AMD Ryzen 9 3900X, SMT on vs SMT off, vs Intel 9900K

[deu]

Awaiting my 3900X delivery, so good to see that gaming performance can improve further since I game and create. To the people wanting to remove the 720p form tests; eventhough they dont say anything about a realistic scenario they can be used to gain insight into where the limitations of the CPU is. The unfortunate thing is that alot of 'less informed' people take these benchmarks as a direction of general performance.

+1 to the vote for '1%' and 'minimum FPS' : It would be nice to see in game tests since they can say more about how the game is actually running and the user experience. But thanks for a extensive and qualitative and quantitive study to give a nuanced understanding of a nuanced subject!

 

[waltc]

Very nice review and testing! AMD for the win, and they deserve it. Intel's been sitting on its rump milking for far too many years--ever the monopolist is Intel.

Interesting thing is that pretty much all these game engines are highly optimized for Intel architectures--so much so that it isn't possible that they be more optimized for Intel than they already are... I'm looking forward to see Ryzen 3k series-optimized game engines, too! Probably won't have long to wait before the first ones arrive, I'd imagine.

 

[moproblems99]

@W1zzard Could you explain the power consumption tests? My understanding is that 9900k uses considerably more power (than even 8700k) and that 3900X uses considerably less. In fact, in your tests a 9900k system draws the same as a i5-9600k or i3-9100f. I am just not understanding how this is possible.

Spoiler: Yours

Spoiler: Tom's

Spoiler: Anand

 

[Nuckles56]

@W1zzard Could you explain the power consumption tests? My understanding is that 9900k uses considerably more power (than even 8700k) and that 3900X uses considerably less. In fact, in your tests a 9900k system draws the same as a i5-9600k or i3-9100f. I am just not understanding how this is possible.

Spoiler: Yours

Spoiler: Tom's

Spoiler: Anand

The reason for the difference is that Wizz uses full system power vs the CPU package power that everyone else uses.

 

[moproblems99]

The reason for the difference is that Wizz uses full system power vs the CPU package power that everyone else uses.

It may explain some of it but it still doesn't make complete sense.

 

[r9]

very good tests

maybe add same tests between 9700K / 3700X and 9600K / 3600X will be good

sadly amd dont offer cpus without ht and lower price example: 3600X without HT around 150us maybe awesome

Neither does Intel. You are looking at this all wrong. Intel charges more for having it not less for not AMD gives it free.

 

[Camm]

Back when I ran a 1950X, I disabled SMT as I found the same thing.

It'll be interesting to see with the 3950X how this plays out, a full 8 core CCX should see less complex hopping.

 

[Crackong]

@W1zzard Could you explain the power consumption tests? My understanding is that 9900k uses considerably more power (than even 8700k) and that 3900X uses considerably less. In fact, in your tests a 9900k system draws the same as a i5-9600k or i3-9100f. I am just not understanding how this is possible.

Spoiler: Yours

And somehow 2600x uses more power than 2700x in gaming.
1600x also uses more power than 1800x and 1700x in gaming.

That's really strange.

 

[SystemMechanic]

Anyone else noticed how the real "winner" here gaming wise appears to be the i5 9400F? 5% slower than zen2, 9% slower than 9700k/9900k, costs 140€....... Really impressed with these charts. Media seems to ignore this little chip, but damn it has a great perf vs price ratio.

Glad you included it in your tests! Thanks

Can you even open a game and some internet tabs on an i5 ??

 

[ToxicTaZ]

Excellent review W1zzard
Wow

3.6GHz 9900K is rock solid IPC gaming CPU!

Never thought 3.6GHz 9900K would stand up to 12 cores 3.8GHz 3900X.... Dam

Next review will be 3950X vs 9900KS.... Most likely

9900K must be really tweaked in the power department quite impressive good old 14nm++.

Definitely looking forward for my CPU swop-out this fall.... 8700K to 9900KS

Here I thought the 9900KS was built to wipe the floor of the 3800X...

9900KS is for us guys with 300 series boards last hurrah.

Bring out those 16 cores 3950X and 4GHz 9900KS reviews

 

[biffzinker]

Never thought 3.6GHz 9900K would stand up to 12 cores 3.8GHz 3900X.... Dam

Probably the 5 GHz Boost helps it along in games. Imagine what the 3900X would do with a 5 GHz Boost.

 

[EarthDog]

Probably the 5 GHz Boost helps it along in games. Imagine what the 3900X would do with a 5 GHz Boost.

Melt iron?

 

[ToxicTaZ]

Probably the 5 GHz Boost helps it along in games. Imagine what the 3900X would do with a 5 GHz Boost.

3.6GHz 9900K is stock. Turbo is 7 cores 4.7GHz with 1 core 5GHz only.

4GHz 9900KS is with a 5GHz Turbo all 8 cores.

Yeah I would love to see an OC review of the same testing 9900K @5.1GHz

Love too see what happens when 3900X is @5GHz?

 

[R0H1T]

Melt iron?

Or create mini black holes inside that package

 

[W1zzard]

My understanding is that 9900k uses considerably more power

Not at stock, the CPU will throttle from highest turbo speeds after around 30 seconds.

 

[InVasMani]

Wonder what happens if you just keep SMT enabled and use MSConfig to limit it to Ryzen Cores. Like what if you limit the SMT to 1 thread per CCX or just use the SMT threads for only one CCX and leave the other untouched? Middle ground compromise perhaps?

 

[Manu_PT]

Can you even open a game and some internet tabs on an i5 ??

Internet tabs?? You mean with 4gb ram? Nop. With 16gb ram you re fine. Debunked alteady by steven burke from GamersNexus. Tabs wont use your cpu, same with discord, spotify, itunes etc. You do renders while playing game? Hell yeah, your i5 will struggle big time, so will your r5 3600. Next.

 

[EarthDog]

Or create mini black holes inside that package

I've heard a singularity is created...

 

[BorgOvermind]

Hope to see Ryz9 in plenty of notebooks.

 

[bug]

Hope to see Ryz9 in plenty of notebooks.

If that doesn't happens soon, it's going to have some tough competition: https://www.notebookcheck.net/10th-...n-par-with-an-AMD-Ryzen-9-3900X.427965.0.html

Already desktop Zen has pretty poor idle power draw. If that cannot be curbed easily, it's bad news for battery life.

 

[medi01]

very good tests

maybe add same tests between 9700K / 3700X and 9600K / 3600X will be good

sadly amd dont offer cpus without ht and lower price example: 3600X without HT around 150us maybe awesome

2600 costs 127 Euro including VAT at the moment.
I also doubt including mainboard into equaton would make it look better for intel.

Why would someone recommend an İntel/Nvidia setup in an AMD SMT On/Off thread?!

Damage control.
Since 3000 release AMD outsells Intel on mindfactory 3.5 to 1.

 

[dirtyferret]

Damage control.
Since 3000 release AMD outsells Intel on mindfactory 3.5 to 1.

Fantastic!!! Now they just need to outsell Intel for the next several years to make a dent in the 80/20 market share lead Intel has (even on STEAM)


According to data from Mercury Research, two generations of Ryzen increased AMD’s x86 desktop unit market share from 12.2% to 17.1% between the first quarters of 2018 and 2019. Its x86 notebook unit market share rose from 8% to 13.1% during the same period. Its market share rose as its rival Intel suffered supply shortages, which encouraged customers to switch to AMD’s CPUs.

AMD Gains PC CPU Market Share from Intel with Ryzen - Market Realist

Advanced Micro Devices (AMD) has come a long way in the PC processor space.

articles2.marketrealist.com

 

[geon2k2]

My take from this is that with any gpu lower than 2080ti there are like 12 cpus which offer virtually indistinguishable performance (<5% difference), so either R5 3600 or i5 9400 will suffice.

I'm looking at the 1440p results and I assume that the lower gpu at 1080p matches 2080 ti at 1440p)

 

[r9]

Anyone else noticed how the real "winner" here gaming wise appears to be the i5 9400F? 5% slower than zen2, 9% slower than 9700k/9900k, costs 140€....... Really impressed with these charts. Media seems to ignore this little chip, but damn it has a great perf vs price ratio.

Glad you included it in your tests! Thanks

Talking about winners Microcenter was selling Ryzen 1600 for $80.

 

[waltc]

Talking about winners Microcenter was selling Ryzen 1600 for $80.

Absolutely! Before jumping to x570 Aorus Master and the 3600X, I came out of an R5 1600 x370/x470 chipset mobo (one board for each chipset, 2 boards, total)--that little cpu did a permanent 800MHz OC ROOB, from 3.2GHz stock clock to 3.8GHz by simply changing the multiplier to 38...and on both motherboards, too! Even the max stock boost was only 3.6Ghz...! Couldn't do 3.9GHz with any degree of stability no matter what I did with voltage--but I saw no reason to complain! Sweet! $80 is a great deal, but recall that it's Ryzen 1, so no need to buy an x570 mboard as they are incompatible with Ryzen 1--I'm very curious as to why that is so--academically speaking--probably something to do with boost & voltage differences, although Ryzen Zen+ was just like Zen in that regard--but I didn't own Zen + so I'm not positive about it. On a budget? How can you beat this? Now the 3600X is in another category altogether, imo. @ 3.8GHz stock, the 3600X begins in frequency just where my 1600 left off at its max OC--but cost the same as a 1600X, or a 2600X cost at introduction! ANd of course the 3600X includes IPC improvements well over 15% in many situations! And the 3600X includes doubled L3--which really does make a difference in gaming. I get the 4.4Ghz boost regularly and reliably with the 3600X under air (stock cooler)--but it cost considerably more than this sale of R1600 for $80.