*** AMD "Zen 4" thread (inc AM5/APU discussion) ***

[CAT-THE-FIFTH]

So at 1080p and 1440p the X3D excels but doesn't make much difference at 4K or above, apart from the 1% lows.

The 1% lows are the metric you should be looking at anyway because it means much smoother FPS. This is why frametimes are far more important to look at if possible.

What is even more important is the Ryzen 7 5800X3D matches or beats Intel ADL CPUs in games where normal Zen3 CPUs do relatively poorly. This is because these sorts of games tend to be more sensitive to memory-CCX latency. That massive amount of L3 cache helps massively in that regard.

It wouldn't surprise me if the doubling of the L3 cache size on Zen4 and DDR5 memory is helping in games compared to Zen3. Zen4 is basically Zen3 on steroids. Zen5 is the bigger change AFAIK.

 

[Freddie1980]

I agree, the lows should be the measurement of a CPU's performance as that's where all the design differences and choices manifest themselves. But I do understand why reviewers prefer to use averages (lows can be effected by other system components or slight changes in the benchmark run although if your running a benchmark multiple times and the results are repeatable there's an argument it should stay in).

 

[RBImGuy]

I agree, the lows should be the measurement of a CPU's performance as that's where all the design differences and choices manifest themselves. But I do understand why reviewers prefer to use averages (lows can be effected by other system components or slight changes in the benchmark run although if your running a benchmark multiple times and the results are repeatable there's an argument it should stay in).

Issue with tech reviewers is that the data dont apply for you.
Benchmarks runs the same but gaming does not.

Its why a 5800x3d can be the better buy than an intel set up due to the sheer memory cache advantage tech.
Its why I went with a zen4 7600x and then at some point upgrade to the zen4 x3D version next year.

The reviewers rarely if ever run games that push your hardware while gaming.
So, there findings don't apply for you really.

You dont feel fps as you feel delay aka latency which amd tech with x3D excels at removes as a factor while gaming

 

[Dave2150]

Well that's a load of rubbish. All the games I play utilize <16GB of RAM. RAM is very easy to upgrade later also.

What games do you play?

Screenshot below is from Plague Story Requiem, check the memory in use. 32GB has been required for smooth gameplay (without page file usage) for a while now.

 

[Vigorito]

Sure,if you insert 64gb pc will eat almost all 64gb old story,16gb is okay,tonight i will installl same game and test with 2x8gb 6000mhz on 7600x vs 4090,and see if game run smooth but i allready know a answer

 

[Dave2150]

Sure,if you insert 64gb pc will eat almost all 64gb old story,16gb is okay,tonight i will installl same game and test with 2x8gb 6000mhz on 7600x vs 4090,and see if game run smooth but i allready know a answer

Combining 16GB RAM with a RTX4090. As much a good idea as buying a £10k dolby atmos setup to go with a black and white TV.

That's enough forum madness for me today

 

[Vigorito]

Really? CP 77 works pretty good with 2x8gb,why not any other game

 

[Noxia]

I don't know enough to comment but I know plague tale is really badly optimised on pc as well as console so maybe not a good example.

 

[Vigorito]

People are commenting and giving false advices nothing new,i guess they were bored

 

[Dg834man]

People are commenting and giving false advices nothing new,i guess they were bored

Got to flex the e-peen and get a 13900ks, 128gb ram, 4090 on a £700 motherboard and do nothing apart from run benchmarks on it

 

[g67575]

2x16GB is a lot of RAM, for most it's not needed, you can easily get by with 2x8GB.

For many, making minor adjustments to game settings may be preferable to buying 2 16GB modules.

 

[g67575]

I'm having some trouble getting Prime95 stability with AVX512 instructions enabled (it's a been quite a tedious process to test it, with Prime95 passing with no errors after 1-2 hours, other times failing after about 15 minutes or so).

I'm trying a PBO curve magnitude of -24 now (down from -30).

The stability has been fine in P95 with AVX instructions disabled, but getting stability with AVX enabled is more time consuming.

 

[chroniclard]

What games do you play?

Screenshot below is from Plague Story Requiem, check the memory in use. 32GB has been required for smooth gameplay (without page file usage) for a while now.

Citation required.

 

[g67575]

Can't Windows use more of the pagefile during gaming anyway, and not games themselves?

 

[Frenzy]

Gah I thought it would be clear cut on which cpu to buy now but having not kept up with the information my head is spinning.

 

[g67575]

It does seem like you can use a 7900X or 7950X with SMT disabled, to get improved performance (in games).

Has anyone tried this yet?

 

[g67575]

What's a reasonable PBO curve optimizer magnitude setting?

Is -20 likely to be stable in Prime95 (AVX512 enabled)?

 

[chroniclard]

What's a reasonable PBO curve optimizer magnitude setting?

Is -20 likely to be stable in Prime95 (AVX512 enabled)?

Each CPU is different, you have tune each core individually don't you?

 

[g67575]

Each CPU is different, you have tune each core individually don't you?

That would take too long in my opinion. I've just been using the all core setting.

 

[chroniclard]

That would take too long in my opinion. I've just been using the all core setting.

There are a few useful tools that speed it up somewhat, but yes it takes a while.