120W TDPThats a really high base freq compared to the rest of Zen5 lineup 🙃
Wait, what?? Is that legit??Thats a really high base freq compared to the rest of Zen5 lineup 🙃
Yeah, Intel looks bad there.
Not really, just a different appproach.
AMD is 26% more performance at 27% more power.
Intel is 18% more perfomance at 3% more power.
Overall it's just different, I would even argue that Intels approach is better, because it's basically free performance at same power.
The figures from David Huang's tweet are single-core results (single-thread vs. dual-thread runs on same core) at fixed clock speed,¹ relative to Vcore VRM power.It's weird because other tests (i.e. Phoronix Epyc tests) show that enabling SMT has little to no impact on power draw but generally increases performance.
So I wonder what can be extrapolated from either set of tests.
Of the CPU makers who put two different core microarchitectures into one and the same CPU, Intel was the only one who additionally put HyperThreading into one type of these cores. As others already pointed out, how is an operating system's thread scheduler supposed to handle this? I don't know though if, or how much, this played a role in Intel's decision to remove HT in upcoming heterogeneous client CPUs (in which wide computing parallelism is not the top concern, Cinebench marketing figures aside). BTW, in another segment, small server CPUs = Xeon E, Intel did something else and disabled heterogeneity instead.[....] it's basically free performance at same power. Which makes it even more unbelievable that they removed it for new consumer CPUs.
Well, at least the headline on the slide is "Lion Cove", not "Arrow Lake".[...] Intel says "Optimized for PPA" on their slide for the Arrow Lake P-cores:
View attachment 109751
This. Doesn't. Make. Sense.
That would be hypothetical 9950x3D with 2 v-cache dies9800X3D looking to be the first X3D that will be good at everything, if not stellar at everything
Guys, post your best results.Folks, an assignment for you and the sooner you complete it, the sooner I can share a juicy screenshot
View attachment 109769
Screenshot to be posted after sharing of 10 results at above setting in FF Dawntrail benchmark
Intel is using superior silicon on Arrow Lake compared to their own Intel fab. That is where the power efficiency for Intel comes from.
You have picked wrong game ... but I will run it on 9950X + 7900XTXGuys, post your best results.
Please.
I'm dying to share a little something...
TSX? Is there some interesting story out there?TSX in Zen3.
Thanks!Hah! My performance is Extremely High!! Eat that console peasants 😅
9950X 102FCLK CO1 -20 CO2 -30 MEM6120MHz PBO 220W GPU Stock
It doesn't make sense to me either. Essentially, Intel is going with SMP vs SMT like AMD did with Bulldozer. Furthermore, having SMT in both P and E cores provides the OS with a very simple scheduling task compared to having P cores with SMT and E cores without.The aspect of performance per power is only one of the reasons for implementing SMT. It's nice to get it without increasing power usage much, but even at linear power increase it would be desirable because that's better than the less than linear performance-per-power increase at high frequencies. The other argument for SMT is more performance per area because the additional transistors require much less area (around 5%) than they add in performance (15-30%). The source is this document, and the money quotes are:
So...
Why do I say this? Because Intel says "Optimized for PPA" on their slide for the Arrow Lake P-cores:
View attachment 109751
This. Doesn't. Make. Sense.
And sorry for this now basically being a post for the Intel thread, but we are also discussing AMD's more effective implementation of the same technology.
What makes validation easier is having only ONE CPU architecture stack to debug. This is truly a crazy argument IMO.Maybe the simplest answer is correct: removing SMT simplifies validation of the design, and makes it easier to schedule threads between the P cores and the E cores. That’s it. Anything beyond that is just marketing so that the consumer doesn’t feel like they got a downgrade.
I think the "whole frickin sea of cores" concept is where Skymont is going. This flies in the face of the PPA SMT is proven to provide. I think Intel will need to rethink this approach as they will certainly pay for it in DC for sure.I would disagree here. I think the main reasons are "lots of design decisions made around improving SMT performance", making AMD CPUs great scalable multicore processors and not wanting to spam their CPUs with smaller cores (reasons for which I'm not entirely sure of but I think they have Mont type cores undergoing development and they are not yet ready for prime time).
We have seen many examples where AMD has shown Intel how to do things right. SMT is just one of such things where they took a security first approach to implementing secure boundaries between the physical and virtual threads. Another is AMD's mitigations against Spectre/Meltdown which Phoronix showed that turning off these mitigations actually makes AMD CPUs run a bit slower whereas doing the same on Intel CPUs makes them run faster. So Intel's mitigations are doing additional checks or preventing some optimized pathways from working whereas AMD's mitigations were built into the core design itself and they figured out optimizations to steal the performance back from these mitigations and turning the mitigations off also turns off those mitigation "mitigating" optimizations.
I think if AMD gives up SMT without ever exploring SMT4, it will be when they can have a whole frickin' sea of cores in an area smaller than Intel Monts.
Whoa! Talk about a 1-2 punch! So we are going to have a ZEN5 CPU on N4P massively outpacing a newer Intel Arrow Lake processor on N3B (the most dense process TSMC currently offers using the most EUV layers.... and the most expensive process TSMC offers). That is going to sting for this cycle. Still, the more important battles will be in the laptop and DC markets, so it may not be that bad for Intel.
I keep wondering when games will be developed that are designed to operate on processors with more than 8 cores. Within a year or two, I think the bottom end CPU's will likely have 16 cores minimum.That would be hypothetical 9950x3D with 2 v-cache dies
BTW, we always talk here about how using cores on the other die is detrimental for gaming, cause of the higher latency, and thats why all the core-parking shenanigans and whatnot. Now this goes for sure for games, that use 8 cores max, in which case you dont want it randomly to use one of the second die cores, fair enough. But what about games (or apps potentially benefitting from v-cache) that can use more than 8 cores? Is that latency so big, that those cores are pretty much useless and make the experience actually worse, or do the additional cores on the second die actually provide tangible benefit? There seems to be little info on this topic, granted most games/apps dont need more than 8 cores, so thats i guess, why. Anyway, i would be interested.
I keep wondering when games will be developed that are designed to operate on processors with more than 8 cores. Within a year or two, I think the bottom end CPU's will likely have 16 cores minimum.
Maybe so in general. I still think that in the high end desktop, we can expect more cores moving foward. I am pretty sure AMD's Zen 6 will feature an e Core CCD having 32 Zen 6c cores in it.Nah, way too soon. also see Amdhal's Law. I expect 8 core CPU's to be around at the bottom end for some time.
Thanks!
Well, all I can say is, you are quite behind compared to my screenshot
I'm not that well-versed in this benchmark's scores. Would the score increase with a faster GPU? How much would your score decrease if you power limit your GPU, by say, 100W?
Rest of you rich brats, where are the scores???
Yeah, "Lion Cove" is the name of the Arrow Lake P-cores. You can read the whole slide deck here.Well, at least the headline on the slide is "Lion Cove", not "Arrow Lake".
@CouncilorIrissa answered that: the higher TDP automatically increases the base clock.Wait, what?? Is that legit??