Ah this makes more sense then. So IPC gains are lower than predicted from just object detection? Well that sucks.Per Geekbench 6 Internals, SME is used in three different subtests (Photo Library, Object Detection, and Background Blur.)
Ah this makes more sense then. So IPC gains are lower than predicted from just object detection? Well that sucks.
But yeah even then IDK, is it blowing up power? With three subtests from the GB6 test we should be able to find out just based on how high they are in a graph. At any rate, if it is, that still counts in some sense.
"dull iSheeps", huh? Relax friend. Do people who accuse others if being "iSheeps" know that they themselves are sheeps of something else? Usually "AYYMD-sheeps".Lmao. Here we go, muh PowerMetrics from senttoschool. This isn’t accurate, Geekbench is a composition of different tests, and power metrics doesn’t even measure package power anymore. It just measures “CPU power”.
You want to include the whole platform. That’s what Andrei does, that’s what people who aren’t lobotomized dull iSheep do. Your M1 Pro is drawing much more than 5W to begin with, and Apple has pushed up power.
Even an iPhone draws 4-5W for GB5/6 or Spec dude.
I am not an AMD fan and dislike the behavior of both Apple and AMD fans (you know the whole gang here) in this forum, but Apple fans are really taking the cake lately with this BS. “My Mac is doing like 3W in full ST workloads” lmao, no it isn’t."dull iSheeps", huh? Relax friend. Do people who accuse others if being "iSheeps" know that they themselves are sheeps of something else? Usually "AYYMD-sheeps".
Weird deflection. I’m fully aware they can measure GPUs and NPUs too, but we’re talking about CPU workloads, and in those (well just like others), DRAM and yes even the VRMs are important too, and PowerMetrics doesn’t measure that. The composition is important because if anything it’ll spike above 5W instantaneously, but powermetrics may not display that. It’s a composition of very bursty tests.When did I say Geekbench isn't a composition of different tests?
Powermetrics measures CPU+GPU+NPU. It's not just CPU power.
it wasn’t M1 package power, it was the full platform because it was from the wall with a Mac Mini which does not in fact have batteries to throw anything off. You’re not measuring the same thing.Yes, and Andrei measured M1 package power at 6.3w peak for average ST load. GB6 is an average ST workload.
I mean, the fact that an iPhone consumes more than your eyeballed measurements for the average of Spec or GB for the most part, and those iPhones are including DRAM and VRMs is telling. By extension your power is going go be at *least* theirs for a chip of the same generation, and likely more than the regular M1 due to your bus width.Regardless, that's not too far off from my eyeballing GB6 running between 3w-5w without rare occasions going above 5w (probably around 6.3w that Andrei measured).
Dude. It’s a new chip, and an M chip, not an A chip, and Apple puts M chips in their iPads. It doesn’t matter what you think based on device size. We’re asking what the chip draws.The iPhone is 63% thicker than the iPad. While the iPad is much bigger overall, there's no reason to think that the iPad Pro draws more than double the wattage of an iPhone 15 Pro Max. Dude.
Going off A14 results (albeit GB) looks like they’re using 62.5% than the similar generation of A series chip for the M series here. So I don’t find twice the total power in ST for A17 to M4 to be something ridiculous, laptop chips have more headroom and the iPad probably just throttles by default to a degree.
it wasn’t M1 package power, it was the full platform because it was from the wall with a Mac Mini which does not in fact have batteries to throw anything off. You’re not measuring the same thing.
Other than power supply losses and spinning a little fan what about measuring at the wall with a Mac Mini is incorrect for establishing package power? It isn't exact but it should be pretty close. At the very least it provides an upper bound - if you measure 6.3 watts at the wall you know the true figure is less than 6.3 watts.
Sure yeah.You could increase the accuracy by bypassing the power supply to feed it DC directly and measure with something like a Fluke that's designed for more accurate readings than a Kill-A-Watt device (which is known to overestimate small loads) and powering the fan separately so its power isn't counted as package power.
Andrei measured 6.3w from the wall. Dude.it wasn’t M1 package power, it was the full platform because it was from the wall with a Mac Mini which does not in fact have batteries to throw anything off. You’re not measuring the same thing.
I know? What’s your point. Apple has pushed power up with M2 and M3 both just a bit, so 11W for a 3800 GB6 is not that ridiculous. That’s still best in-class. God the cope is so weirdAndrei measured 6.3w from the wall. Dude.
They went for speed and it shows.@adroc_thurston how do you feel about the M4 and Apple’s watts?
be honest. 3330-ish.for a 3800 GB6 is not that ridiculous
I don’t agree but it’s true it’s the meta. Everyone’s pumping. Straight up Venice Beach roids with watts rn.They went for speed and it shows.
Watts (more is better) is the meta anyway.
You think 3300? Might be fair, IDK. Did they get any IPC gains outside of SME or no?be honest. 3330-ish.
SME pump is baaad, shared cluster-level accelerators are so Sun Niagara.
No, seems to be 9% moar perf for 9% moar freq. It's a speedbump basically, like Comet Lake (APPL edition).Did they get any IPC gains outside of SME or no?
Look harder.Right now it looks like Zen 5 might not pass Apple which I was previously looking forward to.
Wait until Techinsights tears the thing into bits.I don’t agree
Yeah pretty rough. This time seems to be no LITTLE update either so it's literally cometlake AAPL edition (+2c, more watts, fun!)Lol yeah it has strong “riding off the foundation the Nuvia guys built” vibes.
It's a real mean shredder and they've hidden it so well from the prying eyes of comp analysts everywhere.Even if they don’t, getting IPC close to M1 is big enough to me.
fmax chasing is fine if you're willing to trade off the area and have the physdes crews ready to go at it.I meant that I personally think it’s stupid
Yeah, lmao.Yeah pretty rough. This time seems to be no LITTLE update either so it's literally cometlake AAPL edition (+2c, more watts, fun!)
I’m looking forward to it. Computex full deets?It's a real mean shredder and they've hidden it so well from the prying eyes of comp analysts everywhere.
Yeah. I will say this. At least AMD’s strategy is non-stupid with the Big.Compact strategy on Strix. Good way to do this. Big fan of the simplicity.fmax chasing is fine if you're willing to trade off the area and have the physdes crews ready to go at it.
Still a pretty good part but the issues are showing clearer than ever.Yeah, lmao.
Yeah and as you've seen, they're throwing the HotChips party for it with Mikey C. unlike the usual.Computex full deets?
Oh it’s still on top IMO because of that efficiency and the e cores, and clocking that thing down would be revealing but Apple won’t let you do that, and the second issue is they’re still throwing down 8GB of RAM. Yuck. Fine for a tablet but not a laptop.Still a pretty good part but the issues are showing clearer than ever.
Yeah we’ll see if they even have anything in the tank really. Pegasus is more exciting.Hopefully A18 is a real new big C.
Yep, looking forward to that and the Oryon one. As you know Mike C has been waiting for this for years now.Yeah and as you've seen, they're throwing the HotChips party for it with Mikey C. unlike the usual.
He is a longtime AMD bear who continues to be proven wrong, June 3 isn't far away now.I am not an AMD fan and dislike the behavior of both Apple and AMD fans (you know the whole gang here) in this forum, but Apple fans are really taking the cake lately with this BS. “My Mac is doing like 3W in full ST workloads” lmao, no it isn’t.
M4 isn't in any Macs as of yet anyway.Fine for a tablet but not a laptop.
Pretty far away unfortunately.Pegasus is more exciting.
Oryon is a bit less interesting since LLVM has the actual machine model now.Yep, looking forward to that and the Oryon one.
Continues to be proven right.He is a longtime AMD bear who continues to be proven wrong, June 3 isn't far away now.
Sure but I’d be surprised if they did 16GB standard. Maybe they will, but idk.M4 isn't in any Macs as of yet anyway.
True. But it’ll just be the M4 core I bet.Pretty far away unfortunately.
A18 is clost at least.
Ehhh that’s only so much, I did see though. Both will be interesting.Oryon is a bit less interesting since LLVM has the actual machine model now.
Who milkegg? Him and there’s one other apple guy here who are the same way and don’t understand basic stuff. Adroc is manipulative when his hype gets going and it’s pathetic how many here bend over, but at least understands some ground truths (among other things) senttoschool/milkegg doesn’t.He is a longtime AMD bear who continues to be proven wrong, June 3 isn't far away now.
It's plenty to give a general overview.Ehhh that’s only so much, I did see though
Hearing this from a guy singing gospels to 8cx g4 is funny.Adroc is manipulative when his hype gets going and it’s pathetic how many here bend over
There’s other things you’d want to see.It's plenty to give a general overview.
it’s a good part. The design wins say as much too. That it irritates people makes it funnier at this stage.Hearing this from a guy singing gospels to 8cx g4 is funny.
Sort of.There’s other things you’d want to see.
It's ok.it’s a good part
Only Dell really committed. Rest are so-so.The design wins say as much too.
It's kinda mid for how much reddit wanks it, plus they missed clk targets in public!That it irritates people makes it funnier at this stage.