Quote from: chicanto on March 28, 2020, 20:50:23
It's fair to compare to the 3950X becuase the 3950X has the highest performance 8-core CCXs.

fair to compare what, 8 cores vs 8 cores? sure but you need a strong context about exactly 8 core limit, no more.

4900H is more efficient per-core because it is a single monolithic die. There is no separate 14nm I/O die.

It's fair to compare to the 3950X becuase the 3950X has the highest performance 8-core CCXs.

Title is clickbait as hell. Compare only 8 cores on a 16C chip and call it 'outscore'? What a joke.

Quote from: Duckers on March 25, 2020, 18:29:49
why would you even consider using userbench as a source? wait for actual proper benchmarks from anandtech instead. Userbenchmark has really unrelieable numbers.

userbenchmark is the best place for users..and benchmarking. best interface&widest hardware.

This'll just be because userbenchmark hasn't yet worked out how they can mess with the scores to make it look bad. Once they give these APUs the same treatment they gave all AMD's other chips they'll be in a more realistic place within AMD's lineup.

@gc - You would have to be bottlenecked by something to get that. Multi-threads aren't magic.

End of the day 4800HS has shown Geekbench 5 scores of 1164 single thread and 7544 multi thread (which is NOT over 10X). So I find it hard to believe that the 4900HS has 3x less single thread performance than a 4800HS single core and almost 2x less multicore.

Even a 4500U gets more than 398 single core performance. Hell even an old 3500U gets 730 geekbench 5 performance. Are you really gonna tell me a 3500U is 2X faster in single core performance than a 4900HS?

why would you even consider using userbench as a source? wait for actual proper benchmarks from anandtech instead. Userbenchmark has really unrelieable numbers.

10x is possible when comparing single thread vs. 16 threads.

lol, so many people being outraged based on the headline without even bothering to read the actual article.

The article clearly said that the results should be taken with a grain of salt, even in the results, it won on average is a single benchmark. Again, on "AVERAGE" and this is a single benchmark.

As benchmarks vary on what they test, it isn't uncommon for a new CPU to beat an old one in some aspects.

In this case, the userbenchmark checks for integer speeds. Not to mention it is only talking about use of 8 cores, while 3950x is 16 cores.

@RallyMax - As for the geekbench numbers, where? I only found geekbench 4 numbers which show 4900HS at 5,309 points single core. In comparison, the 3950X is 5,570. So it isn't impossible for the 4900HS to beat the 3950X core for core. The same applies to the 3700X.

Overall, since ram speed plays a big role in AMD processors, that might cause variance in the average.

I find it hard to believe that the 4900HS gets 4x less performance in geekbench 5 then the 3700X. If anything, I find your numbers way more suspicious than the userbenchmark numbers. Why? How would you get 398 on single core and 4546 on 8 core multicore? That is a over 10X increase in performance for 8 cores. That would require some very serious voodoo magic.

The level of either stupidity or incompetence expressed in this article is unreal.
None of this is accurate.
It's so bad that I filtered this organization out of my news feed.
Short term clickbait for long term failure.

Absolutely flawed article. The same machine popped up on geekbench.

Single core: 398
Multi-core: 4546

The scores are nowhere 3700X, which posts averages of

Single Core: 1258
Multi Core: 8738

It is possible because these are monolithic processors so there is much lower latency as compared to desktop chiplet parts

Quote from: william blake on March 25, 2020, 09:32:40
people are unable to read

To me, it feels more like a click bait.

people are unable to read
1440 peak multicore for 4900hs
3000+ peak multicore for 3950x