• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Discussion Intel current and future Lakes & Rapids thread

Page 390 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
DrMrLordX said:
Well a certain other company just respan their entire core on the same process with a net increase in transistor count and hit paydirt doing it. The real problem is that Cypress Cove wasn't purpose-built for 14nm. It seems like a hackjob.
Click to expand...
I assume you mean to say 'respun'. Actually, they came up with a new architecture (evolutionary, not blue sky).
 

AnandTech Forums: Technology, Hardware, Software, and Deals

Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
www.anandtech.com

Pivotal moment for Intel. After selling their NAND flash memory production facilities to SK Hynix, can they really afford to maintain another proprietary standard ?

If Intel does decide to buy Micron's facilities then I guess they don't really care about the quality of their fabs output anymore since they'll be wasting time doing other things. Otherwise Intel should just let some outside party acquire 3D XPoint technology or just let it die ...
 
Very interesting, people have to work quite hard on tuning and testing to get 5.1 stable with varied workloads. And this pretty much means that as long as MB has decent power delivery and proper 280+ AIO is on CPU it will run 5.1 24/7.
I think manual tuning might still be preferred to avoid excess voltages and watts, but for average Joe this is great tech.

Intel must be binning these i9 K CPU very very hard.
 
JoeRambo said:
Intel must be binning these i9 K CPU very very hard.
Click to expand...

All of these various turbo modes is really unnecessary. Each part of the Rocket Lake stack basically runs at max programmed speeds if cooling is available. Intel should just state that and the max frequencies/core(s).

The "K" parts are also getting superfluous because as Joe Rambo stated Intel is obviously clocking the 11900K to an inch of it's life. Unless you are running exotic cooling there won't be much left.

Now with all of that being said, as yields increase over the next 4 or 5 months perhaps you will be able to pick up a 11700K that can be overclocked to perform like a 11900K, which is why we all started overclocking in the first place.

Oh and I just can't wait to see the power draw from a well-cooled 11900K running 5.1GHz all cores with a heavy load saturating all cores.

I think the betting line should be "under/over 325W."

I'm going say over with AVX512 engaged.
 
Hulk said:
Oh and I just can't wait to see the power draw from a well-cooled 11900K running 5.1GHz all cores with a heavy load saturating all cores.
Click to expand...

For static overclock it would be true ( and would crash in Prime95 AVX512 ). Intel's solution would be limited by amps and temps and CPU would downclock to stay within Intel defined limits.

But is it really relevant? During gaming it won't go above 100w, while having 5.1-5.3ghz clocks depending on workload. And those who love to heat their rooms with CPUs are better off with buying 10980xe or 5950x, both can do 300-500w when pushed, at least you have cores to show for it.
 
JoeRambo said:
For static overclock it would be true ( and would crash in Prime95 AVX512 ). Intel's solution would be limited by amps and temps and CPU would downclock to stay within Intel defined limits.

But is it really relevant? During gaming it won't go above 100w, while having 5.1-5.3ghz clocks depending on workload. And those who love to heat their rooms with CPUs are better off with buying 10980xe or 5950x, both can do 300-500w when pushed, at least you have cores to show for it.
Click to expand...

Bull. A 5950X will not use nearly that power. A 10980XE will not either unless overlocked to its limit.
 
Thunder 57 said:
There is no point to OCing a 5950X. Try again.
Click to expand...

On topic of "no point", i thought this was enthusiast forum, sorry, my mistake.
But lets stop these discussion right here, i was just commenting that for proper MT loads, there are way better CPUs to generate heat, ones that have 16+ cores. No need to beat dead 14nm "efficiency" horse any more.

Zucker2k said:
And you'd need a chiller to cool it at that sustained power.
Click to expand...

I'd say anything above 200W is ridiculous already. If you are doing anything that is using >200W on desktop CPU you have to ask yourself if Threadripper or 3090 would not serve that workload better
 
JoeRambo said:
On topic of "no point", i thought this was enthusiast forum, sorry, my mistake.
But lets stop these discussion right here, i was just commenting that for proper MT loads, there are way better CPUs to generate heat, ones that have 16+ cores. No need to beat dead 14nm "efficiency" horse any more.



I'd say anything above 200W is ridiculous already. If you are doing anything that is using >200W on desktop CPU you have to ask yourself if Threadripper or 3090 would not serve that workload better
Click to expand...
Those are prohibitively more expensive. Thank you AMD. Also, desktop users generally don't need all those cores, yet demand all the performance they can get. Basically, AMD is maxing TSMC 7nm, while Intel is maxing its own 14nm. The consumption difference, outside of AVX is within ballpark of the efficiency delta between both processes.
 
eek2121 said:
Zen 3 is notable because it is the first CPU where AMD has Intel beat in single/multicore workloads while consuming 25-50% less power with twice the core count.
Click to expand...

Yeah, and ZEN2 was already damn efficient, lets not discount them.

I have ML ( AI or whatever buzzword to call neural net driven algos is nowadays ) workload that i run on two system:

10900K 5.1Ghz static OC - runs ~170w
3950X 4.05 static undervolt - runs ~80w

They both complete the workload in same time, 10C vs 16C.

The problem? My 3090 runs order of magnitude faster while consuming 300W.
 
You could probably buy some kind of FPGA that would do it even faster for even less power. Of course the CPUs will still be good for generalized compute tasks that the FPGA would either need to be completely reprogrammed to handle (if it could do it at all) and that the GPU might similarly balk at because it doesn't have any hardware acceleration for that work or the ability to parallelize it well.

Both of those CPUs could render Crysis (The AT review says the 10900K can do a little over 6 FPS) but the 3090 would be several orders of magnitude faster. The 3090 probably sucks at compiling a Linux kernel though. The problem only exists if you narrowly define workloads to include things that are better done on some other piece of hardware.
 
Mopetar said:
Both of those CPUs could render Crysis (The AT review says the 10900K can do a little over 6 FPS) but the 3090 would be several orders of magnitude faster. The 3090 probably sucks at compiling a Linux kernel though. The problem only exists if you narrowly define workloads to include things that are better done on some other piece of hardware.
Click to expand...

Once we remove laptops and professional workstations from equation, i think what is left are average consumer guys who mostly game, browse web and watch videos on their system.
Despite all appearance on these forums, only fraction of % actually do DC, encode videos or render. And even those niches are shrinking fast due to GPU renderers, builtin encoders, DC/miners running on GPUs.

When all is said and done, desktop CPUs will sit most of the time idle and in sub 100W region when loaded with game or Youtube in Chrome. I think both AMD and Intel serve this niche very well currently, and Intel is gonna get more competitive performance wise with RKL-S. They already have amazing price/perf deal with CML-S, which i hope will push AMD to release more SKUs.
 
DrMrLordX said:
I was tempted to say that, but I think that would be wrong.

www.vocabulix.com

Conjugation of spin

Spin - Verb conjugation in English. Learn how to conjugate spin in various tenses. Present: I spin, you spin, he spins ...
www.vocabulix.com

I was using past tense, so span/respan would be the proper conjugation.
Click to expand...
It’s the past tense, which is spun. Using the present tense it would be 'respin'.
 
eek2121 said:
A 5950X will use...143W of power.

Zen 3 is notable because it is the first CPU where AMD has Intel beat in single/multicore workloads while consuming 25-50% less power with twice the core count.

Intel really has to turn it around.
Click to expand...
Intel does need to turn it around but, financially they are still doing well because AMD is so constrained on the production side. So Intel still has time on their side...for now.
 
JoeRambo said:
But is it really relevant? During gaming it won't go above 100w, while having 5.1-5.3ghz clocks depending on workload. And those who love to heat their rooms with CPUs are better off with buying 10980xe or 5950x, both can do 300-500w when pushed, at least you have cores to show for it.
Click to expand...

No, you are right, it's not really relevant. But we discuss "fringe" cases here and I am really curious how much power the 11900K can consume at peak load with proper cooling. Intel is pushing the boundaries with this part to be as competitive as possible and I'm looking forward to see what they consider "safe" limits for a consumer product that will have a long lifespan. Besides the curiosity part of it it also informs us of the reliable overclocking limits of the 14nm process (per Intel), which of course gives us insight to overclocking in general. Intel has to sell and warranty these parts.

Of course we all remember when Intel pushed it just a bit too far with the PIII 1133 in an effort to defeat the Athlon. They drove a little too far down that road once...
 
JoeRambo said:
Yeah, and ZEN2 was already damn efficient, lets not discount them.

I have ML ( AI or whatever buzzword to call neural net driven algos is nowadays ) workload that i run on two system:

10900K 5.1Ghz static OC - runs ~170w
3950X 4.05 static undervolt - runs ~80w

They both complete the workload in same time, 10C vs 16C.

The problem? My 3090 runs order of magnitude faster while consuming 300W.
Click to expand...
Really? Also, my car goes infinite times faster while only consuming 100x the power.
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top