OcUK Ryzen 3000/Zen 2 review thread

MartinPrince · 18 Jul 2019 at 23:32

Zernath said:
I suppose a plateau is inevitable, companies are getting reluctant to keep pushing the envelope, as this costs a lot of money. They would much rather sell refresh after refresh, with minimal R&D outlay!

Precisely. Well at least AMD are giving us more core count is good for those folks doing well threaded productivity programs.

Zernath · 18 Jul 2019 at 23:37

MartinPrince said:
Precisely. Well at least AMD are giving us more core count is good for those folks doing well threaded productivity programs.

Competition drives innovation, the problem with this industry has been a lack of competition. Hopefully this is changing!

Hotwired · 18 Jul 2019 at 23:54

MartinPrince said:
Precisely. Well at least AMD are giving us more core count is good for those folks doing well threaded productivity programs.

Seeing as we get regular reports of the highest core count cpus selling hugely every single launch I don't believe AMD has found a limit to the number of cores that the market will buy.

Mercutio · 19 Jul 2019 at 00:15

Hotwired said:
Seeing as we get regular reports of the highest core count cpus selling hugely every single launch I don't believe AMD has found a limit to the number of cores that the market will buy.

TR3 with 64/128 might start to steady the market a bit

Hotwired · 19 Jul 2019 at 00:35

Mercutio said:
TR3 with 64/128 might start to steady the market a bit

I have the feeling they won't. People with the means and the desire will keep on leeching tasks onto their personal desktops which were traditionally impossible to run there (due to not having a high clock speed 64c128t processor in a desktop).

chrcoluk · 19 Jul 2019 at 03:58

Journey said:
Kind of ironic really though, given that people expect to see reviews for free, no? I do see your point, but even back in the magazine buying days with all the reviews in there, they didn't buy the gear it was loaned or given to them.

Sell it after review and the net cost is low just need the capital available for purchases. Then got patreon etc.

The established reviewers are able to afford studios and get a full time living out of it so they are making money from somewhere. Some reviewers when looking at their lifestyles are clearly comfortable.

CAT-THE-FIFTH · 19 Jul 2019 at 15:40

@MartinPrince

Puget Systems have published their Photoshop benchmarks:

https://www.pugetsystems.com/labs/a...adripper-2-Intel-9th-Gen-Intel-X-series-1529/

Nutella33 · 19 Jul 2019 at 15:48

Those i9's getting beaten by a Ryzen 5 3600 is something.

CAT-THE-FIFTH · 19 Jul 2019 at 17:10

Nutella33 said:
Those i9's getting beaten by a Ryzen 5 3600 is something.

It looks an excellent value CPU.

TPU just reviewed it too:
https://www.techpowerup.com/review/amd-ryzen-5-3600/

easyrider · 19 Jul 2019 at 17:11

Nutella33 said:
Those i9's getting beaten by a Ryzen 5 3600 is something.

Impressive!

CAT-THE-FIFTH · 19 Jul 2019 at 17:12

So not only is Asus enabling PCI-E 4.0 on some of its B450 and X470 motherboards so is Biostar:
https://www.techpowerup.com/257515/...pcie-gen-4-on-its-amd-400-series-motherboards

opethdisciple · 19 Jul 2019 at 17:53

See now these ram speed results are confusing.

There are benchmarks showing that the difference between 3000MHz and 3600MHz aren't that huge.

Here tho the 1% lows show huge difference.

Here for example doesn't show a big difference.

Nutella33 · 19 Jul 2019 at 18:32

opethdisciple said:
for example doesn't show a big difference.

I've found that the only review going over memory impact so far but it baffles me the choices of speed and timings they used to test - 2666 at C16, 3600 at C17 and 4000 at C20 and a 3200 C14 generally slower than 2666 at C16 which we know is not true.

https://www.phoronix.com/scan.php?page=article&item=amd-3900x-ram&num=3 which shows what we would expect and completely disproves the TP results. I can only assume they had some weird bios issue or simply didn't care about the timings? (I know its linux but it shows what the processor is truly reacting like imho and gives the expected 5-8% bump)

scrivz69 · 19 Jul 2019 at 18:54

MartinPrince · 19 Jul 2019 at 19:39

CAT-THE-FIFTH said:
@MartinPrince

Puget Systems have published their Photoshop benchmarks:

https://www.pugetsystems.com/labs/a...adripper-2-Intel-9th-Gen-Intel-X-series-1529/

Cheers Ctd, makes for interesting reading, especially the detailed breakdown of certain tasks. Either way there seems very little in it between the 3900X and the 9900K. It will be good to test this out myself.

opethdisciple · 19 Jul 2019 at 20:22

Nutella33 said:
I've found that the only review going over memory impact so far but it baffles me the choices of speed and timings they used to test - 2666 at C16, 3600 at C17 and 4000 at C20 and a 3200 C14 generally slower than 2666 at C16 which we know is not true.

https://www.phoronix.com/scan.php?page=article&item=amd-3900x-ram&num=3 which shows what we would expect and completely disproves the TP results. I can only assume they had some weird bios issue or simply didn't care about the timings? (I know its linux but it shows what the processor is truly reacting like imho and gives the expected 5-8% bump)

Ah yes. But those are workstation class tasks. I think games are far more reactive to memory speed.

At least that's what I thought?

Or rather one review put it like this: "Ryzen 3000 gaming performance is not as reactive to memory speed as Ryzen 1000 and 2000."

opethdisciple · 19 Jul 2019 at 21:37

Do we believe this guys benchmark results? He has the 3800x effectively faster than the 9900k.

From about 14mins on wards he starts talking about the gaming performance.

Firejack · 19 Jul 2019 at 23:10

Yes you can believe Wendell @ Level1techs results. He is well respected in the tech press. You often seem him invited to appear on big tech Youtube channels such as Linus Tech Tips, Gamers Nexus etc. He really knows his stuff.

chrcoluk · 20 Jul 2019 at 03:35

opethdisciple said:
See now these ram speed results are confusing.

There are benchmarks showing that the difference between 3000MHz and 3600MHz aren't that huge.

Here tho the 1% lows show huge difference.

Here for example doesn't show a big difference.

3rd link doesnt show lows thats why, just crappy avg's.
2nd link you didnt put in post so cant see.

When textures are been loaded from gpu memory ram wont help, but when they first loaded from disc ram performance can help, some games dont use the gpu at all for certain textures or data and those games are more likely to see larger differences such as lightning returns. (games with these characteristics usually dont get reviewed by reviewers).

Some games are coded in a mess like FF15 where all gpu textures are duplicated into ram, although I never tested FF15 performance to see if ram latency has any impact.

Also i dont think memory bandwidth helps with games, its more so latency, so what would help is people stopped doing 3000 vs 3600 tests but rather did CL12 vs CL14 vs CL16, CR timings etc. I know 3600CL14 latency is lower than 3200CL14 latency but it adds more confusion and unneeded variables. Especially when the latency is increased on the higher speeds.

VoG · 20 Jul 2019 at 09:09

Steve at Gamers Nexus doing one of his ridiculously in depth reviews of the 3800X vs 3700X cpu's, long story short he doesn't think the 3800X is worth it, your better of getting the 3700X & putting the cash difference into a better gfx card.