Discussion in 'Frontpage news' started by Hilbert Hagedoorn, Jul 21, 2021 at 8:35 AM.
You are acting rather pathetic. Go get some air or something.
Some people worry about games jumping on the little cores ? Is not hard to set core affinity!
I hope this is true the cpu battle hopefully is about to get wild !
Will this be Fabbed in the U.S. ?
Looks cool but, I will have to see further evidence before I consider this real. However, isn't competition a great thing?
It'll only be 5000$!
That's my main worry tbh. I just don't see any incentive for Intel to compete on price. Especially if these numbers come true.
ok! Well done ! But what will it cost? I mean performance/price is important ....
Gotta agree with others here that the numbers seem highly dubious. 30% on a single generation while not necessarily impossible, is unlikely given Intel's track record of 5%-9% gains. Though we do still have rather ambitious boosts of 5.3 all core. Unless Intel have finally fixed their 10nm node, I can't see 5.3 without some silly power targets.
The multicore is a more interesting prospect, A 16 core, 24 thread (the ".little" cores are reported to not have multithreading via Tweaktown) with a supposed 5.3 all core should perform well, but still beating out their own 18c/36t monolithic makes me question it. While yes it may have the clockspeed to beef it up a bit, it's still going against a unified core design, and a final product to boot, vs a QS. Again we're not really seeing "realistic" figures here either, it's still a 31% increase.
While this no doubt will be an improvement over the 11900, I can't see that improvement being much to say the least.
At the very least BIG.little shouldn't see many problems on arrival thanks to a lot of software already being tweaked to deal with Ryzen's CCX design, so accounting for different core clusters is basically already done there. Though I am struggling to think of a market for something like this outside of "Performance" Ultrabooks.
All of this is purely conjecture though. If Intel have somehow managed to, without a node change achieve the same kind of gains AMD made on a 28nm to 14nm move, and an entirely new architecture, then others are right, they won't have kept it quiet for this long. They'd have been shouting it from the rooftops.
The less complex your architecture is, generally speaking, the more responsive it is. Take this graph for example:
Notice how the moment the 1800X needs to use both CCXs, the latency skyrockets. But it's not just a CCX issue, because notice how even though each of the Intel CPUs are very similar to each other and use the more efficient ringbus approach, each time the core count goes up, the latency goes up with it.
Also notice how after 2 threads (and 8 threads, for AMD), it doesn't matter how many more threads you ping - the total complexity of the chip is the only thing that matters once you start multitasking. So for something like the 5960X, it doesn't matter that 8 of the 10 cores are being ignored. Just simply the fact those cores exist is enough to affect the latency.
I can't remember where I saw it, but I remember ARM having some inter-core communication being a small fraction compared to x86 parts.
It's not just a CPU thing either - there is such thing as too much RAM. Granted, nowadays the point of diminishing returns is much harder to reach. It's been known for a long while too that the bigger the CPU cache, the higher the latency gets. In modern CPUs, the tradeoff is worth it in order to fit more complicated instructions. I figure that's why the closer you get to the core, the smaller the cache gets.
So - if the little cores can clock high enough and if the game can be locked to only run on the little cores, they should be able to calculate a frame in less time. This is of course assuming the little cores can run the game at all, and more importantly, don't have to use up more clock cycles in order to compensate for missing instructions. So, what I'm talking about here is pretty conditional, and theoretical.
As pointed out earlier, it doesn't matter how big and fancy your CPU is if you can't take advantage of the transistors. Software has to be programmed to use additional instructions or cores. It doesn't matter how big your cores are if they're mostly going to waste. That's why Apple's M1 can achieve similar performance to an i7 at 1/10 the power consumption - it only contains what is necessary for optimal performance.
My only issue with what you're saying here is with Ryzen 5000/Zen 3, AMD actually beat Intel when it comes to gaming broadly speaking at release so they were ahead in both categories for productivity/gaming. 5900X is an absolute beast for games and is generally faster than even the 10900K going off the test suites I've seen. Of course the price reflects this and Intel have provided better gaming performance per dollar recently from what I gather. If all I wanted to do was play games I'd p
People have been disappointed with Intel I think since they probably could've done better sooner, but didn't to chase better profit margins while they were ahead. I'll buy whatever CPU gives me the performance I want at a reasonable price, but I do understand why people are pissed at Intel. 11th gen was fairly disappointing in my estiation. iirc Gamer's Nexus called it a "waste of sand". I wouldn't say it's quite that bad, but it doesn't look good when you're getting performance regressions in some tasks at a higher price with minimal gains when it is faster VS your last generation.
Somewhat related I'm still kinda miffed Digital Foundry skipped Ryzen 5000/Zen 3 testing all together. Seems to me it was one of the biggest hardware launches of the year and they usually test those sorts of things, but did not in the case of Ryzen 5000 for some reason. Just seems weird they wouldn't test them, historically they have tested lots of CPUs for gaming perf.
Intel inside, keep a chiller at it's side
You don't think there's any validity to the criticism towards Intel? Surely it's at least a mixture of some deserved criticism aimed their way rather than all "hate" and "bullying". My observation has been different than yours, in that there's definitely "toxic" exchanges, sure, but I don't think it's reasonable to dismiss it all as that I mean. What is the root of the toxicity? Why are people upset with Intel? Why are people happy with AMD? There are at least some valid reasons/answers to those questions and sometimes their discussed in a fair way which is fine.
You mention you've not seen any actual contribution from anyone on this forum. Do you mean regarding CPU discussion only? For general troubleshooting I've gotten help tons of times just in recent months (you can check my thread posts, there are tons of useful replies for questions I had about X or Y software feature). People are definitely still helping each other out, perhaps not as much as they used to if that's what you mean, I'm not sure, but this is definitely a much better PC gaming/help forum than say, reddit, where there is just a wild amount of misinformation circulating.
The Little and Big cores can all work simultaneously for heavy workloads no? Or does only one of the sections operate at a given time (so, BIG cores only OR little cores only)?
Trust me, AMD would have done exactly the same thing if Intel performed like those Bulldozers.
Dig a little deeper in Intel history and you will arrive at Core 2 Duo - it was a tremendous step in performance vs previous generation so it is very possible.
I hope that it is true - I don't like my 5950X at all - a big furnace of heat, expensive and also the platform is not as stable as Intel.
I totally agree/you're correct, but then people would be criticizing AMD for having crappy CPUs like they did in that time would be my expectation. I'm just trying to say that sometimes people are hard on a company and it's not for no reason. Other times, yes, it can get out of hand and go beyond what's reasonable and I can see why people are getting annoyed with the "AMD good, Intel bad" stuff at this point, sure.
I certainly agree with you that in terms of price to performance in games, probably 10th gen Intel would be my buy right now, but if money was no object the 5950X is just an incredible CPU (assuming you can cool it, my lord these things get hot these days).
I'm not sure where you are getting this information.
5.3ghz is single core boost speed (maybe more the one?) And the all-core is 5.0ghz, however, that's only for 8 cores, not 16.
The other 8 cores all core is 3.6ghz.
That's probably the most believable part of this story, it's not far off the 11900k 8 core parts.
He's a cool member...always upbeat, we could use a few more like that around here.
Thanks Doggie, i kinda deserved it tho, love you @Airbud folks like you make this place home.
instead of those who always bitch and moan,like they were in agony.