• 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.
Soldato
Joined
18 May 2010
Posts
22,297
Location
London
I will be very surprised if they are not backwards compatible as that's what was promised from AMD.

HOWEVER... I think it's up to the discretion of the motherboard manufactures to produce updated bioses for the previous gen boards.
 
Soldato
Joined
9 Nov 2009
Posts
24,765
Location
Planet Earth
I remember seeing that Linux was getting an update for a lot more microcode for Zen 3. If it's just a storage-space issue, there may be some B450's that can run the new chips. I guess we'll find out tomorrow.

My point is that it matters to me if a new MB will be required for arbitrary reasons or for engineering reasons. If the new CPU's require a bunch of new microcode to perform better, I would rather have new CPU's that use more microcode than some boards can hold rather than have AMD release less powerful CPU's just for expanded backward compatibility.

It's not some,it's almost all mainstream Zen2 buyers. Zen is an SOC,so in theory does not need a traditional chipset chip like Intel does. The chipset just is a PCI-E conversion hub,to add extra ports to the native ones in the SOC.

Far more people use B450 motherboards,than there are X570 motherboards,and plenty have 32MB BIOSes,which are the same as a X570 motherboard. So if 32MB is not enough then a lot of X570 owners are in trouble too. Mine is from 2018,and had a 32MB BIOS,and a proper 6 phase VRM,and a PCB with extra layers,and it was literally the same as the X470 version.

Personally I would be annoyed,but at least I bought it with a Zen+ CPU,and had usage before Zen2 was released so have two generations of CPUs,but so many have bought similar motherboards this year,so it's a dead end IF this happens!

I would rather AMD makes better compatibility,then a useless update which causes a lot of 2019/2020 B450/X470 and maybe X570 motherboards to be stuck with no upgrade path after one generation. Also Intel made that same excuse,that they needed new microcode for years hence a new socket,until ASRock made some weird motherboards which showed it was a load of tosh on their part.

AM5/DDR5 is probably due in 2022. So any new X670/B550 is going to be the last line of AM4 CPUs. Even if AM5 Zen4 can kind of work in those motherboards,what is the likelihood it actually gets gimped more using DDR4 than DDR5,if they double core counts??

If AMD made this move last year,it wouldn't be so bad,but to hold back until Intel has some new CPUs,2/3 through the life of Zen2 is a douche move. It affects the mainstream users who most likely want an upgrade path due to cost.

So I sincerely hope,this is a misconception by the reviewers hinting at this.

I will be very surprised if they are not backwards compatible as that's what was promised from AMD.

HOWEVER... I think it's up to the discretion of the motherboard manufactures to produce updated bioses for the previous gen boards.

Until those multiple rumours hit,I was certain it would be fine. But I don't know now,as if AMD orders OEMs not to have compatible BIOSes they won't do it,ie,PCI-E 4.0 BIOSes never appeared with later AGESA revisions for B450.

All speculation about mobo's until release, be nice if they were backwards compatible all back to x350, though.

I hope it is wrong speculation,by those reviewers!
 
Last edited:
Soldato
Joined
26 Sep 2010
Posts
7,146
Location
Stoke-on-Trent
Jesus Christ, is this still going on?

I don't give a rat's ass how cheap that Chinese B550 looks, if it's PCIe 4 compliant then it has a thicker PCB to carry all the extra traces. You can huff and puff all you want, but if you don't have decent traces and a PCB to support it then you're not getting reliable PCIe 4. That's why AMD pulled PCIe 4 for 300 and 400 boards. Nothing to do with the chipset, it was PCB quality. End of. Now could AMD have let board vendors do unofficial BIOS updates to enable PCIe 4? Yes, they could have. but then tell me one board vendor who is going to kill consumer incentive to purchase an upgrade if they enabled PCIe 4 where possible. And open up a massive can of worms with customer support if some of their boards support PCIe, some don't, and some top slots aren't actually 16x slots.
 
Soldato
Joined
26 Sep 2010
Posts
7,146
Location
Stoke-on-Trent
And there are no retimers on TRX40 either. Combined with the repurposed Zen IO die, it's looking like X570 was a rushed hatchet job that has since been refined.

Why do you keep harping on about chipset? Retimers has nothing to do with the chipset, it's about boosting PCIe signals. And with a good enough board design, you don't need retimers to get PCIe 4 on the slot closest to the CPU and the CPU-attached M.2. That's what was planned with 300 and 400 but ultimately didn't work because it was too erratic.

Stop with the conspiracy theories. I bet you a fiver that if AMD did enable PCIe 4 on 300 and 400 boards and allowed the patchwork, **** show of incompatibility and fragmented support you'd be bitching about it. AMD could not guarantee consistency of PCIe 4 support so they pulled it, why is that so hard for you to accept? And yes, a monster Asus board is unlikely to have a flimsy PCB compared to that SOYO unit, but if the PCIe traces were only designed for Gen 3 then you're just not going to get Gen 4 out of it. Power delivery and PCIe signalling are not the same thing. AND this is Asus, there is no way in hell they'd give customers free upgrades. The second AMD said "hang on, PCIe 4 might not be viable across the board" you just know Asus will go "fine, no Gen 4 then. Nope, nope, none".

And as for B550A, has it occurred to you that, despite being B450 chipset, the board as a whole was designed to support PCIe 4 in limited capacity? It's nothing to do with chipset for crying out loud, it's board design.

Right now I just have no idea what you're crying about. I've read your post 3 times to make sense of it and you're just making random accusations and using your own interpretations of things I never even insinuated as some weird defence.

It's Thursday now, we'll see what happens. Try not to burst a blood vessel if it turns out B450 is locked out, it really doesn't matter. I know that I won't be getting flaming pitchforks and attacking AMD for the first time in about 10 years they've done a single generation motherboard.

"Fall on their sword for AMD" lol such overly dramatic rubbish.
 
Caporegime
Joined
17 Mar 2012
Posts
47,381
Location
ARC-L1, Stanton System
Well... here we go.

Steve from GN spent a lot of time not long ago constantly banging on about how AMD's CPU TDP claims are "Fake", even made a dedicated video about it, and really oddly never said anything about Intel's utterly ridiculous TDP ratings, which as far as "Fake TDP' Claims" go are vastly worse than AMD's.

So here he is with Blender power consumption chart and i'm seeing something very familiar here, a while ago a bunch of reviewers were claiming Coffeelake Power consumption figures just under or at Intel's TDP ratings, really odd that given we know Intel's CPU run at least double the TDP ratings.

Turns out what they were doing was locking the CPU to TDP in the BIOS or switching to TDP locked Motherboards after performance tests, the problem with that is locking Intel CPU's to TDP drastically reduces performance, so they were performance testing the 9900K at 200 Watts and then locking it to 95 Watts limit for power consumption testing.

What does this have to do with Steves AMD TDP rants while ignoring Intel's far worse?

He's locking the Intel CPU's to TDP.

What a ######.....

This matters because Intel are actively claiming their CPU's are class leading power efficient.

Watch out for this sort of crap when he starts testing the 10K Series.....

Mt52Lzx.png


This slide shows the power difference in Blender vs locking the 9900K to TDP and not...

phHms5i.png


 
Soldato
Joined
9 Nov 2009
Posts
24,765
Location
Planet Earth
@humbug Good observation there. It happened with the 6 core CFL generation too,in OEM systems,which were TDP locked. Performance wasn't as good! But I was reading the AT review of the new Zen2 based Acer Swift 3 ultrabook(1.2KG),and AT complained of throttling. But they compared it to IceLake 15" laptops which were 1/3 heavier,and had more cooling for half the cores. They only used the equivalent size and weight XPS13 in a limited number of tests.
 
Last edited:
Soldato
Joined
6 Aug 2009
Posts
7,070
Jesus Christ, is this still going on?

I don't give a rat's ass how cheap that Chinese B550 looks, if it's PCIe 4 compliant then it has a thicker PCB to carry all the extra traces. You can huff and puff all you want, but if you don't have decent traces and a PCB to support it then you're not getting reliable PCIe 4. That's why AMD pulled PCIe 4 for 300 and 400 boards. Nothing to do with the chipset, it was PCB quality. End of. Now could AMD have let board vendors do unofficial BIOS updates to enable PCIe 4? Yes, they could have. but then tell me one board vendor who is going to kill consumer incentive to purchase an upgrade if they enabled PCIe 4 where possible. And open up a massive can of worms with customer support if some of their boards support PCIe, some don't, and some top slots aren't actually 16x slots.

Agreed, if the reason for the extra cost of X570 was PCIE 4.0 support through the use of higher quality PCBs I can see that trying to enable it on B450 would have opened a real can of worms. Can you imagine the commotion if there were lots of RMAs for defective motherboards?

@humbug interesting observation. You should publicly challenge their TDP testing. GN take great pride in their methodology, it would be interesting to see what their response was.
 
Soldato
Joined
15 Oct 2003
Posts
14,742
Location
Chengdu
humbug, you really need to lose the chip on your shoulder.
There is no agenda at GN. BS is called when it is witnessed, towards any company.

Edit: I'm pretty sure there was a video on Intel TDP limits also. Didn't it call out the mobo manufacturers, but also say Intel knew fine well it was going on.
 
Caporegime
Joined
17 Mar 2012
Posts
47,381
Location
ARC-L1, Stanton System
@humbug your two charts are comparing two different readings. One is overall system load power draw, the other is measuring the EPS 12v rail.

The CPU runs off the 12v EPS rail, its not as if there is anything else to measure that would give you a different reading, yes they are measuring the CPU alone, as opposed to System, IE from the wall.

I made it clear the chart that i'm using illustrates the difference in power consumption when you lock the CPU to TDP as opposed to the way it would actually run if you didn't touch it, if you just installed it and left it, its written on the chart (95Watt) vs unlimited
That chart is from the original 9900K review, while a lot of reviewers just locked the CPU to 95 Watt for power consumption testing and left it at that, the author of that review (Steve from hardware unboxed) pulled those reviewers up on it and deliberately put the power consumption difference on the slide.

humbug, you really need to lose the chip on your shoulder.
There is no agenda at GN. BS is called when it is witnessed, towards any company.

Edit: I'm pretty sure there was a video on Intel TDP limits also. Didn't it call out the mobo manufacturers, but also say Intel knew fine well it was going on.

Do you actually understand what is going on here?

It would be like me performance testing my 3600 in CB R20 and getting a score of 3800 and then run a power consumption test also in CB R20 but going into my BIOS and locking the CPU to 20 watts beforehand to claim in the power consumption results that the CPU only uses 20 Watts, in which case it would, what i'm not telling you is if i did that CB R20 run again it would only score about 2100.

If i did that you would be all over me with criticisms and you would be right.
 
Caporegime
Joined
17 Mar 2012
Posts
47,381
Location
ARC-L1, Stanton System
@CAT-THE-FIFTH watch that video you posted, there is nothing real in the way of criticism about the way Intel calculate TDP, just an explanation of how it works, more a justification.

Intel base their TDP off PL states, the lowest being PL1 which is 3.6Ghz, that's how they get to 95 Watts, if you put the CPU in a really bad motherboard it will be locked to PL1 and its performance would be far less than a higher end Motherboard that didn't lock it to PL1

AMD are no different, my CPU has a 65 Watt TDP, also calculated from a PL1 3.6Ghz state, its actually BIOS locked to 88 Watts allowing it to boost to 4.2Ghz, i can set that higher or lower in the BIOS or Ryzen Master.

The difference here is while both AMD and Intel do exactly the same thing Intel is far worse for it, the 3600 uses about 80 Watts in Blender with a claimed 65 Watt TDP, the 9900K about 170 Watt with a claimed TDP of 95 Watts and Steve is not locking any Ryzen CPU to TDP in his power consumption tests.

TDP is bull crap, AMD could just put 20 Watts TDP on the box and that would be right, so long as that's what it is in PL1 state, Intel's 10K series on the face of it have lower power consumption, they put lower or the same TDP with higher clocks / cores counts vs Coffeelake on the slides, do you believe that the 10970 is really a 65 Watt CPU?

What i'm watching for is reviewers saying that it is, much like Gamers Nexus here is saying the 9100F is a 48 Watt CPU and the 9900K a 93 Watt CPU.... pop-weasel.
 
Last edited:
Soldato
Joined
27 Mar 2009
Posts
3,290

Im gutted as had a B450 board waiting for an upgrade. On the flip side AMD must feel they have a real winner with Zen3 to upset such a large amount of customers but feel confident they will buy anyway. As many have already realised its time to stop thinking of AMD as this do good for the consumer brand as accept they are only looking after shareholders.

If in 6 months time modders get Zen3 working on B450 with just PCI 3.0 instead of 4.0 then like Intel, AMD should face all the bad PR they deserve
 
Status
Not open for further replies.
Back
Top Bottom