two IODs, the picture doesn't change that much overall.
After all, 16 x 3 + 35 (for Zen 5) is still 83 watts, while 16 x 3 + 30 = 78 watts.
And the perf difference between the two samples, at those power figures are still not that large.
I've already redone the math lmao... 34 minutes agoThese are the numbers that will be measured for the IOD in the reviews.
Page 626 - Discussion - Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)
Page 626 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.forums.anandtech.com
In power needed sure, but that's not perf/watt, which equalizes wattage and looks at perf improvements.Maybe we're talking around each other (?)
At the 60w power target: (and please forget about the IOD for a second)
9950X had 25w divided over its 16 cores = 1.5625w per core
7950X had 30w divided over its 16 cores = 1.8750w per core
Thats a 20% difference
Its sad you had to point this out multiple times and the guy still argues. I thought thats common knowladge after 5sec of looking at the screenshots.30/25 = 20% difference
There is your effiency difference at the lower end of the power scale
Ok i guess i know why u argued. But idk why you are still arguingI didn't know the fabric clock was OC'd, that's fair.
In power needed sure, but that's not perf/watt, which equalizes wattage and looks at perf improvements.
efficiency is generally refered to iso perf, perf/watt is iso watts.
Because the 5 watt difference between the IODs doesn't change the perf/watt match much at all.Ok i guess i know why u argued. But idk why you are still arguing
Equalizing something iso power vs equalizing something iso performance are two very different things.Lol, efficency and perf/watt are exactly the same thing, seems that you are quite confused.
To summarize if Zen 5 is 15% more efficent at 7.5W than Zen 4 then it will be also 15% more efficent at 3W, , or else the laws of physics would mean nothing.
Power scaling is surely not uniform over frontend logic, execution logic, the various buffers and caches, fabrics, and so on, is it?
On the one hand, the power scaling characteristics of, for example, one ALU in Zen 5 is not going to be exactly the same as of one ALU in Zen 4, because the physical designs and the process nodes are not exactly the same.Yes but the scaling will be the same for both Zen 4 and 5, why should those elements scale differently with Zen 5.?
The IOD is the same, only the software has changedDunno about exact same unless we can see them side by side. I think Zen 5's IOD may be more refined where needed to support the newer cores.
There are quite a few settings in the BIOS that can change this number completely. Even the physical hardware can change how much power the IOD consumes. It is also irrelevant beyond client, since the IODs aren’t used beyond non AM* platforms.Someone literally just posted a picture of the Zen 5 IOD consuming 35 watts (as it was tuned and OC'd) and the Zen 4 one using 30 watts.
True. I assumed both of them were at stock.The IOD is the same, only the software has changed
There are quite a few settings in the BIOS that can change this number completely. Even the physical hardware can change how much power the IOD consumes. It is also irrelevant beyond client, since the IODs aren’t used beyond non AM* platforms.
Mine is using 21W idling at the desktop right now, but if I push the machine it jumps up to 27W.
On the one hand, the power scaling characteristics of, for example, one ALU in Zen 5 is not going to be exactly the same as of one ALU in Zen 4, because the physical designs and the process nodes are not exactly the same.
On the other hand, Zen 5's architecture (thus, balance of frontend...backend, various widths and buffer sizes, frontend features, and more) differs. How a particular workload is going to cause two different architectures to expend fractions of the overall task energy at the front/ at the back etc. won't be the same.