• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

AMD Zen 3 (5000 Series), rumored 17% IPC gain.

Status
Not open for further replies.
You probably know, but to be clear for anyone reading this that doesn’t, the 4000 series APUs are based on Zen2 so don’t really indicate what we can expect from the “proper” 4000 series CPUs.
Not at all. We can see how well the Zen 2 cores operate without the cross-CCX latency found on Matisse's chiplets (much like the 3300X). Since a large chunk of Zen 3's IPC increase is coming from moving to an 8-core CCX design, the monolithic APUs serve as a very nice indication of what a "proper" 4000 CPU can do.

AMD’s naming conventions are a travesty at this point!
How? You only think it is because, as an enthusiast, you're aware of the name "Zen" and therefore see the numbering discrepancy between the architecture codename and the product marketing name. But honestly, even then it's not that hard to keep track of.

If it has a "G" in it, it's old cores ;)
 
Disagree.

First of all I was replying to a post that said the clock speeds were disappointing... while I don’t expect big increase in clock for zen 3 these APUs have Little to no bearing on what we will see.

Secondly, cross ccx latency is only one part of the equation and even then as far as I can see this is actually still using 4 core ccx with 8 cores split across two CCXs that communicate over the infinity fabric as usual:

“AMD stated that the Zen 2 design in this chips follows the same CCX layout as the desktop hardware, which means the 8 cores are split into two CCX units which communicate over the internal infinity fabric.”

https://www.anandtech.com/show/1532...-apus-7nm-8core-on-both-15w-and-45w-coming-q1

Just because it’s monolithic doesn’t mean it changes how the cores are structured.

So no, I stand by the comment that we learn very little indeed from these chips as to the potential of zen 3.


The naming convention is simply confusing... look at the person I replied to - they didn’t know the 4000 APUs are architecturally a generation behind the CPUs. Look at the current 3000 series line up where from 3100 up you alternate jumps between different architectures. The average consumer would be extremely confused by how a higher numbered CPU just with a graphics chip performs worse than a lower numbered CPU without.

I am a big fan of AMD but they have made the lineup unnecessarily confusing to the average consumer and could do with a revamp. They still aren’t as bad as Intel but intel shouldn’t be the benchmark for naming!
 
I am a big fan of AMD but they have made the lineup unnecessarily confusing to the average consumer...

I agree, it is confusing. Unfortunately, this is standard practice for both Intel and AMD. OEMs want constant upgrades to help sell their hardware, so they demand new SKUs which end up just being the existing or last gen hardware, but with next gen naming. It's horribly anti-consumer and it should be done away with.
 
You're missing the point that the "average consumer" doesn't know what "Zen" is. How is a bigger thousand confusing? Is 4600G a bigger number than 3400G? Yes. Is it a better APU? Yes. So what's the problem? The letter "X" is ingrained in the public consciousness as something big, wow, impactful. So is "X" better than "G"? Yes. So what's the problem? Even without the "X", 3600 is better than 3400G, so what's the problem?

Is it confusing for us enthusiasts to keep track of architecture names vs product series? It has its moments for sure, but literally ignore "Zen", it means zero for the product and marketing names.
 
Alright well we can just agree to disagree on that front as it’s subjective... I think the naming is disingenuous and unnecessarily confusing. You didn’t reply to how it isn’t confusing to the average consumer that the CPU part of a 3100 outperforms a 3200G, but ultimately this is a circular argument and if you’re convinced that the naming is fine then great. I have a different opinion, but that’s what they are ultimately, opinions.

As for “missing the point” as you put it, what about your comments regarding the CCX?
 
Last edited:
I'll save you the bother then, it does :D

These chips will almost certainly have improved memory latency, but they still aren't representative (in either direction tbh) of an 8 core Zen3 "CCX" that doesn't need to communicate between 2 CCXs within a CCD and with different cache structures, but also with separate IO dies and potentially multiple CCDs... and the whole point was that these are not entirely indicative of what we can expect from Zen3 because put simply, they aren't Zen 3 in any way, shape or form.

I put CCX in quotes because I believe the CCX is effectively non existent in Zen3, it's just a unified 8 core CCD but it helps in comparison if anyone isn't familiar with CCX, CCD, IOD, WTF...
 
Last edited:
I'll save you the bother then, it does :D
Thank you.

I don't think there's been much rumoured or discussed about how (or if) the actual CPU cores are changing (I've not watched AdoredTV's recent "Integer monster" video). If the Zen 3 cores aren't much different than Zen 2 cores, then a good chunk of the IPC improvement is coming from dropping the split CCX design on the chiplets. Granted, there are still 2 CCXs in Renoir APUs, but if there's no physically separate I/O die to roundtrip through because it's a monolithic design, surely that is a pretty big indication of what Zen 3 can do when you don't have the inter-CCX latency to worry about? This was part of the excitement when we saw the 3300X's gaming benchmarks.

Of course, if the Zen 3 cores are significantly different to Zen 2 then just ignore everything I've said :D
 
You're missing the point that the "average consumer" doesn't know what "Zen" is. How is a bigger thousand confusing? Is 4600G a bigger number than 3400G? Yes. Is it a better APU? Yes. So what's the problem? The letter "X" is ingrained in the public consciousness as something big, wow, impactful. So is "X" better than "G"? Yes. So what's the problem? Even without the "X", 3600 is better than 3400G, so what's the problem?

Is it confusing for us enthusiasts to keep track of architecture names vs product series? It has its moments for sure, but literally ignore "Zen", it means zero for the product and marketing names.

i agree the naming schemes and what is what is confusing. needs to be simplier.
 
Yep I also agree the naming scheme is unnecessarily over complicated.

Just keeping the cores to the same 1000, 2000, 3000 series would be a start. Releasing zen2 cored apu’s right before zen 3 releasing and using the 4000 series is massively confusing bordering on deceptive.

IMO they are missing a trick waiting so long to produce and release the apu’s. Imagine if they release zen 3 cored apu’s just after zen 3 drops, Intel would be so far behind you would need a bribe to mention them in the same breath.
 
If the biggest thing people have got to complain about is the naming scheme then that is surely a good thing, no?

IMO they are missing a trick waiting so long to produce and release the apu’s.

Maybe they had no choice, given the 7nm monolithic design, and that they used fab space for the Zen2 chiplets for Ryzen 3xxx/Threadripper 3xxx/EPYC Rome first and foremost, then priority went to the laptop/low power parts which ship millions for than the desktop equivalents these days.

The next gen APU's may be later again, since it could be a total redesign with RDNA2 based GPU included in the die, and 4/6/8/10/12 CPU cores.
 
IMO they are missing a trick waiting so long to produce and release the apu’s. Imagine if they release zen 3 cored apu’s just after zen 3 drops, Intel would be so far behind you would need a bribe to mention them in the same breath.

Most likely just comes down to lack of resources... AMD is still a minnow compared to intel and until relatively recently was even more so. These are different to the desktop line (monolithic rather than chiplet) and undoubtedly require substantial R&D that AMD doesn’t, or at least didn’t, have the ability to do concurrently.

I expect as their income and budget grows they’ll be better positioned to put the newer arch into mobile sooner or even concurrently. Also perhaps they could move to a chiplet based design where they plop a GPU chiplet on, this reducing the rework required to make it work.


Of course, if the Zen 3 cores are significantly different to Zen 2 then just ignore everything I've said :D

As far as I’m aware we don’t yet know all the changes... Forrest described it as “an entirely new architecture” for what it’s worth, and of course we have the added complication of it being on a more enhanced node too.

I get what you are saying, yes potentially it gives some insight into what a lower latency can do, but it still isn’t enough to gain anything truly useful about Zen 3 (other than lower latency would be good, which I suggest is fairly obvious!) as we have no idea how it actually compares to what is a “completely new” architecture on a newer node. Will Zen 3 be higher latency than this? Lower? The same? What else has changed in the architecture? None of these questions are answered by the arrival of these chips because Zen 3 they ain’t....
 
Forrest described it as “an entirely new architecture” for what it’s worth
I think a lot of what was planned for the "entirely new" Zen 3 was pushed back to Zen 4 in the end. There will be changes to Zen 3 of course, but I think the radical changes have been held back a generation. Which is fine really, if Microsoft can't sort out a scheduler that works properly with Zen 2 chiplets, how in the blue hell would they handle SMT 4 :P
 
I think we both know the answer - "poorly" :eek:

Of course all that stuff about higher SMT and other changes was all based on rumours, leaks and speculation so whether it was ever truly planned for Zen3 at any point will probably forever be a mystery to us. These comments about "enitrely new architecture" were made back in January so I suggest it's pretty unlikely to have been altered significantly since then if they are planning on a release this year, but then it's also debateable what that comment actually means in reality too as it's hardly specific.
 
Status
Not open for further replies.
Back
Top Bottom