As owner of multiple AMD and Intel CPUs i feel that situation is not as clear cut as You put it with power. For casual desktop user/gamer, i think even 5950x vs 10900K will consume similar total system power over time.
Thing is, ~5ghz is 15-20W affair on AMD and ~25W on Intel per core. So casual gaming or web browsing is not that different and only when rendering or compiling 24/7 AMD's efficiency will be fully realized.
Where AMD suffers is idle power consumption, esp with 2 CCD chips and low load regime things are frankly not rosy and i feel for my average desktop usage and gaming, AMD would come out not that far ahead in power efficiency.
Now where AMD really shines is when tuned for ~4-4.4Ghz clocks, now that is the efficiency Intel's chips can't really touch, but further clock scaling and esp stock boosting algorithms waste power big time.
EDIT: not discussing servers here, nor mobile, just Desktop as it applies for 99% of users + dGPU.
As owner of multiple AMD and Intel CPUs i feel that situation is not as clear cut as You put it with power. For casual desktop user/gamer, i think even 5950x vs 10900K will consume similar total system power over time.
Thing is, ~5ghz is 15-20W affair on AMD and ~25W on Intel per core. So casual gaming or web browsing is not that different and only when rendering or compiling 24/7 AMD's efficiency will be fully realized.
Where AMD suffers is idle power consumption, esp with 2 CCD chips and low load regime things are frankly not rosy and i feel for my average desktop usage and gaming, AMD would come out not that far ahead in power efficiency.
Now where AMD really shines is when tuned for ~4-4.4Ghz clocks, now that is the efficiency Intel's chips can't really touch, but further clock scaling and esp stock boosting algorithms waste power big time.
EDIT: not discussing servers here, nor mobile, just Desktop as it applies for 99% of users + dGPU.
Hard to believe such claims when ARM themselves claim X1->X2 is +16% Integer IPC?
X2 is gonna be huge hit, but beating GC? Not there yet.
Yes it is. I edit my post just now. He explained in later post that arch perf = IPC, means GC and X2 has same IPC in SPEC, from his given info.Having seen the original posts in Chinese, "architecture performance" most likely refers to IPC.
28 seconds or 56 seconds is default depending on the processor. That is, unless you have your motherboard set it to a longer amount of time.PL2 consumption is only for a few minutes.
I don't think your workload is only for a few minutes, so what's the problem? Performance at 125W?
28 seconds or 56 seconds is default depending on the processor. That is, unless you have your motherboard set it to a longer amount of time.
Eek2121 could adjust that amount of time or the max PL2 power to his needs if he really wanted to. Performance for the first minute would a bit lower though.
Depends on the workload and what you define as gimped. Tom's Hardware got a 0.4% multithreaded gain going from an average of 119 W to an average of 150 W with the 11700K: https://cdn.mos.cms.futurecdn.net/38xAAFafxXqYcff5moDEQo.png from this review: https://www.tomshardware.com/reviews/intel-core-i7-11700k-cpu-review/2 If a 0.4% loss is severely gimped to you, then it is severely gimped. I personally would say that had no noticeable effect. I would also say that if multithreaded tasks were your big need, then go Ryzen. Not because of a power reasons, but because the additional cores do better.Both of you are ignoring the fact that Intel 14nm CPUs are severely gimped if constrained to 125W. Golden Cove sounds like it will be a monster, but it also sounds like, despite the gain from 10nm, it will still suck a lot of juice.
Having seen the original posts in Chinese, "architecture performance" most likely refers to IPC. Considering how good the X1's IPC is even with anemic cache (SPEC loves cache), it certainly wouldn't be surprising for the X2 to match if not beat GC in this regard.
Considering that X2 has nearly 10 less pipeline stages, with uop cache miss and about 5 with uop cache hit yes it's possible. At the depth of X2, Golden Cove would be an additional 15-30% faster!
Per clock, they are losing an insane amount of performance trying to reach those absurd frequencies.
In the Anandtech article about the X2, it says how ARM thought reducing 1 stage is worth it for the performance. Also the uop cache on the ARM parts are made to reduce effective pipeline stages. Golden Cove goes and adds an extra stage.
When uop cache was introduced with Sandy Bridge, it went from 16 stages in Nehalem(14 in Core 2), to 14-18. So part of the motivation of the uop cache was to increase clock frequencies. It's not bad as the Pentium 4, where it had 20 stages after the Trace Cache hit(close to 30 on a miss).
Gracemont should be same as Tremont which is 13 stages.
On a desktop where your aim is performance at any cost I guess it's worth it. But elsewhere it really is not.
Depends on the workload and what you define as gimped. Tom's Hardware got a 0.4% multithreaded gain going from an average of 119 W to an average of 150 W with the 11700K: https://cdn.mos.cms.futurecdn.net/38xAAFafxXqYcff5moDEQo.png from this review: https://www.tomshardware.com/reviews/intel-core-i7-11700k-cpu-review/2 If a 0.4% loss is severely gimped to you, then it is severely gimped. I personally would say that had no noticeable effect. I would also say that if multithreaded tasks were your big need, then go Ryzen. Not because of a power reasons, but because the additional cores do better.
We evaluated Intel’s latest generation of “Ice Lake” Xeon processors. Although Intel’s chips were able to compete with AMD in terms of raw performance, the power consumption was several hundred watts higher per server - that’s enormous. This meant that Intel’s Performance per Watt was unattractive.
EDIT: Is Raptor Lake really launching "next year"?
Xeon Icelake lost the competition for the next Cloudflare server:
If it's no good now because of the power consumption, what will be of the next Xeon that may have even more power demands?
Depending on whom you believe, Raptor Lake isn't going to be much more than Alder Lake with some extra Gracemont cores strapped to it, and some tweaks/optimizations to help Golden Cove reach higher clocks.
Cloudeflare was being generous here. IceLake-SP loses plenty of benchmarks to Rome; meanwhile, Cloudflare is big enough that they can probably get early Genoa silicon if they want it. Looks like they went with Milan instead, at least for now. It is not hard to see how perf/watt would favor Milan in a showdown with any IceLake-SP product. A lot of what was holding back IceLake-SP was the bizarrely poor power scaling of 10nm+. Sapphire Rapids will hopefully do a better job at the same clocks, but of course it will have to contend with Genoa. That is not the fight Intel wanted.
Maybe they didn't want to publicly shame who they might want to use again in the future? Intel is really hyping up SPL. Can't wait to see how that turns out.
If they didn't want to publicly shame them, they could have made no announcement at all. Let's not forget Cloudflare adopting Centriq rather publicly in 2018.
If those are not inflated "early" numbers then the price increase is consistent across the entire product stack.As I understand it, it seems that the i9 has increased their prices, with the rest remaining pretty consistent with previous generations.
Given that they launched cannon lake as well - not to mention how AMD "launched" Vega Frontier Edition (calling it FE of all possible things... still a classic Raja move) literally on the last day of the promised quarter - there is ca. zero chance of them not "launching" it in a timeframe that looks good for them.How likely do you think it is that Intel will say desktop Alder Lake launches in October, even if it is a paper launch?
Depending on whom you believe, Raptor Lake isn't going to be much more than Alder Lake with some extra Gracemont cores strapped to it, and some tweaks/optimizations to help Golden Cove reach higher clocks.
There is a big core IPC increase with Raptor Cove and improvements in performance/watt. Intels says it has a much better CPU cache capability for gamers.
I don't quite remember that one as I wasn't paying as much attention back then, but perhaps my above statement was incorrect?
I have doubts of any real IPC gain. It's going to be big enough as it is with the extra small cores. I'm beginning to think the "Improved Performance" is the small cores and that's it. Maybe enabling AVX-512. I don't know how much L3 they can add, that can't be much but you could say it's better.
It's funny when you compare modern core CPUs to the Pentium 4. The final version of the Pentium 4 had a 31-stage pipeline.