- Mar 3, 2017
- 1,659
- 6,101
- 136
If anything, uzzi kind of indirectly said that Strix Halo will have that kind of unified cache. Point lost it, but Halo is a plain gaming oriented device.
Yeah it's just the obvious conclusion. Why else would you have that in Point but not in Halo?I wasn't really implying that when I wrote it, but in any case I'm sure other's have said that dozens of times by now.
Strix Point is a tier above Phoenix. It's meant to slot into Premium segments only.
Kraken is the successor of Phoenix for the average laptop.
Holy f* sh* lol I knew it was bad but good god.???
No, you're underestimating how arse MTL is. LNL being as good as it is at 15W relative to MTL only puts it on par with PHX at 15W.
MTL is utterly unusable at 15W, it consistently falls far short of Phoenix on average frame rates and has utterly unplayable frame times to boot at 25W. It just gets even worse at 15W.
Keep in mind that while this is true that MTL is even worse at low wattages in anything — they’re talking about a game, so the GPU as much (more) the issue.View attachment 96702
Interesting. According to this graph, MTL is better than Phoenix at higher wattages, but the opposite is true at lower wattages.
View attachment 96702
Interesting. According to this graph, MTL is better than Phoenix at higher wattages, but the opposite is true at lower wattages.
But even through the lie you can see the truth. That's what he means (I think).That s made up numbers, MTL is not competitive CPU wise at 30W in respect of Phoenix, there s scores of tests that say so, even at 50W it s still barely competitive.
Those curves do not make sense, look at the Snapdragon curve from 5W to 22W, the perf increase almost linearly with power, there s no process that allow for such a curve, at best perf increase as the square root of power with a perfect process that do not exist because of physics limitations.But even through the lie you can see the truth. That's what he means (I think).
PHX performs better than MTL at every tier under 50W, but it gets worse the lower you go, and that's death for a mobile thing.
LNL when?
I wonder how Z5 will perform in low power usage actually.And what does this have to do with Zen 5 ?
The efficiency is so good, you cannot believe it?Those curves do not make sense, look at the Snapdragon curve from 5W to 22W, the perf increase almost linearly with power, there s no process that allow for such a curve, at best perf increase as the square root of power with a perfect process that do not exist because of physics limitations.
Having a wider core allows you to achieve the same performance but at lower frequency, right?I wonder how Z5 will perform in low power usage actually.
If you have a fatter core with a wider front-end, isn't it a given that it'll do worse at the lowest power? Doesn't it need more base Watts to run vs the smaller Z4?
Spectacular.I wonder how Z5 will perform in low power usage actually.
But it's not going to beat the best, right?Spectacular.
Never designed electronics in my life, but that would make sense.Having a wider core allows you to achieve the same performance but at lower frequency, right?
That's high power throughput logic though. "My 8 wide decode can run 33% more stuff at the same frequency as my former 6 wide decode".By reducing frequency, you can gain substantial power savings, that may offset the power increment of going wider.
I guess I'm totally wrong then.Spectacular.
Strix Point will have higher CB 2024 MT score at 20W than Apple's M3. A totally useless metric but I can reveal this to you now.But it's not going to beat the best, right?
*best = Apple
The efficiency of the marketing eventually, what i m talking about is that no process can yield the Snapdragon curve, this has nothing to do with CPU design but with the way transistors works.The efficiency is so good, you cannot believe it?
pffft... Apple cheats. Solders ram right next to the soc. Adds specialist instruction sets to their cpu their own software stack leverages since compatibility is null and void in their world.But it's not going to beat the best, right?
*best = Apple
No, AMD's cheating too. They have at least 1½ cores.AMD does it the hard way. A single core to serve 15W tablets to 500W serverside beasts.
That's fair, but cheating is literally the job.pffft... Apple cheats. Solders ram right next to the soc. Adds specialist instruction sets to their cpu their own software stack leverages since compatibility is null and void in their world.
AMD does it the hard way. A single core to serve 15W tablets to 500W serverside beasts while maintaining compatibility to an astonishingly large amounts of existing software.
Why is it that when I limit my 7950X to 50W it achieves a GB6 MT score that is less than half the score at 100W?The efficiency of the marketing eventually, what i m talking about is that no process can yield the Snapdragon curve, this has nothing to do with CPU design but with the way transistors works.
A mosfet transistor is at best a square law device, wich mean that frequency, and hence perf, increase as the square root of power, that is, you need at least 4x the power to increase perf by 2, or 2x the power to get 1.4x higher perf, if you dont understand this basic law then it s moot to discuss CPUs tech.
I could have guessed that. Of course 4 Zen5 + 8 Zen5C will beat 4P+4E in MT, especially considering that the latter doesn't have SMT either.Strix Point will have higher CB 2024 MT score at 20W than Apple's M3. A totally useless metric but I can reveal this to you now.
Yes, yes I know about frequency/power scaling law. But to call that graph fake without taking into consideration any nuances that could explain it, would be irresponsible. gdansk happily points out this:The efficiency of the marketing eventually, what i m talking about is that no process can yield the Snapdragon curve, this has nothing to do with CPU design but with the way transistors works.
A mosfet transistor is at best a square law device, wich mean that frequency, and hence perf, increase as the square root of power, that is, you need at least 4x the power to increase perf by 2, or 2x the power to get 1.4x higher perf, if you dont understand this basic law then it s moot to discuss CPUs tech.
Why is it that when I limit my 7950X to 50W it achieves a GB6 MT score that is less than half the score at 100W?
It's because there is SoC power vs core power. And while it's at 50% less total power, it goes from about 70W for the cores to play with to about only 20W.
On the low end of the Snapdragon Elite curve, don't you think it's possible there's a similar cause but to a lower degree?