Discussion Speculation: Zen 4 (EPYC 4 "Genoa", Ryzen 7000, etc.)

[Vattila]

Except for the details about the improvements in the microarchitecture, we now know pretty well what to expect with Zen 3.

The leaked presentation by AMD Senior Manager Martin Hilgeman shows that EPYC 3 "Milan" will, as promised and expected, reuse the current platform (SP3), and the system architecture and packaging looks to be the same, with the same 9-die chiplet design and the same maximum core and thread-count (no SMT-4, contrary to rumour). The biggest change revealed so far is the enlargement of the compute complex from 4 cores to 8 cores, all sharing a larger L3 cache ("32+ MB", likely to double to 64 MB, I think).

Hilgeman's slides did also show that EPYC 4 "Genoa" is in the definition phase (or was at the time of the presentation in September, at least), and will come with a new platform (SP5), with new memory support (likely DDR5).

What else do you think we will see with Zen 4? PCI-Express 5 support? Increased core-count? 4-way SMT? New packaging (interposer, 2.5D, 3D)? Integrated memory on package (HBM)?

Vote in the poll and share your thoughts!

 

[eek2121]

It's just trolling. There've been pretty much no performance leaks for either Intel or AMD 2024 server products. Hopefully with silicon starting to make its way out, we see something soon.

…or any product for that matter.

 

[Ajay]

They manage to keep customers buying their inferior offering thanks to pure lies, promising that their future products will be much better than AMD s, and it work apparently, just read this report and take notice of Intel s deceptive discourse to their customers :

Intel Hopes to Stop Server Beating from AMD Next Year

After getting bruised in servers by AMD, Intel hopes to stop the bleeding in the server market with next year's chip offerings. The difference-making products will be Sierra Forest and Granite Rapids, which are due out in 2024, said Dave Zinsner, chief financial officer at Intel, last week at...

www.hpcwire.com

An exemple of the Intel lies to their customers :





See the discourse.?.
Fantastic, they say, yet we all know that their SF and GR will be handily destroyed by already existing Bergamo and Genoa, let alone by AMD s Zen 5 based offering that will come next year at the same time as those fake marvels, as for existing SR 8490H we can see at Phoronix how it perform "extremely well"...

It s all about buying time with the hope of barely catching up at some point because currently they are roughly one if not two generations late.

Customers who currently buy Intel are jeopardising their businesses by a way bigger TCO and even more likely by the fact that 2 Intel s servers are required to compete with a single AMD one in perfs while being a disaster power comsumption wise, at the end of the day Intel with its deceptive discourse will bring them to bankruptcy, and i say it s so far so good, that s what incompetent people deserve.

Wow, this is so delusional I had to laugh. I can only imagine that Intel will have to sell their server CPUs at cost just to keep the FABs spinning and maintaining cash flow.

 

[Abwx]

Wow, this is so delusional I had to laugh. I can only imagine that Intel will have to sell their server CPUs at cost just to keep the FABs spinning and maintaining cash flow.

Even at cost that s a financial disaster for the eventual customers, tests at Phoronix show dual Bergamo at 385W at up to 2 x the perf of a dual SR at 568W.

That s 47% more power at up to half the perf, and 3x the power at same perf assuming one put 2 dual SR to face a single dual Bergamo, actually there s no better way to get your business bankrupted than buying this outdated gear.

If anything customers who take the intel route will lose any competitivness, not only in TCO but in time to market when it comes to products design as well as a huge loss of productivity, think of softwares like LS Dyna wich require huge times for various simulations in the design course, with hardware that inferior crews productivity will be litteraly halved.

 

[Markfw]

Even at cost that s a financial disaster for the eventual customers, tests at Phoronix show dual Bergamo at 385W at up to 2 x the perf of a dual SR at 568W.

That s 47% more power at up to half the perf, and 3x the power at same perf assuming one put 2 dual SR to face a single dual Bergamo, actually there s no better way to get your business bankrupted than buying this outdated gear.

If anything customers who take the intel route will lose any competitivness, not only in TCO but in time to market when it comes to products design as well as a huge loss of productivity, think of softwares like LS Dyna wich require huge times for various simulations in the design course, with hardware that inferior crews productivity will be litteraly halved.

The other thing, is that the power is not directly related to cost. There have been arguments about this as to the amount, but in a server room that power is related also to how much AC is costing, AND how much backup power is required. So, when the power doubles the actual power could go up by 4,6, even 8 times.

 

[yuri69]

Oh, and then comes your trusty OEM salesman and gets you a *nice price* on those new Intel models. "It's a deal, see ya next year".

Business is definitely not objective matter.

 

[biostud]

I work at a school and we have our own server to run our administration network. I'm pretty sure our IT tech guy call a IT provider, tell them what we need, and they then present him with different hardware options depending on the price. Then he chooses whichever fits our budget, has a good service plan and good support. He doesn't really care too much about which processors are in the servers. I guess that is the case for most smaller and medium sized companies, that doesn't need servers for specific computational needs.

 

[Mopetar]

Even if AMD is better for most customers they can't supply 100% of the market so someone is going to have to buy Intel or stay with what they have. I'd imagine anyone buying in volume is getting a big discount that puts the performance per dollar more squarely in line with AMD.

 

[Markfw]

Wow, this is so delusional I had to laugh. I can only imagine that Intel will have to sell their server CPUs at cost just to keep the FABs spinning and maintaining cash flow.

Not only that, the site can't spell, or or totally inept. Here is a quote:
"AMD’s next big chip release is Bergamot, which is coming out in the first half of this year."

Bergamot ??? What am I missing

 

[Abwx]

Even if AMD is better for most customers they can't supply 100% of the market so someone is going to have to buy Intel or stay with what they have. I'd imagine anyone buying in volume is getting a big discount that puts the performance per dollar more squarely in line with AMD.

Servers volume is not as big as the rest of the market, and even if the CPUs use a lot of silicon they could still easily ship 60% of the market, it s not like TSMC is that capacity constrained, only limitation eventually would be for Genoa X but that should be a niche within the whole market needs.

 

[Mopetar]

They could ship more in server if they were willing to abandon desktop and other markets. They'd make more money that way, but it's not a good long term move.

 

[Abwx]

They could ship more in server if they were willing to abandon desktop and other markets. They'd make more money that way, but it's not a good long term move.

They had a big decline in DT recently, so capacities are there, beside when Zen 4 launched they stated that they wouldnt be capacities constrained and that they could feed the market whatever the demand.

Also TSMC pointed recently that they are at lower than optimal use of production capacities.

TSMC fab utilization to recover starting 2Q23

Part of TSMC's fab utilization rates, including those for 7/6nm process nodes, will begin to recover in April, sooner than previously predicted by the market, according to sources at semiconductor equipment companies.

www.digitimes.com

 

[Joe NYC]

They could ship more in server if they were willing to abandon desktop and other markets. They'd make more money that way, but it's not a good long term move.

TSMC has a ton of capacity, spare / idle capacity. Just look at their financials. AMD could take even 100% of server market and TSMC would still be running well below full utilization.

AMD server shipments stagnated last 2 quarters, while the client collapsed. If anything, AMD had cancel a ton of wafer starts / pay penalties to TSMC or come to some other terms with TSMC.

TSMC has been drowning in overcapacity from cancellations from multitudes of customers across the industries.

Which is why NVidia could come in and at moments notice get some of this idle TSMC capacity for their H100 GPUs.

AMD could do the same in the heartbeat, for 100% of the server market and could still at the same time double the client PC production.

There is just no tradeoff when we are at the bottom of the semiconductor cycle. Can we just stop this silly trade-off talk for at least a year? Please?

 

[Mopetar]

As I recall, AMD said they were limited by substrate availability. No sense in buying more wafers than you can make into chips. It doesn't really matter where the bottleneck is, but AMD can't just flip a switch and magically have production capacity to take 100% of the market. Even if they have the substrate shortage in hand they can only produce more until the next bottleneck. There usually isn't massive amounts of unused and readily available production capacity just sitting around. Markets don't tolerate that much inefficiency.

 

[moinmoin]

I think AMD said around late last year that substrate is no longer a limiting factor. The direct result of that was that the PS5 was no longer supply constrained since.

 

[Joe NYC]

As I recall, AMD said they were limited by substrate availability. No sense in buying more wafers than you can make into chips. It doesn't really matter where the bottleneck is, but AMD can't just flip a switch and magically have production capacity to take 100% of the market. Even if they have the substrate shortage in hand they can only produce more until the next bottleneck. There usually isn't massive amounts of unused and readily available production capacity just sitting around. Markets don't tolerate that much inefficiency.

The last time substrate was a bottleneck was exactly a year ago. AMD predicted that by Q4, 2022, supply would catch up to demand.

What happened in subsequent 3 quarters is that datacenter growth stalled, client dropped 66% (plenty of substrate there as well) and the rest of the market tanked as well.

So all the other market participants cancelled portion of their substrate orders, freeing up a lot of capacity.

The only "bottleneck" that AMD is facing is the market that has been rigged by Intel.

That's a demand bottleneck for AMD products "engineered" by Intel's market manipulation, not a supply bottleneck.

 

[Abwx]

The only "bottleneck" that AMD is facing is the market that has been rigged by Intel.

That's a demand bottleneck for AMD products "engineered" by Intel's market manipulation, not a supply bottleneck.

Gelsinger learned his lesson very well from Otellini since he was at Intel at the time of massive briberies, so he s undoubtly relying on the same tricks, not bribing directly this time but throwing the same amounts in a more pernicious way.

If we look at Intel s financials they started being in the red just when AMD DC offering massively outperformed their products, and yet AMD didnt manage to get higher than 30% marketshare the following quarters.

It s likely that they sold, and still selling, their servers CPUs at huge losses to keep AMD from gaining marketshare, compensated with the DT and mobile market revenues and then proceeded to put a little loss in their DC results official numbers to make things look as if they sold DC CPUs for a little profit but in less quantities than expected while they were actually bleeding cash by the bns quarterly in this market.

 

[jpiniero]

It s likely that they sold their servers CPUs at huge losses to keep AMD from gaining marketshare, compensated with the DT and mobile market revenues and then proceeded to put a little loss in their DC results official numbers to make things look as if they sold DC CPUs for a little profit but in less quantities than expected while they were actually bleeding cash by the bns quarterly in this market.

Intel's financial report for Q1 says Server volume was down 50%.

 

[Abwx]

Intel's financial report for Q1 says Server volume was down 50%.

They made up their numbers to make things look like it was due to lower volume, but if their volume actually shrinked by 50% then AMD would be automaticaly at 50% marketshare currently since they were already at about 25% and that the whole market was 25% lower volume wise.

Do the maths, their numbers do not hold unless their volume did shrink much less but was sold at an huge loss, and that they covered this by saying that it was the volume that collapsed.

 

[moinmoin]

Yeah, wanted to point out that Su just said AMD would exceed 25% in the global server CPU market:

"I actually think our server market share is higher than 20%, ... should be over 25%."

AMD's global server CPU market share tops 25%, says Lisa Su

As the market's attention is focusing on AMD's confirmation on having TSMC produce its latest Instinct MI300 series generative AI accelerator to be released in the fourth quarter of the year, AMD CEO Lisa Su disclosed that AMD's server market share has continued to progress and has exceeded 25%.

www.digitimes.com

 

[Exist50]

That's a demand bottleneck for AMD products "engineered" by Intel's market manipulation, not a supply bottleneck.

What particular market manipulation do you claim is occurring today?

 

[Markfw]

What particular market manipulation do you claim is occurring today?

I have no evidence to support this, but I would guess he is talking about Intel virtually or literally giving away server CPU's to stop the loss of market share due to their server chips being so inferior at the moment. It is only logical, as AMD has proven in recent benchmarks to exceed Intel in performance by ~100% and use ~50% of the power in doing so. Why else would data centers keep buying CPUs that are inferior and cost them more to run, even at the same price !

 

[Hitman928]

Yeah, wanted to point out that Su just said AMD would exceed 25% in the global server CPU market:

"I actually think our server market share is higher than 20%, ... should be over 25%."

AMD's global server CPU market share tops 25%, says Lisa Su

As the market's attention is focusing on AMD's confirmation on having TSMC produce its latest Instinct MI300 series generative AI accelerator to be released in the fourth quarter of the year, AMD CEO Lisa Su disclosed that AMD's server market share has continued to progress and has exceeded 25%.

www.digitimes.com

Not mentioned in the article but she is talking about revenue share specifically, not volume share.

 

[Exist50]

but I would guess he is talking about Intel virtually or literally giving away server CPU's to stop the loss of market share due to their server chips being so inferior at the moment

So it's market manipulation to sell at low margins? That's a conspiracy? AMD spent the better part of the 2010s doing that themselves.

as AMD has proven in recent benchmarks to exceed Intel in performance by ~100% and use ~50% of the power in doing so. Why else would data centers keep buying CPUs that are inferior and cost them more to run, even at the same price !

You're referring to Bergamo, correct? So can you point me towards where I can buy a Bergamo server? Or rent a Bergamo VM? You can't expect to see revenue and marketshare impact until people can actually start buying a product. Enterprise in particular is a slow ramp. AMD themselves have spoken about this on occasion.

 

[Markfw]

So it's market manipulation to sell at low margins? That's a conspiracy?

You're referring to Bergamo, correct? So can you point me towards where I can buy a Bergamo server? Or rent a Bergamo VM? You can't expect to see revenue and marketshare impact until people can actually start buying a product. Enterprise in particular is a slow ramp. AMD themselves have spoken about this on occasion.

Genoa.... I will not bother to link, as its all in benchmarks in this thread.

 

[moinmoin]

Not mentioned in the article but she is talking about revenue share specifically, not volume share.

Isn't it about revenue most of the times anyway?

Maybe Intel may change to units since that's the more favourable number for them.