AMD Announces Ryzen 7000 Series - Launching This Fall

[BiG StroOnZ]

Hmmm, now this is getting interesting.

 

AMD Corrects Socket AM5 for Ryzen 7000 Power Specs: 230W Peak Power, 170W TDP ~

 

Quote

AMD would like to issue a correction to the socket power and TDP limits of the upcoming AMD Socket AM5. AMD Socket AM5 supports up to a 170W TDP with a PPT up to 230W. TDP*1.35 is the standard calculation for TDP v. PPT for AMD sockets in the “Zen” era, and the new 170W TDP group is no exception (170*1.35=229.5).

 

https://www.tomshardware.com/news/amd-corrects-socket-am5-for-ryzen-7000-power-specs-230w-peak-power-170w-tdp

[porina]

9 minutes ago, BiG StroOnZ said:

AMD Corrects Socket AM5 for Ryzen 7000 Power Specs: 230W Peak Power, 170W TDP ~

Was just going to post that but you got there quicker!

 

AMD - nani the f...!

 

The communication from Robert Hallock (AMD's Director of Technical Marketing)* AMD to Ian Cutress (formerly of Anandtech) was that 170W figure was PPT. The keynote presentation presenter said TDP. The slide just said "W". Now Tom's Hardware's unnamed AMD contact issued that "clarification".

 

We're going to need a clarification to the clarification. One speculation is that it may be currently planned early products will be 170W PPT, but future products may be 170W TDP.

 

* I just rewatched that part of Ian's stream again. He said it was from an AMD representative previously, and he would have to poke Robert Hallock to check that.

[KaitouX]

17 minutes ago, porina said:

Was just going to post that but you got there quicker!

 

AMD - nani the f...!

 

The communication from Robert Hallock (AMD's Director of Technical Marketing)* AMD to Ian Cutress (formerly of Anandtech) was that 170W figure was PPT. The keynote presentation presenter said TDP. The slide just said "W". Now Tom's Hardware's unnamed AMD contact issued that "clarification".

 

We're going to need a clarification to the clarification. One speculation is that it may be currently planned early products will be 170W PPT, but future products may be 170W TDP.

 

* I just rewatched that part of Ian's stream again. He said it was from an AMD representative previously, and he would have to poke Robert Hallock to check that.

Maybe the one they shown is a 16-core with 170W PPT, but they have/plan another SKU on top of that with 230W PPT and higher clocks, like Intel have the 12900KS?

Or it's just for future 24 or 32 core parts with Zen 5 or later.

[sounds]

AMD is doing amazingly well. They have a fierce competitor - Intel - and PCIe gen 5 is still a tough beast to tame. DDR5 seems to have stabilized, other than that retail prices are still high.

 

The Socket AM5 specs pass muster. The Zen4 CPUs are ok, even though they don't really impress me. The X670/X670E/B670 boards will be fine. Intel is king again for this round.

[tim0901]

2 hours ago, LAwLz said:

I think AMD will be more efficient as well. 

Maybe not as efficient as they are now, but still better than Intel. 

 

Personally, I don't think efficiency or power consumption matters that much, as long as it is somewhat reasonable and the performance is good. 

It's only when a product performs worse and still uses more power than it hink its worth bringing up. 

A product performing 20% better and uses 30% more power? Who cares... 

A product performing 20% worse while still using 30% more power? That's embarrassing. 

 

 

Of course I'm taking about desktop products here. With laptop and other battery powered things efficiency is important. 

 

 

I also think people stare too blindly at peak power consumption when making comparisons. Power consumption at medium loads are arguably even more relevant, if we are talking about things like laptops. Most people rarely if ever fully load their system. 

Completely agreed - power consumption really doesn't matter much on desktop. With cost of power skyrocketing in many parts of world it might matter a bit more than usual, but it still isn't a significnant factor. And it's very true about peak power consumption being a rarity - would be interesting to see better metrics for measuring efficiency at more standard workloads. But given how poorly most outlets measure power consumption already I don't have much faith in that coming to fruition any time soon (unless Steve wants to take a crack at it).

 

2 hours ago, KaitouX said:

Power consumption might not be that great of a deal with Raptor Lake, at least if Intel doesn't do what they've done to the 12900K, and increased the power consumption in almost 100W for less than 10% more performance, at 190W it's less than 5% difference. The 12600K and 12700K are already basically as efficient as the 5600X and 5800X, while the 12700 is similar to the 5900X, they don't need much to compete in efficiency, only the 5950X and it's successor are probably going to stay isolated as the most efficient consumer CPUs in MT tasks.

True - the lower end Intel chips are very much more efficient. But who knows if that will change as their e-core numbers are pumped up. I was also thinking mobile where - as far as I'm aware - Intel still has some catching up to do in terms of efficiency compared to AMD or Apple, although I'll fully admit that I don't follow that space as much and could be wrong.

[Mark Kaine]

So in real world performance,  and im quoting here, this looks 

Quote

disappointing 

(again)

 

 

 

even a measly 12400F *without AVX* might mop the floor with this (which is good because a 12400 F or even 11th gen should be cheap by then ~)

 

looks 13th gen will really eat 7000 for breakfast? 

 

Maybe AMD can turn it around with 3d cache, but it all does look mightily unexciting and honestly not what i was expecting or hoping for...

 

 

[tim0901]

25 minutes ago, KaitouX said:

Maybe the one they shown is a 16-core with 170W PPT, but they have/plan another SKU on top of that with 230W PPT and higher clocks, like Intel have the 12900KS?

Or it's just for future 24 or 32 core parts with Zen 5 or later.

I think future proofing the socket makes a lot of sense - it might encourage board vendors to build better boards as well, preventing the issues we saw around Zen 2 where early Zen 1 motherboards weren't built well enough to handle the newer chips, despite otherwise fully supporting them.

[LAwLz]

1 hour ago, BiG StroOnZ said:

Hmmm, now this is getting interesting.

 

AMD Corrects Socket AM5 for Ryzen 7000 Power Specs: 230W Peak Power, 170W TDP ~

 

 

https://www.tomshardware.com/news/amd-corrects-socket-am5-for-ryzen-7000-power-specs-230w-peak-power-170w-tdp

SIgh... AMD giving different answers to different people. Ian said he specifically asked AMD about this and they said PPT, not TDP.

It's one thing to say something wrong and then correct it. It's another to give several different answers to different people. AMD should have handled this way better. 

 

 

 

 

1 hour ago, porina said:

* I just rewatched that part of Ian's stream again. He said it was from an AMD representative previously, and he would have to poke Robert Hallock to check that.

Robert Hallock also said it was PPT, not TDP. He even clarified it himself in this video with PCWorld at 4:00:

 

Right now I will believe in the written clarification that was went to Tom's Hardware over what was communicated verbally.

It is more likely that something written is correct over something that was said in an interview.

[porina]

57 minutes ago, LAwLz said:

Right now I will believe in the written clarification that was went to Tom's Hardware over what was communicated verbally.

It is more likely that something written is correct over something that was said in an interview.

It depends who wrote it?

 

I'm now wondering why didn't they make it clearer on the original slide. It would have been reviewed before use and made it unambiguous. There may be an argument that most communications do refer to TDP, not PPT, so TDP could be implied that way. I don't have time right now, who from AMD gave the presentation? I'd have to assume they would be pretty senior to appear in a keynote, although it remains a possibility they're just reading what was given to them.

 

One scenario that remains is everyone is right, but they're talking about slightly different things. That would need further clarification. 

 

And I'll throw in a tin foil hat scenario. This is intentional confusion to maximise talk about the product!

[Quackers101]

2 hours ago, porina said:

And I'll throw in a tin foil hat scenario. This is intentional confusion to maximise talk about the product!

nononono, its for "AMD is losing to intel on every front?!" clickbait (before its real).

[BiG StroOnZ]

5 hours ago, porina said:

Was just going to post that but you got there quicker!

 

AMD - nani the f...!

 

The communication from Robert Hallock (AMD's Director of Technical Marketing)* AMD to Ian Cutress (formerly of Anandtech) was that 170W figure was PPT. The keynote presentation presenter said TDP. The slide just said "W". Now Tom's Hardware's unnamed AMD contact issued that "clarification".

 

We're going to need a clarification to the clarification. One speculation is that it may be currently planned early products will be 170W PPT, but future products may be 170W TDP.

 

* I just rewatched that part of Ian's stream again. He said it was from an AMD representative previously, and he would have to poke Robert Hallock to check that.

 

To quote the Tom's Hardware article:

 

Quote

The increased 170W TDP also means that we could see souped-up 12- and 16-core Ryzen 7000 chips with a 170W TDP for extreme users, while 105W 12- and 16-core models slot in for more mainstream uses.

 

This new tier of 170W Ryzen chips could be AMD's attempt to create a new quasi-HEDT tier of chips on its standard mainstream PC platforms, which would make a lot of sense given that Threadripper has largely disappeared from retail in favor of pricier Threadripper Pro models for professional users. Quad-channel memory bandwidth was one of the key reasons to jump up to a Threadripper platform, but dual-channel bandwidth from DDR5 on the AM5 platform could be potent enough for the semi-professionals and content creators that AMD targets with the standard non-Pro Threadripper lineup

 

Increasing the TDP and PPT will help AMD deliver more performance, particularly for its higher core-count models, during heavy multi-threaded workloads. In many cases, AMD's previous limit of 142W with the previous-gen AM4 socket held back performance, so the additional 88W of power will be particularly helpful with the newer 12- and 16-core models.

 

I like this surmising the most, as it is in line with what to expect for the increase in / uncertainty about the finalized TDP/PPT. It leaves room for future products to exist that have more power: either through higher clock speed or more cores or both.

 

3 hours ago, LAwLz said:

SIgh... AMD giving different answers to different people. Ian said he specifically asked AMD about this and they said PPT, not TDP.

It's one thing to say something wrong and then correct it. It's another to give several different answers to different people. AMD should have handled this way better. 

 

Tom's updated the article at the bottom (might slightly clear up some of the confusion):

 

Quote

Update: We asked AMD about the TDP and PPT values that the company used for its Computex 2022 demo against the Core i9-12900K. AMD tells us:

"The Computex demo utilized a 16-core pre-production sample not yet fused to specific power values, but was operating below our final 170W TDP spec."

Naturally, that doesn't tell us if the demo processor consumed 50W below the 170W spec, or just one single watt below the spec. However, that's the information the company has to share on the matter. We'll update if we learn more. 

[porina]

32 minutes ago, BiG StroOnZ said:

I like this surmising the most, as it is in line with what to expect for the increase in / uncertainty about the finalized TDP/PPT. It leaves room for future products to exist that have more power: either through higher clock speed or more cores or both.

The big problem with that is they're not saying AMD said any of those things. It is their speculation on the possibilities, no better than what we're doing in this thread.

[BiG StroOnZ]

9 minutes ago, porina said:

The big problem with that is they're not saying AMD said any of those things. It is their speculation on the possibilities, no better than what we're doing in this thread.

 

Yes I agree, it is their speculation (Tom's Hardware), and it could be a number of reasons why the TDP/PPT increased. Could be simply Zen 4 is power hungry for all we know. 

[porina]

1 minute ago, BiG StroOnZ said:

Yes I agree, it is their speculation (Tom's Hardware), and it could be a number of reasons why the TDP/PPT increased. Could be simply Zen 4 is power hungry for all we know. 

I think we've gone over the logical and even less logical possibilities. Can only wait and see what really happens. It's going to be interesting towards the end of the year one way or other.

[Blademaster91]

1 hour ago, BiG StroOnZ said:

I like this surmising the most, as it is in line with what to expect for the increase in / uncertainty about the finalized TDP/PPT. It leaves room for future products to exist that have more power: either through higher clock speed or more cores or both.

I agree that seems like it makes sense, at least with the info we have now. But I wonder how the TDP ratings will compare to Zen 3, with Zen 3 TDP was accurate for average or gaming use power consumption, PPT was boost clocks or overclocking power consumption, socket power of 230W leaves room for future cpu's.

Though it would be nice if Zen 4 isn't significantly more power hungry, I actually care about power consumption as I don't like my system heating up a whole room. And GPU's are consuming 250-300w on mid range parts,  a more power hungry CPU requires more expensive cooling, and a larger case with more fans running.

[BiG StroOnZ]

1 hour ago, porina said:

I think we've gone over the logical and even less logical possibilities. Can only wait and see what really happens. It's going to be interesting towards the end of the year one way or other.

 

It's also going to be an interesting summer with all the leaks too! 

 

36 minutes ago, Blademaster91 said:

I agree that seems like it makes sense, at least with the info we have now. But I wonder how the TDP ratings will compare to Zen 3, with Zen 3 TDP was accurate for average or gaming use power consumption, PPT was boost clocks or overclocking power consumption, socket power of 230W leaves room for future cpu's.

Though it would be nice if Zen 4 isn't significantly more power hungry, I actually care about power consumption as I don't like my system heating up a whole room. And GPU's are consuming 250-300w on mid range parts,  a more power hungry CPU requires more expensive cooling, and a larger case with more fans running.

 

Well as far as it comparing to Zen 3, from what we have currently, information wise; Zen 4 is a 65W TDP and an 88W PPT increase over AMD's current Zen 3 flagships. They are definitely going to require more expensive cooling solutions as not only do you have the increase in TDP/PPT but also increase in thermal density from the new process node. No matter how you slice it, Zen 4 is going to be a different beast to tame compared to Zen 3; which is relatively mild in comparison to Alder Lake. AMD might have cooked up their own Alder Lake here. 

[leadeater]

9 hours ago, tim0901 said:

I think future proofing the socket makes a lot of sense - it might encourage board vendors to build better boards as well, preventing the issues we saw around Zen 2 where early Zen 1 motherboards weren't built well enough to handle the newer chips, despite otherwise fully supporting them.

A lot of that was faith issues too though, AMD boards just didn't sell previously and the products had really bad reputation. Would have taken a lot of faith to make X570 quality and effort boards back then. Fully understand why they didn't.

[BiG StroOnZ]

Look what I found ~

 

During Computex, we (TechPowerUp) caught up with Robert Hallock and asked him many questions we had, and also included things our community members brought up in numerous forum discussions in recent days:

 

Quote

16-core, 32-thread is the maximum core configuration for the Ryzen 7000 at launch?

That is correct.

 

At Computex, you showed a 15% single-thread performance gain over the Ryzen 9 5950X. Wouldn't that only put the gaming performance on par with 5800X3D?

I think it's too early to say actually. We were deliberately conservative with our number on single-thread performance. We do intend to publish the exact breakdown of IPC vs. frequency contribution later in the summer, also including performance, power, and area on the new process. As far as what stacks up against what, I think it's too early to say, we're still in silicon bring-up.

 

Your presentation mentioned "AI acceleration." Is that AVX-512 or something more exotic, like Intel GNA?

Yes. Specifically, AVX 512 VNNI for neural networking and AVX 512 BLOAT16 for inferencing. Both are pretty nice speedups, we're not using a fixed-function acceleration, this could be something we could do with our Xilinx acquisition. We are starting to see more consumer applicability of AI workloads, like video upscaling, which has grown a lot in the last two years. I think there is a general trend for the average enthusiast to take on more AI-type workloads. The time felt right to bring these features into the chip, given we moved to a smaller process node with better performance, power, and area capabilities.

 

Will AM4 coolers provide the optimum experience or just be "compatible"? i.e., will there be new cooler releases for AM5?

 

I think it's a little bit of both. We still intend to offer 65 W and 105 W CPUs even though the socket power limit has gone up to 170 W. Not every single processor will use that power envelope. Coolers designed for the 65 W and 105 W parts will be equally appropriate for AM5. For the 170 W socket power CPUs, I expect the existing higher-end air and liquid-cooling units to be perfectly fine, but there will probably be a wave of new solutions positioned as "designed for" these CPUs. Personally, I have a 5950X in my system on a Noctua NH-D15, and I fully intend to reuse that cooler on Socket AM5.

 

What can we expect from the processors in terms of CPU overclocking?

I'm not gonna make a commitment yet on frequency, but what I will say is that 5.5 GHz was very easy for us. The Ghostwire demo was one of many games that achieved that frequency on an early-silicon prototype 16-core part with an off-the-shelf liquid cooler. We're very excited about the frequency capabilities of Zen 4 on 5 nanometer; it's looking really good, more to come.

 

https://www.techpowerup.com/review/amd-zen-4-ryzen-7000-technical-details/

 

I, for obvious reasons, didn't quote the whole article; check the link above for more questions being answered. 

 

Also from TPU:

 

Quote

Apparently, there will be certain SKUs of Socket AM5 processors with TDP of 170 W. 

 

This does not necessarily mean that there will be a Ryzen 7000-series SKU with 170 W TDP. AMD plans to give AM5 a similar life-cycle to AM4, which is now spanning five generations of Ryzen processors, and the 170 W TDP and 230 W PPT figures only denote design goals for the socket. AMD, in a statement, explained why it needed to make AM5 capable of delivering much higher power than AM4 could—to enable higher CPU core-counts in the future, more on-package hardware, and for new capabilities like power-hungry instruction-sets (think AVX-512).

 

https://www.techpowerup.com/295301/amd-clarifies-ryzen-7000-zen-4-tdp-and-power-limits-170w-tdp-230w-ppt

[starsmine]

I strongly suspect that even though the socket can do 170W, that no 7000 chip will have a 170W tdp, that exists more for a 24core 8000 or 9000 ryzen r9 chip.

[Rauten]

5 hours ago, BiG StroOnZ said:

"AMD plans to give AM5 a similar life-cycle to AM4, which is now spanning five generations of Ryzen processors"

I'd like to see an actual AMD spokeperson/slide talk about this. If this is true, I'll splurge on a capable PCIe 5.0x16 board and just run it for years and years. Other than being limited to PCIe3.0, my current X370 motherboard has been amazing.

So even if Zen4 isn't all that great, I can just stay on the same platform for ages and upgrade the CPU later.

 

As for the chipset hellscape:

Quote

On the X670 (non-Extreme) only the M.2 NVMe slot is required to be Gen 5, the top slot for graphics will optionally be Gen 5.

Fears confirmed. Goddamnit AMD. 

Inb4 the forum gets swamped with "hwy dosnt my moderbord run GPU 5.0¿?" threads.

[LAwLz]

35 minutes ago, Rauten said:

Quote

On the X670 (non-Extreme) only the M.2 NVMe slot is required to be Gen 5, the top slot for graphics will optionally be Gen 5.

Fears confirmed. Goddamnit AMD. 

Inb4 the forum gets swamped with "hwy dosnt my moderbord run GPU 5.0¿?" threads.

What?

AMD specifically wrote "X670 - PCIe 5.0 Storage and Graphics" on their slide, while B650 only said "PCIe 5.0 Storage". My assumption when I see "B650 will only have PCIe 5.0 to the storage, X670 will have it to the storage and graphics card" is that X670 will actually have it to the graphics card, not that it is optional.

 

If PCIe 5.0 to the graphics card is optional on X670, what differentiates a PCIe 4.0 X670 board from a B650 board? Seems like they will be the same thing to me.

Anandtech wrote that B650 will not support overclocking but that seems to be incorrect as well.

 

AMD did a really poor job communicating and clarifying things with this presentation.

 

 

 

By the way, I saw someone in the TechpowerUp comments misunderstand Robert when he talked about the AI instructions and I fear more people will misunderstand him.

Robert did not confirm that Zen 4 has AVX 512 support. In fact he basically did the opposite and all but confirmed that Zen 4 do not have AVX 512 support.

 

AVX 512 is a collection of multiple new and reworked instructions. This is essentially how the conversation went:

TechPowerUp: So will Zen 4 support the hundreds of new instructions that are called AVX 512?

Robert: Yes, we will support 10 of those new instructions, but not in dedicated hardware.

 

 

Don't get me wrong, those instructions are pretty damn important for AI workloads, but they are by no means the "AVX 512" that people like the developers of RPCS3 are excited about. The fact that they are not implementing it in fixed-function hardware also does not sound very positive to me. 

[porina]

19 minutes ago, LAwLz said:

If PCIe 5.0 to the graphics card is optional on X670, what differentiates a PCIe 4.0 X670 board from a B650 board?

On GPU lanes support, to my understanding X670E is 5.0 mandatory, X670 is 5.0 optional, and B650 is 4.0 only. Is there somewhere we can get the full slide deck from AMD? It appears the X670 may be covered in a footnote. On differences between B650 and X670, if we look back to 500 series, the B550 used and offered 3.0 connectivity, with 4.0 on the X570. Assuming similar that could contribute to cheaper B650 boards.

 

I feel AMD could have done better on chipset names here, but to be fair I suppose X670 with 4.0 or 5.0 is still the same chipset, and it wouldn't make sense to change the chipset name to differentiate between them. Making mandatory 5.0 would have been nice, and would put them on a par with Intel.

 

 

Random thought on chipset naming wars: Intel started the B series with B150 and B250. AMD stole B350, making Intel move to B360 and so on. With AMD's slowing of chipset releases, AMD are now releasing B650 after Intel already used B660. I guess that answers the question I was about to ask, which is if Intel will reclaim B750 with Raptor Lake? The answer is no. If they already passed on returning to B650 before AMD, and are sticking to Bx60.

 

19 minutes ago, LAwLz said:

AVX 512 is a collection of multiple new and reworked instructions. This is essentially how the conversation went:

TechPowerUp: So will Zen 4 support the hundreds of new instructions that are called AVX 512?

Robert: Yes, we will support 10 of those new instructions, but not in dedicated hardware. 

On Intel side, if you have AVX-512 support, there are mandatory and optional levels. My personal interests would make use of AVX-512F which is mandatory on Intel. This provides extension to FP64 operations. Even in implementations not providing additional FP execution resource (one unit) it can provide a good boost in throughput over AVX2, on the order of 40% IPC for Prime95 and similar. Two unit was around 80%. This would have made Zen 4 much more interesting to me had it supported it.

 

The TPU article does seem to imply Robert confirmed they have an implementation of VNNI. On Intel side this was offered as an optional extension to AVX-512. It may be in this or other similar thread I learnt it can be offered independently of base AVX-512. This does seem like the most logical route unless we hear otherwise. It would not be good for the wider community to have different instructions doing pretty much the same thing.

[Rauten]

52 minutes ago, LAwLz said:

AMD did a really poor job communicating and clarifying things with this presentation.

It's a mess is what it is.

5 minutes ago, porina said:

I feel AMD could have done better on chipset names here, but to be fair I suppose X670 with 4.0 or 5.0 is still the same chipset, and it wouldn't make sense to change the chipset name to differentiate between them. Making mandatory 5.0 would have been nice, and would put them on a par with Intel.

Yeh, the chipset is exactly the same. The "E" is merely a branded thing, basically a way to say that the motherboard, not the chipset, has everything wired for 5.0 and has been designed with OC in mind (god only knows what that will actually mean in the products).

 

This should have been:

X670E - EVERYTHING is wired for 5.0

X670 - One PCIE x16 5.0 slot and one NVME x4 5.0 slot

B650 - One NVME x4 5.0 slot

 

That way it would've made a modicum of sense - there would be clear differences between the motherboard "tiers". Instead we've ended up with an X670 that's going to be utter chaos.

 

Yes I keep railing at AMD about this but it really pisses me off.

[FentonW]

AMD is defiantly sandbagging on this one.

Consider the 6000 Series mobile got about a 10% performance improvement just by moving to DDR5.

We are also getting a 10% clock boost + doubling of L2 Cache.

These improvements alone should give you the base 15% improvement. So any IPC improvements will be
on top of this.

However even if it only 15% improvement but at a lower cost (going from 7nm to 5nm), AMD could still hold the price performance
crown. If they come in somewhere between Alderlake and Raptor lake at current 5000 series prices they are still on to a winner.
Then they just need to create an extra 3D V-cache SKU to beat Raptor lake in games.

 

 

[BiG StroOnZ]

13 hours ago, Rauten said:

I'd like to see an actual AMD spokeperson/slide talk about this. If this is true, I'll splurge on a capable PCIe 5.0x16 board and just run it for years and years. Other than being limited to PCIe3.0, my current X370 motherboard has been amazing.

So even if Zen4 isn't all that great, I can just stay on the same platform for ages and upgrade the CPU later.

 

It seems AMD has plans to do some revealing of more details during the summer, so I'm sure you will get a better understanding from AMD on this topic before launch. But it makes sense to future proof yourself by buying a better motherboard, even if anti-future proofing has become the newest fad (saying there is no such thing as future proofing).