Permabanned
Someone needs to bench 9900K stock on linux
-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
AMD Zen 2 (Ryzen 3000) - *** NO COMPETITOR HINTING ***
- Thread starter Jamin280672
- Start date
Soldato
- Joined
- 16 May 2007
- Posts
- 3,221
That’s one of the two things for my decision to upgrade, the single core speed on Ryzen 2 and the other being the IPC improvement. Just waiting for reviews and real life opinions now to decide, assuming it works fine on my x370 overclocked.
Permabanned
Caporegime
Mate, i've gone from a 4.6Ghz 4690K to a 3.9Ghz Ryzen 1600 and i've seen my FPS go from anything the same to double.
Caporegime
Multicore has come a long way, it’s only single core performance that is meh, even on Zen 3
Single core isnt going to get any better than it is, physical trade offs on heat/power are too much.
There will still be an advantage in only having a single chip die (wrt to latency specific tasks), but that's far more expensive than chiplets.
There will still be an advantage in only having a single chip die (wrt to latency specific tasks), but that's far more expensive than chiplets.
Caporegime
Why do people still believe clock speeds are everything?
SunnyCove with its up to 40% IPC uplift thinks differently.
Based on a single leak. Plus it's a 4 physical core cpu that wont be at all useful going forward once Consoles come out.
All the while 10nm isnt even going to grace desktops as far as i know, so it's irrelevant until the 7nm comes out, god knows how relevant it will even be by then, i'd hope that intel realises it has to speed up the process.
Regardless the physical constraint is still memory speed and DDR5 is coming out soon, so maybe there will be some movement on CPU speed again.
Last edited:
Caporegime
CPU-Z benchmark, i saw those slides, more than half of that huge bar chart was 10% or less.
every time Intel bring out a new CPU they make claims of +15% IPC, which then turns out to be 3% unless in memory benchmarks
The saviour is they go back to the drawing board and do what they when going from Pentium 4 to core2duo a massive increase in efficiency even if performance is lost, I know some people think performance is absolute king and whatever else happens is collateral damage, but I dont think that way of thinking is the way forward, their core products have turned to poo.
I dread to think what kind of watts and heat a 16 core variant of the 9900k would be like. A cooler that fills an entire case and a 1000 watt psu been required.
Associate
Dude please stop peddling nonsense. You were all over Zen 2 until it hurt your feelings. Stop being the salty fanboy.
40% IPC in a single test that was an OUTLIER. Would have thought you'd have been able to see that by now. If it's more than 10% uplift I'll eat two breakfasts.
With Linear scaling a 5ghz 16 core 9900k would have a peak draw of 500w. This would only be possible under LN2 - the amount of heat in that small space is too much to cool with other methods. To keep power draw and temps in check a 16 core 9900k would have to kept around 3ghz to 3.5ghz me thinks. So a 40% loss in single core performance for maybe 50% gain in multi ore performance - in other words it’s not worth it
Associate
Looking back to 1.78 when Ryzen outperformed Intel's offerings by a substantial margin, I think this is just a recurrence. The benchmark isn't optimized for Sunny Cove's cache hierarchy.
Also note that in Intel's slides, the graph they showed was anywhere from 99.4% of the IPC, up to 140%, with 118% being the average.
That means there's 18% on the left and 22% on the right. Most workloads will be lucky to benefit by 18%.
Also, if that's an AVX-512 workload (again, no actual description of how they get to 140%), then yes that is highly likely, but given that AVX is still lacking in adoption, it won't matter to most. They won't see those gains.
Last edited:
The 40% figure comes from one type of calculation - it's not representative of day to day or real life performance - the actual average is much lower
And as good as Sunnycove's IPC may be - they max out at 3.6ghz boost clock - so all the IPC gains is lost by slow clockspeeds
And as good as Sunnycove's IPC may be - they max out at 3.6ghz boost clock - so all the IPC gains is lost by slow clockspeeds
Associate
That's because it's a 15 W mobile processor.
Associate
- Joined
- 27 Apr 2007
- Posts
- 966
The clock-speed is slow for that class of chip also; just look at what the previous generation manages.
Associate
It doesn't mean they actually achieve that frequency though. With Intel defining TDPs at base frequency, if the manufacturers don't account for that in their design, it's unlikely to ever sustain any meaningful turbo. What Intel once described as "guaranteed turbo" now no longer applies. It's a d**k move, in my opinion.
Sunny Cove will have a much more powerful integrated graphics solution and better FPU capabilities, so it makes sense for clocks to fall.
Also, 10 nm is no 14 nm++++++ lol.
Last edited: