***OFFICIAL*** Ryzen 5000 / Zen 3 Launch Thread REVIEWS BEGIN PAGE 39

[SteveGrabowski]

Why do you go on about this like you do? The 3600 will be replaced with an eventual cheaper R5, the 5600X is replacing the 3600X ffs. It's not the end of the world _yet_, just because you won't get it on the 5th of November. Same with cars. The lowest and highest motor options don't always launch right away.

You're just guessing there will be a cheap hexacore. FFS just because it has an X by its name doesn't mean it's not a 50% price increase for their hexacores when you could get one on Zen2 for $200. Might as well have called it the 5600Z and charged $350 since Z > X > ∅ .

 

[tamz_msc]

Statistically, any dataset is more likely to have "outliers" than not.

That does not mean they are not pertinent.

Say, comparing Zen2 to Skylake - in most comparisons Zen2 would win comfortably. But software that benefits from AVX512 would more likely see a big Xeon win. Is that an outlier? Perhaps. Despite that, would it be relevant to a HPC farm running said code? Certainly.

Furthermore, if the benchmark comparisons are themselves just a sample (i.e. 10 games) rather than population of all games - then without confidence your sample represents the population you cannot conclude to what degree one is an outlier. 10 games isn't a large enough dataset* compared to hundreds of games in the world to do anything meaningful except draw premature conclusions.

*assuming said dataset is even accurate!

All that is true, but lets assume for the moment(for lack of independent reviews, hence limited data) that there are 100 games that AMD benchmarked out of which they chose to represent 10(which is a reasonable assumption, IMO). Now there is no reason to assume that the sample is representative, after all it's marketing we are talking about here(even though AMD's marketing have been trustworthy most of the time in the recent past). Let's also assume that the data set is accurate. Under these simple assumptions, it is clear that there are outliers in the usual sense of the term - whether or not they should be considered or discarded will depend on the use case. In your example, someone who uses AVX512 all day will have to consider those outliers, others don't have to.

However, when it comes to averaging a limited, non-representative sample drawn from a population, the average of the sample will get inflated if outliers get included in it - which is precisely what is happening here.

 

[tamz_msc]

Waiting for the official reviews is what a logical person would do. There's no need to overhype or downplay at this time.

I pointed out three things-

1. That there are outliers in AMD's gaming performance numbers.
2. That Comet Lake gains a lot(more than Zen 2) with a memory overclock.
3. That Rocket Lake will end up faster in games if it can clock high enough.

You could say that I'm downplaying AMD's marketing with respect to gaming performance, which is the logical thing to do and I see nothing wrong in that.

 

[tamz_msc]

No

Nope, it's you who's trying to apply it falsely. When comparing CPUs, you're dismissing numbers from games that are actually really CPU bound, this is not using statistics correctly but deliberate deception. Who exactly are you want to be kidding especially on AT forums of all places?

CPU-bound can mean many things - as @Carfax83 pointed out even Doom Eternal is CPU-bound at 1080p, but it does not mean that the gap from Intel will be around the 20% mark every time in CPU-bound scenarios when other games are showing a 5% gap. Such large gains occur less frequently and I have shown that they are the exception rather than the norm and should be excluded when calculating averages from a sample that is not representative.

 

[Markfw]

I pointed out three things-

1. That there are outliers in AMD's gaming performance numbers.
2. That Comet Lake gains a lot(more than Zen 2) with a memory overclock.
3. That Rocket Lake will end up faster in games if it can clock high enough.

You could say that I'm downplaying AMD's marketing with respect to gaming performance, which is the logical thing to do and I see nothing wrong in that.

1) I won't get in in that, except to say as a former statistician, you are wrong.
2) You are wrong. It gains, but AMD gains more, I won't bother to link it as you will never admit its true.
3) This is a pro-intel post with NO links or possibility of being true, based on current information.

This is exactly why I say you are posting negatively in this thread.

 

[tamz_msc]

1) I won't get in in that, except to say as a former statistician, you are wrong.
2) You are wrong. It gains, but AMD gains more, I won't bother to link it as you will never admit its true.
3) This is a pro-intel post with NO links or possibility of being true, based on current information.

This is exactly why I say you are posting negatively in this thread.

1. Appeal to authority - the cheapest way out of an argument.
2. I have proved that Intel gains more with a reference but you will not accept it because you are part of the pro-AMD echo chamber.
3. Based on current information there is every possibility of it being true, but of course you will deny it because of point number 2.

 

[Carfax83]

CPU-bound can mean many things - as @Carfax83 pointed out even Doom Eternal is CPU-bound at 1080p, but it does not mean that the gap from Intel will be around the 20% mark every time in CPU-bound scenarios when other games are showing a 5% gap.

I think you should probably read my post again, because I was making the exact opposite claim; that Doom Eternal was GPU limited at 1080p. I was drawing a parallel between Dota 2 and Doom Eternal to show that a high FPS game can still be GPU bottlenecked at lower resolutions if the engine is very good at exploiting CPU parallelism, which Doom's id Tech 7 engine definitely is.

I'm surprised that AMD never thought to include it in their benchmarks, as it would have been perfect to demonstrate Zen 3's prowess assuming it would have beaten a 10900K; which I am almost certain it would have as id Tech 7 can scale across many threads.

 

[Kenmitch]

You could say that I'm downplaying AMD's marketing with respect to gaming performance, which is the logical thing to do and I see nothing wrong in that.

The logical thing to do is wait for the official reviews before coming to a conclusion. Maybe you like the taste of shoe?

 

[Zucker2k]

I'm surprised that AMD never thought to include it in their benchmarks, as it would have been perfect to demonstrate Zen 3's prowess assuming it would have beaten a 10900K; which I am almost certain it would have as id Tech 7 can scale across many threads.

Maybe they didn't because they didn't get the desired result?

 

[Zucker2k]

The logical thing to do is wait for the official reviews before coming to a conclusion. Maybe you like the taste of shoe?

He's not the only one speculating here. Why single him out? Or, we're all supposed to swallow AMD slides and jump on the bandwagon?

 

[JoeRambo]

I think this all discussion is a bit premature. We need ZEN3 chips in hands of all 3 categories:

1) Anandtech and co reviewers, so we can see how good inner core IPC is in things like SPEC and GB5, but since they are done with "stock" settings with "GTX 1080" at "8k", they are not really relevant for situations where CPU is bottleneck.
2) In hand of competent forum reviewers like legendary "the Stilt", so we can get proper pecking order in wide selection of apps
3) In hands of enthusiasts, that fine tune the memory on both systems to see what it is really capable off.

If you think memory is not important:

Jaka pamięć RAM do Intel Core i9-10900K? Test DDR4 2133-4000 MHz | PurePC.pl

Jaka pamięć RAM do Intel Core i9-10900K? Test DDR4 2133-4000 MHz (strona 27) Jaka pamięć operacyjna RAM do procesora Intel Core i9-10900K? Test pamięci DDR4 2133-4000 MHz na platformie LGA 1200 z procesorem Intel Core i9-10900K i HyperX Predator.

www.purepc.pl

Test pamięci do procesorów AMD Ryzen - Taktowania vs opóźnienia | PurePC.pl

Test pamięci do procesorów AMD Ryzen - Taktowania vs opóźnienia (strona 25) Test pamięci RAM DDR4 do procesora AMD Ryzen 5 3600. Porównanie wpływu taktowania i opóźnień na wydajność w grach i programach. Które ustawienia pamięci RAM są najlepsze?

www.purepc.pl

Anandtech setup is in the bottom of those graphs, in case you wonder

 

[coercitiv]

I pointed out three things-

1. That there are outliers in AMD's gaming performance numbers.
2. That Comet Lake gains a lot(more than Zen 2) with a memory overclock.
3. That Rocket Lake will end up faster in games if it can clock high enough.

You could say that I'm downplaying AMD's marketing with respect to gaming performance, which is the logical thing to do and I see nothing wrong in that.

It's not your downplaying AMD marketing that's the problem, there have been plenty of voices in this thread that declared they will wait for reviews before anything else, and they were accepted as such.

The problem is your selective approach towards data, exactly the one flaw you're accusing this entire thread of. When it comes to the new AMD product you're very quick to judge outliers and proclaim limited to non-relevant memory overclock results in gaming, yet when it comes to Rocket Lake you don't even require a preview to estimate performance gains and proclaim a winner.

You simply can't make up your mind between using objective data observation in the case of AMD marketing slides and using speculative estimates with no requirements for pre-existing data in the case of Rocket Lake and/or other Zen 3 performance metrics such as custom memory timing scaling. Your initial evaluation of Zen 3 gaming performance could have been entirely based on AMD marketing slides, with strong emphasis on probable high variance in gaming performance results, cautioning others to take into account a low average gaming performance lead of only 5%. Instead you chose to talk about overclocking and custom memory scaling on Skylake against a CPU you have no other data for.

With a modest all-core OC of 5.1GHz coupled with fast memory the 10900K will have no problems taking the gaming crown, and Rocket Lake will be comfortably ahead if it can clock to ~5GHz. So near yet so far.

"So near yet so far" - you judged before you measured.

 

[TitusTroy]

so the 5800X will be a single 8 CCX design and the 5900X will be 6+6?...so for gaming the single CCX design is more important then the higher cache on the 5900X (64 MB)?...in other words which is the better buy for gaming?

 

[DrMrLordX]

If AMD will not release the 5600 or any other cheaper 4C 8T ZEN 3 CPUs before or after Intel release its new RocketLake Desktop CPUs in Q1 2021 , they will have problems competing with the old Ryzen 2xxx and Ryzen 3xxx.

AMD doesn't seem to be too worried about it. It will be interesting to see what other SKUs they release as time goes on. Remember that Vermeer compatibility for 400-series boards won't come until January at the earliest. That may be a hint.

BUT WE WANT THE RETAILERS TO KEEP GETTING OUR MONEY AND NOT AMD! RAAWARG!

Nahhh. Etailer gouging will happen early on due to supply shortages, but over the long term (3 months out and later), prices will stabilize to MSRP.

Yeah I don't want them to gouge, just as I don't want AMD to raise prices. I knew the XT CPUs were a precursor to this increase in prices though. Didn't want to be right, but I was.

Intel is offering 6C below $200 both with and whiout gpus AMD needs something on there ASAP, once Rocket Lake launches Zen+ APUs and Zen 2 CPUs are a no-go.

See reply to AtenRa above. AMD probably isn't too worried. Those Rocket Lake CPUs are gonna be power hungry.

I like AMD's design changes, but these prices are bad.

They aren't awful, especially not for the 5900x. It'll be a beast of a workhorse CPU. 10900k doesn't even belong in the same conversation.

" Our CPUs need a nuclear reactor. AMD CPUs run on solar power and unicorn farts!" Lulz!

All true, and all hilarious.

 

[coercitiv]

so the 5800X will be a single 8 CCX design and the 5900X will be 6+6?...so for gaming the single CCX design is more important then the higher cache on the 5900X (64 MB)?...in other words which is the better buy for gaming?

We don't know yet, altough an argument can be made in favor of 5800X being more consistent at least. We need data.

 

[exquisitechar]

We don't know yet, altough an argument can be made in favor of 5800X being more consistent at least. We need data.

Robert Hallock said on Discord that the 5900X is the top gaming CPU, later said that the 5950X also equals it (although I think it was actually ever so slightly slower in AMD's comparisons? Not sure). I can also imagine the 5800X having an advantage in some cases, though. Either way, I'm sure the differences will be minute in general.

I think this all discussion is a bit premature. We need ZEN3 chips in hands of all 3 categories:

1) Anandtech and co reviewers, so we can see how good inner core IPC is in things like SPEC and GB5, but since they are done with "stock" settings with "GTX 1080" at "8k", they are not really relevant for situations where CPU is bottleneck.
2) In hand of competent forum reviewers like legendary "the Stilt", so we can get proper pecking order in wide selection of apps
3) In hands of enthusiasts, that fine tune the memory on both systems to see what it is really capable off.

If you think memory is not important:

Jaka pamięć RAM do Intel Core i9-10900K? Test DDR4 2133-4000 MHz | PurePC.pl

Jaka pamięć RAM do Intel Core i9-10900K? Test DDR4 2133-4000 MHz (strona 27) Jaka pamięć operacyjna RAM do procesora Intel Core i9-10900K? Test pamięci DDR4 2133-4000 MHz na platformie LGA 1200 z procesorem Intel Core i9-10900K i HyperX Predator.

www.purepc.pl

Test pamięci do procesorów AMD Ryzen - Taktowania vs opóźnienia | PurePC.pl

Test pamięci do procesorów AMD Ryzen - Taktowania vs opóźnienia (strona 25) Test pamięci RAM DDR4 do procesora AMD Ryzen 5 3600. Porównanie wpływu taktowania i opóźnień na wydajność w grach i programach. Które ustawienia pamięci RAM są najlepsze?

www.purepc.pl

Anandtech setup is in the bottom of those graphs, in case you wonder

I agree.

 

[DrMrLordX]

5800X may have half memory write bandwidth the same as the single-chiplet Matisse CPUs.

 

[majord]

I remember a time , not so long ago when AMD pricing too low would be taken as a sign they wouldn't deliver on performance , or there'd be catch (poor perf/watt, weak ST, or , or in the case of past Ryzen processors - not so competitive gaming performance) .

Perhaps instead of all the outrage over $50, people could see this a sign these CPU's mean business and will meet expectations with little in the way of 'strings attached'.

In all honesty I think the 'sticker' shock brought on by the previous generations pricing is the only thing preventing AMD from pricing these CPU's even higher. Whilst it's disapointing to see Perf/$ won't be pushed higher than their predecessors on at least ONE SKU -the 5600x ( something I think is important to do when introducting a new generation ) , it is still well ahead of the competition, and I think is AMD's justification here.

 

[coercitiv]

Robert Hallock said on Discord that the 5900X is the top gaming CPU, later said that the 5950X also equals it (although I think it was actually ever so slightly slower in AMD's comparisons? Not sure). I can also imagine the 5800X having an advantage in some cases, though.

Interesting. Yes, the limited benchmarks we saw on 5950X showed little difference between the two, with the advantage being in MT limited gaming benchmarks which are not quite representative for many games.

 

[inf64]

I cannot wait for AT Deep Dive and comparisons with Zen2 and Willow Cove, it will be awesome article I bet. 25 days seems like a lot though

 

[Abwx]

1. That Comet Lake gains a lot(more than Zen 2) with a memory overclock.
2. That Rocket Lake will end up faster in games if it can clock high enough.

No, because you are talking of gain in respect of the 2660Mhz official specs, but overclocked to the , CPU clock and RAM, it will gain 13% at best in respect of AMD slide where 3600MHz RAM was used for Intel as well...

Core i9-9900K vs. Ryzen 9 3900X mit RAM-OC in Spielen: Testergebnisse und Fazit

CPU-Spieleleistung im Test: Testergebnisse und Fazit / Testsysteme und Testmethodik / Getestete Spiele / So wurde getestet / Testergebnisse

www.computerbase.de

And something else, how do you know that CML gain more with RAM overclocking than Zen 3..?...

Dare posting some links..?.

 

[krumme]

The 19%
How? Is this a completely new arch?

 

[BTRY B 529th FA BN]

I hope the IF on the chip I get can do 1900. My 3800X can do it all day long with the kit I got.. such a pita and to think doing all that stress testing over again is kinda depressing.

 

[DAPUNISHER]

The 19%
How? Is this a completely new arch?

The doc has a youtube channel now. He gives his thoughts on it. And his subtle Rick Roll game is on point.

 

[inf64]

The 19%
How? Is this a completely new arch?

Well AMD set a goal that was to get a gain as if it were. They said that Zen3 is a back to back update on Zen(2). So it's not a radical departure like Bulldozer-> Zen was but still it's a massive upgrade to all parts of the core. They managed to keep power under control which is super impressive given the same node. This bodes well for Zen4 and even Zen5. I expect that Zen4 is 10-15% IPC uplift over Zen3 ( similar to Zen1+ ->Zen2) and Zen5 is ~20% jump from Zen4. This way they can also get a total of ~40% IPC uplift going from Zen3->Zen5, just like they did with Zen1->Zen3 and Excavator->Zen1(although that was overshot by quite a bit).