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

[Mopetar]

I don't appreciate the accusation of forming a conclusion based on an assumption as if I was refusing to watch the video. I tried to watch the video multiple times, it wouldn't load so I was relying on those who had watched it, hence why I was asking questions to try and get more information.

No offense meant, but it just felt like I was having that kind of an argument. At the time I was exceptionally under-caffeinated and not in the mood, but having obtained some coffee my disposition is much less like an irritable dragon than it was previously.

With that said, I can watch the video this morning and don't even see where they tested Intel at all in the video. Granted, I skipped through it due to time, but I didn't see any charts where they tested Intel comparatively.

They only did it for a small set of titles at the tail end of the video which should be timestamped to that point:

That's why I said it's probably not conclusive or that their results are more limited.

TPU has new results now so it's probably best just to read through their new write-up now and see what they found.

The newest TPU results tested a lot of things, but not the differences between 2 and 4 sticks of RAM. If they do run a full set of tests for both Intel and AMD using an x4 memory configuration it should answer the question definitively either way. As I said before, maybe there is some other explanation for why everyone else saw a bigger performance gap, but the best data available at the time pointed to that being a possible culprit.

 

[wahdangun]

I do wonder if the dissapearing of the strange performance dip (where Zen3 falls just a little bit behind Intel when going to higher resolutions) as is seen with Turing but not at all with Ampere is the advantage of twice the CPU to GPU communication speed of Ampere with PCIe4.


TURING with PCIe3:

View attachment 33709

AMPERE with PCIe4:

View attachment 33710

A graphics driver that strives for the lowest possible latency between user input and screen output will have the CPU to GPU communication in the critical path, even though the CPU to GPU communication takes less then a frame time.

I'm not sure, it's a complex dynamic with many factors but it can be tested by limiting PCIe4 to PCIe3.

Example:
GREEN/BLUE/GREY = TURING
RED/BROWN = AMPERE
View attachment 33712

that the definition of GPU bottleneck as you can see when the GPU performance increased the gap widening,

 

[Asterox]

Ok, no need for further comment.

 

[PhonakV30]

I feel this Ryzen 5600X is an anomaly in PC world.The only 6 cores CPU that can surpass 10900K.

 

[coercitiv]

I feel this Ryzen 5600X is an anomaly in PC world.The only 6 cores CPU that can surpass 10900K.

We did have a 6-core Intel CPU which surpassed an 8-core AMD CPU a few years ago.

 

[Mopetar]

I feel this Ryzen 5600X is an anomaly in PC world.The only 6 cores CPU that can surpass 10900K.

Well just wait until the 5600 comes out.

 

[Ajay]

I feel this Ryzen 5600X is an anomaly in PC world.The only 6 cores CPU that can surpass 10900K.

Taking a quick look at Borderlands 3 running on my PC - it seems that there are only 6 threads doing almost all the work. And those threads aren't even maxing out the CPU cores they use.

 

[lobz]

Taking a quick look at Borderlands 3 running on my PC - it seems that there are only 6 threads doing almost all the work. And those threads aren't even maxing out the CPU cores they use.

'Welcome to my world', says the 1800X to the 10900K

 

[CakeMonster]

Many people get confused about threads, I can have task manager open in WoW to AC:O to HZD on my 9900K and see that since there's 20-70% activity on every thread (and in some cases 1 thread on 100%), that my CPU is holding back the game. But reading benchmarks, everything indicates that its GPU limited.

I'm guessing a lot of people are fooled to think they need 10 or 12 cores because of looking at that utilization that put some load on every available thread.

 

[software_engineer]

Regardless of whether one thinks that the price of the Ryzen 5600X is too high, I don't think it can be taken for granted than AMD will release a Ryzen 5600 or a Ryzen 5700/5700X outside of the OEM market. Let's take a look at the history of these product numbers throughout the various Ryzen generations.

For Ryzen 1000 all of the following models were available to consumers:

• Ryzen 1600 (65W TDP)
• Ryzen 1600X (95W TDP)
• Ryzen 1700 (65W TDP)
• Ryzen 1700X (95W TDP)

For Ryzen 2000 all of the following models were available to consumers:

• Ryzen 2600 (65W TDP)
• Ryzen 2600X (95W TDP)
• Ryzen 2700 (65W TDP)
• Ryzen 2700X (95W TDP)

However things changed a bit for Ryzen 3000, where the TDP of Ryzen 3700X dropped to 65W. The following models were available to consumers:

• Ryzen 3600 (65W TDP)
• Ryzen 3600X (95W TDP)
• Ryzen 3700X (65W TDP)

Presumably because of power efficiencies gained from the move from the Global Foundries 12nm process to the TSMC 7nm process, there was no longer a place for both a Ryzen 3700 and Ryzen 3700X in the product lineup in addition to the Ryzen 3800X. I'm not sure why the part was not labeled Ryzen 3700 instead of a Ryzen 3700X. In the case of the Ryzen 3900X (105W TDP), their was a Ryzen 3900 (65W) part made available to OEM customers.

Returning to the current Ryzen 5000 series, the following models are currently available to consumers:

• Ryzen 5600X (65W TDP)
• Ryzen 5800X (105W TDP)

If AMD were to follow a similar strategy to the one that they followed for the Ryzen 3000 series in respect to the Ryzen 3700X, one might predict that we might not see the release of a lower clocked Ryzen 5600 (65W TDP) part to consumers. However we might see the release of a Ryzen 5600 part to OEM customers. There does still seem to be a place in the lineup for a Ryzen 5700X (65W TDP) part, at least from a TDP point of view.

 

[uzzi38]

Honestly the more graphs and the like I see on Zen 3 the less I think Rocket Lake will offer meaningful competition and the even less I think AMD will ever actually launch a 5600 and/or 5700X.

Between Gamer's Nexus showing Zen 3 besting highly tuned Intel systems running high all core OCs and highly tubed memory just at stock with DDR4-3200cl14 memory and TechPowerUp's testing at differing levels of GPU bottlenecks showing that with no GPU bottleneck at all Zen 3 was up to 20% faster than Comet Lake I really don't see AMD feeling pressured by Rocket Lake in the slightest.

With that in mind... I can't help but feel we'll just see the 5600X settle down in price to $249, the 5800X settle down to $399 and AMD just releasing XT SKUs at the original prices just like Zen 2.

 

[moinmoin]

Honestly the more graphs and the like I see on Zen 3 the less I think Rocket Lake will offer meaningful competition and the even less I think AMD will ever actually launch a 5600 and/or 5700X.

There are many chips AMD could have avoided launching with that line of thinking, the last one was 3100. AMD will fill the product range over time, not because competition forces it to but because it can. More telling will the amount AMD allocates for each of them, as 3300X has shown (3100 is readily available, 3300X is like a ghost).

 

[uzzi38]

There are many chips AMD could have avoided launching with that line of thinking, the last one was 3100. AMD will fill the product range over time, not because competition forces it to but because it can. More telling will the amount AMD allocates for each of them, as 3300X has shown (3100 is readily available, 3300X is like a ghost).

That's a fair point actually, yeah from just the optics perspective of filling out the gaps in the lineup that makes sense.

 

[exquisitechar]

der8auer tried the Intel Cryo Cooler with a 5950x.

 

[zinfamous]

Honestly the more graphs and the like I see on Zen 3 the less I think Rocket Lake will offer meaningful competition and the even less I think AMD will ever actually launch a 5600 and/or 5700X.

Between Gamer's Nexus showing Zen 3 besting highly tuned Intel systems running high all core OCs and highly tubed memory just at stock with DDR4-3200cl14 memory and TechPowerUp's testing at differing levels of GPU bottlenecks showing that with no GPU bottleneck at all Zen 3 was up to 20% faster than Comet Lake I really don't see AMD feeling pressured by Rocket Lake in the slightest.

With that in mind... I can't help but feel we'll just see the 5600X settle down in price to $249, the 5800X settle down to $399 and AMD just releasing XT SKUs at the original prices just like Zen 2.

If Rocketlake ends up underperforming the Zen3X series, but at a decent price, that puts AMD in the position to then release the non-X series and still match/beat their tiers with respect to Intel...and also even potentially beat them on price, just gobbling up more and more of whatever market share Intel is praying to retain this generation.

The tides have turned!

 

[AtenRa]

Well Rocketlake is a 14nm back ported TigerLake @ 10nm used in Laptops.

Tiger lake brings up to 20% IPC performance increase over last gen.
It will not be far fetched to expect Rocket lake to be able to compete against Zen 3 in gaming performance.
Back porting it to 14nm will maintain 5GHz+ clock speeds. Only downside will be higher power consumption (125W TDP)

And since Intel always release lots of SKUs, im expecting cheaper 6Core and 8Core ZEN 3 SKUs from AMD close to Intel RocketLake release.

 

[DrMrLordX]

Well Rocketlake is a 14nm back ported TigerLake @ 10nm used in Laptops.

It's actually more related to Sunny Cove. We'll have to wait for the final product to be sure, but so far it doesn't look like Rocket Lake gets +20% IPC over Skylake.

 

[amrnuke]

Well Rocketlake is a 14nm back ported TigerLake @ 10nm used in Laptops.

Tiger lake brings up to 20% IPC performance increase over last gen.
It will not be far fetched to expect Rocket lake to be able to compete against Zen 3 in gaming performance.
Back porting it to 14nm will maintain 5GHz+ clock speeds. Only downside will be higher power consumption (125W TDP)

And since Intel always release lots of SKUs, im expecting cheaper 6Core and 8Core ZEN 3 SKUs from AMD close to Intel RocketLake release.

I can't wait to see this competition! If Intel reclaim the gaming crown, even at a power cost, consumers will end up winning regardless.

 

[ondma]

It's actually more related to Sunny Cove. We'll have to wait for the final product to be sure, but so far it doesn't look like Rocket Lake gets +20% IPC over Skylake.

Ice Lake was generally accepted to give ~18% over Skylake, no? And Tiger lake gave little if no further IPC over ice lake. I think some in these forums even claimed a slight regression.

 

[DrMrLordX]

Ice Lake was generally accepted to give ~18% over Skylake, no?

So far it doesn't look even that good, though leaks are limited to some GB scores and CBR20 scores. Not comprehensive. We'll need to see the actual retail parts.

 

[Atari2600]

Well Rocketlake is a 14nm back ported TigerLake @ 10nm used in Laptops.

Tiger lake brings up to 20% IPC performance increase over last gen.
It will not be far fetched to expect Rocket lake to be able to compete against Zen 3 in gaming performance.
Back porting it to 14nm will maintain 5GHz+ clock speeds. Only downside will be higher power consumption (125W TDP)

They'll never hit that 20%.

If Intel have to sacrifice cache (or other functionality) in order to make it economically manufacturable, then the entire design can be put out of balance. Think Barcelona vs. Thuban.

If they don't sacrifice something, i.e. clock rates, they'll blow straight through their power budgets.

It also would require the TigerLake design to be laid out for hitting high clocks. Intel may not have been thinking along those lines when starting the design work.

 

[krumme]

Makes no sense to me this zen3 is in anyway similar to zen2. Its 20% ipc, a tad more fmax, more efficient and yet only like 12% larger. 12%! Its counter intuitive. It ought to be 120% larger.
What similar examples have we?
I mean at high level this might be similar to zen2 but it looks more clean sheet from a result perspective than anything i remember.

 

[amrnuke]

Makes no sense to me this zen3 is in anyway similar to zen2. Its 20% ipc, a tad more fmax, more efficient and yet only like 12% larger. 12%! Its counter intuitive. It ought to be 120% larger.
What similar examples have we?
I mean at high level this might be similar to zen2 but it looks more clean sheet from a result perspective than anything i remember.

Yep. Papermaster said it was a complete redesign, he said from way above it might look similar, but no area of the design/uarch was untouched.

 

[Mopetar]

Makes no sense to me this zen3 is in anyway similar to zen2. Its 20% ipc, a tad more fmax, more efficient and yet only like 12% larger. 12%! Its counter intuitive. It ought to be 120% larger.
What similar examples have we?
I mean at high level this might be similar to zen2 but it looks more clean sheet from a result perspective than anything i remember.

I suspect moving to an 8-core CCX over using a two 4-core setup as in the past resulted in a lot of transistors saved. Having everything as part of a single CCX means anything that had to be duplicated for both essentially became redundant.

Sometimes optimization can result in using less silicon as well as bringing better performance. A fresh design can do this because it isn't trying to build on top of existing silicon so it isn't constrained with keeping things as they are while bolting on additional functionality or creating additional paths for special cases.

 

[CakeMonster]

Are there any real world scenarios where the lower base clock of the 5950X could be a disadvantage over the base clock of the 5900X?

After all the difference is 3.7 (5900X) vs 3.4 (5950X), while the boost difference is 4.8 (5900X) vs 4.9 (5950X).