Discussion Ryzen 3000 series benchmark thread ** Open **

[mohit9206]

The Ryzen 3700X CPU looks the sweetest deal for most users out there. Relatively future-proof, relatively inexpensive (still I'd be glad if it cost, say $270). A CPU to run for the next 7 years, lol. Time to upgrade my aging Intel Core i5 2500.

True, and the 3600 best value gaming CPU. The only question is whether one should go for 3600 or discounted 2700.

 

[Shivansps]

From what im seeing so far (im still watching reviews) Ryzen 3000 is impressive as a whole but i expected A LOT more for gaming.

 

[TheGiant]

It does give you a reason to still buy Intel, especially when 5 Ghz should be easily doable on the 9600K/9700K, and much less reason to buy any Matisse other than the 3600.

The Kraken score is terrific btw, the giant L3 cache might be the reason.

this

my thoughts overall:

mixed feeling 3700X vs 9900K the 8C battle

• performance productivity a tie
• performance gaming still a lose, albeit better, by far the best gaming test is here with complete frametime distribution https://techreport.com/review/34672/amd-ryzen-7-3700x-and-ryzen-9-3900x-cpus-reviewed/10
  • the intel 9 and 8 series are still better, when oced even more, but r3x is finally competitive and recommendable
  • streaming https://techreport.com/review/34672/amd-ryzen-7-3700x-and-ryzen-9-3900x-cpus-reviewed/15 so for the AMD marketing
• power varies very much, especially the 9900K vs 3700X power, sometimes its https://www.techspot.com/review/1869-amd-ryzen-3900x-ryzen-3700x/ here its a little more than 7700K power and 80W less than 9900K and https://www.techpowerup.com/review/amd-ryzen-9-3900x/18.html it looks only little less than 9900K

overall we got AMD skylakeIPC1.03 in 2019 with better power (still looking to it, but definitely better)- for me power is the key and it looks like we can get 9900K performance with much less power

the real deal-12C 3900X

• skl-x obsoleted and tr too, unless you have IO or the mem bandwitch need
• Intel has zero response for this, even if man overclocks the skl-x 12C to like 4.7Ghz to be comparable, gaming wise skl-x is much worse so the 3900X is no weakness chip for 500 EUR-the real threat to intel atm
• I must say I dont like how AMD made their TR line pretty much obsolete, it doesnt give a good signal

the 6C 3600- the sweet spot- IMO the best buy, for 200 EUR CPU and 100EUR board you can get a 8700K (non oced) performance and even good with gaming

upgrade- 8700K or 9 series intel chip users- no upgrade, even downgrade atm when oced, but for the first ryzen users this is definitely an upgrade

IMO best summary https://techreport.com/review/34672/amd-ryzen-7-3700x-and-ryzen-9-3900x-cpus-reviewed/17 so if you OEE is high enough, get that r3

the bottom line

• AMD should silence those morons creating the overhype and using their tech knowledge to lie, just stupid lies
• AMD marketing didnt show us how good the 12C really is

 

[Asterox]

Solid CPUs from AMD but I’m a bit disappointed with the OC potential on these. Either way anyone on the 2600kish era seems to have a nice upgrade path here.

But you dont have "a logical reason to be disapointed".

 

[tamz_msc]

Re: the der8auer video about him not reaching the advertised boost clocks, TPU claims otherwise:

The max boost clock was always supposed to be on 1C.

 

[cortexa99]

Last 3 lines are power consumption.

 

[Mopetar]

Existing TR is obsolete but they will have new replacement chips eventually. The good news for now is that mainstream desktop can now pull what previously required HEDT platforms (and HEDT prices) so people can buy in at lower cost.

Don’t know if they’ll offer 64 core parts right out of the gate, and it’s questionable if they’ll even offer a 16-core part, but it’s not as though they’ve killed off the platform.

 

[exquisitechar]

From what im seeing so far (im still watching reviews) Ryzen 3000 is impressive as a whole but i expected A LOT more for gaming.

It's a huge improvement over the second gen. If you expected it to take the crown in gaming, you were a bit too optimistic. It is very hard to accomplish on a new node due to the clock speed deficit, but 7nm+ EUV should help with that.

Anyway, to me, Ryzen 3000 is as expected. Amazing for my use cases, and the gaming performance is more than enough for me since I barely game, and @60Hz. Can't wait to get my hands on a 3700x.

 

[Kedas]

Seems that the advertised boost clocks on the 3900x are not even reached in reality. Even with very high voltages, 4.6GHz is barely reachable on a single core for brief moments.

Maybe it only does this extra with certain software loads or there is something not working on some motherboards.

Despite the long time they had, MFS were still giving bios updates during the reviews.

Anyway looks like AdoredTV will have some work to do to make sense of it

 

[Abwx]

The Stilt's posts on Matisse.

IPC is on par with Coffee Lake S parts in his suite. Many interesting observations, same as last time.

Most interesting is that it show that TSMC s process is awfull beginning at 3.4GHz, Intel is really lucky on this one and it explain why they didnt switch to smaller nodes as laws of physics are the same for everybody...

Up to 3.4GHz power/frequency is of the form P = F^2.64, meaning that 10% higher frequency increase power by 30%.

At 3.4GHz there s an abrupt knee, it become of the form P = F^4 and 10% higher F will result in 50% more power, we are essentially in the same case as GF s 14nm before they improved their process at 12nm.

 

[cortexa99]

Maybe it only does this extra with certain software loads or there is something not working on some motherboards.

Despite the long time they had, MFS were still giving bios updates during the reviews.

well.... I dont know if anything is set in stone I really hate to say but, I actually read that PBO up to 4.85Ghz in some cases, but I don't know which cases.
But I agree that most important is performance not Gigahertz. IMO won't complain about OC headroom anymore after reviews show up.

Most interesting is that it show that TSMC s process is awfull beginning at 3.4GHz, Intel is really lucky on this one and it explain why they didnt switch to smaller nodes as laws of physics are the same for everybody...

Now I know why Intel 10nm postpone, but still, if the situation goes on I think in the near future AMD would be able to bring <7nm that clocks much near to Intel's offering. That's Intel's problem.

 

[coercitiv]

The Stilt's posts on Matisse.

Sounds like The Stilt is a bit pissed at Asus for delivering a "faulty" BIOS with altered power consumption behavior.

In case of the ASUS Crosshair VIII Hero Wi-Fi motherboard, the media was instructed to use 0066 bios build, which had been vetted and approved by AMD. However, newer bios builds were available and ASUS has also (allegedly) told the media to use those versions. What exactly has transpired here is still under investigation, but regardless of the actual reasons behind it, the consequences might be rather significant. In practical terms, all reviews which were done on ASUS Crosshair VIII Formula or Hero motherboards using other than 0066 bios build must be considered invalid, at least partially. Reviews using other ASUS motherboard models (not provided by AMD) are under suspicion as well.


AMD had no part in it, and the actions by ASUS are the sole reason behind it. The investigation revealed that ASUS is altering one or more power management parameters of the CPU, causing it believe it consumes less power than it actually does. As a result, the frequencies will be higher than the actual power budget would normally allow to. Tricks like this are pretty much a common (mal)practice these days however, there is a good reason why this must be considered worse than the others: this "thing" is completely undetectable without external measurements and rather deep knowledge, but also there is no way to disable it either. Even a person such as myself, who can control most things on these platforms cannot disable this "thing". As you may notice, at the moment I call this issue the "thing", since I'm giving ASUS the benefit of a doubt.

 

[Asterox]

Re: the der8auer video about him not reaching the advertised boost clocks, TPU claims otherwise:

The max boost clock was always supposed to be on 1C.

Well, what can you expect from Pro Youtube Overclocker.

"He probably cant sleep, if he is just a little disappointed" with some CPU overclocking.

Logicaly, it is expected that on Ryzen 3000 max advertised CPU clock is not All Core Turbo.

 

[guachi]

Looking at the Anandtech gaming results and *only* looking at 1080 results (because it's the most popular resolution):

The 3700X and 3900X are 98% as fast as a 9700K/9900K. (The two AMD chips are almost equal and the two Intel chips are almost equal). And the 3700X is 88% the cost of the 9700K.

Yes, the Intel chips are faster but not meaningfully so considering everything else you get. For comparison, the 2700X is 88% as fast as a 9700K.

 

[Kedas]

Logicaly, it is expected that on Ryzen 3000 max advertised CPU clock is not All Core Turbo.

That's not the problem, he couldn't hit this clock with 1 thread

 

[TheGiant]

Well, what can you expect from Pro Youtube Overclocker.

"He probably cant sleep, if he is just a little disappointed" with some CPU overclocking.

Logicaly, it is expected that on Ryzen 3000 max advertised CPU clock is not All Core Turbo.

Breathe please, breathe

 

[Abwx]

That's not the problem, he couldn't hit this clock with 1 thread

He s no officialy supported reviewer, how did he receive the proper bioses.?.

 

[guachi]

Just checked performance/$ for the chips tested in the Anandtech review. If you eliminate the A12-9800, the 3700X was the best performance/$ in every test I checked (I skipped checking the 4k gaming results).

All of them.

 

[mohit9206]

There is no reason for anyone to buy an Intel CPU anymore (mainstream consumers). I feel pretty good about it as its really been a long time coming. I just hope AMD doesn't get overconfident and start screwing up again.

 

[AtenRa]

Last 3 lines are power consumption.
View attachment 8114

According to this in CB R20 Multi , the 3700X has almost the same performance vs the Core i7 9900X at half the power and $200 cheaper.

 

[Fanatical Meat]

There is no reason for anyone to buy an Intel CPU anymore (mainstream consumers). I feel pretty good about it as its really been a long time coming. I just hope AMD doesn't get overconfident and start screwing up again.

They will, that is what all businesses do

 

[CHADBOGA]

It is a pity that thread about Overclocking being outdated these days got filled up with nonsense about cars and then closed.

 

[HurleyBird]

The Stilt's posts on Matisse.

IPC is on par with Coffee Lake S parts in his suite. Many interesting observations, same as last time.

This is a pretty good confirmation of what many have expected then, that The Stilt's suite is a bit biased towards Intel. I don't think he's biased though... just random chance that any assortment of benchmarks will end up favouring one architecture or another to a certain amount.

 

[IEC]

Looks like there is a lot of per-core variation in maximum frequency achievable within the FIT limits. And your 6-core chiplet SKUs will on average do better in regards to those versus the 8-core chiplets.

That tells me that TSMC's 7nm process is not all that mature, even if yields are supposedly good. It remains to be seen whether the 3950X will change that, though I suspect we will need to wait until next year for 7nm+ for anything better.

I was correct to temper expectations on overclocks. Looks like for 3900X and the 8c SKUs running stock will be the best option. 3600 is the only possible exception (and best value).

That won't stop me from picking up a 3950X when they are available, but I will be wary of picking up an early sample unless it shows improvement in per-core variation and hotspots. As well as start shopping for a new case to hold a custom loop with at least 720mm of rad space...

 

[lixlax]

Note that the few reviews that have done memory BW tests regarding write speeds: single chiplet processors are ~25GB/s while the 3900X has ~50GB/s. Quite interesting.