AMD Pushes Ryzen 7000 "Zen 4" Availability Date to Clash with Intel "Raptor Lake" Announcement Date

[Minus Infinity]

AMD has learned their lessons the hard way in dealing with Intel's shonky benchmarks that most of the media are all too eager to promote with no questions asked. Smart move IMO.

 

[Dirt Chip]

Do whatever you can to maximize your profit and better your position against compatition.
Switching a date is a very easy and cheep way to do it- I see no reson not to do it if AMD think it is better for them and remember that thay konw thing we dont.

In any case, I dont intend to purchase any new AMd CPu untill intel RL is lunches and I will see compression review from YOU and others so this supposed 'delay' changes nothing for me personally.

 

[nguyen]

I had my x370 + 1800x + 32GB 3600 ram working out of the box too.

I also tried to install esxi on a 12th gen platform and nope it runs like trash.

Sample size of one doesn't mean anything.

but Yes, new platform always introduces new bugs.

Yeah I don't believe you, 32GB 3600mhz working with 1800x? haha good joke

 

[TheLostSwede]

Well I'm still on z370 platform from 2017, Quad Ranks 3600mhz RAM just work out of the box

Then I build x570 with R5 3600 for my nephew and every cold boot system cannot POST with RAM above 3200mhz (dual ranks for quad ranks same problem), platform stability is just not ideal. I'm sure AM5 will have plenty of bugs to be iron out.

Did you 1. update the UEFI and 2. not use Corsair LPX memory?
Been running my X570 system for a few years with the memory overclocked, no problem at all, initially with a 3800X and now a 5800X.
Early UEFI/AGESA releases had issues, but after about six months, everything has been rock solid.

 

[nguyen]

Did you 1. update the UEFI and 2. not use Corsair LPX memory?
Been running my X570 system for a few years with the memory overclocked, no problem at all, initially with a 3800X and now a 5800X.
Early UEFI/AGESA releases had issues, but after about six months, everything has been rock solid.

1. Yes I updated to latest UEFI, still problem with cold boot when RAM above 3200mhz
2. Tried 2 kits, same result, first was Gigabyte Aorus 3200mhz cas16 (would not POST), second kit was Corsair Vengeance 4x8GB 3466mhz cas16 (samsung b-die) that had been running 3600mhz on my 8700K+Z370 system.

 

[HenrySomeone]

Feels like a very blatant move to prevent comparison with Raptor lake on Intel launch. Maybe intel has a clear winner after all in raw performance and AMD has now accepted that.

Precisely! This certainly isn't a move you do when you're full of confidence!

Would it not look worse for AMD's image if Intel compares RL to Z3 and shows big wins (50% or more) and wins in everything?
Compared to comparing RL to Z4 and even if Intel wins it will not be in everything and will likely be single digit percentages in most cases.
I mean think about what would look worse for AMD:

Headline 1: Intel Raptor Lake is 50% faster than AMD Zen3.
Headline 2: Intel Raptor Lake is 10% faster than AMD Zen4.

At the same time AMD can say that yes Intel is slightly faster than us because they have pushed power consumption to 11 and have slight core count advantage.
I mean the second option makes AMD look much better in my opinion if their latest and greatest is only slightly slower but much more efficient and they have X3D model(s) coming that further improve gaming performance.

It would definitely look worse, but likely reason is that it won't be just 10% faster than Zen4, especially not in the mainstream class where 20 thread 13600k is going to obliterate 7600x in ... everything and the only potential win is going to be 7950x vs 13900k in rendering by like 4%...

 

[TheLostSwede]

1. Yes I updated to latest UEFI, still problem with cold boot when RAM above 3200mhz
2. Tried 2 kits, same result, first was Gigabyte Aorus 3200mhz cas16 (would not POST), second kit was Corsair Vengeance 4x8GB 3466mhz cas16 (samsung b-die) that had been running 3600mhz on my 8700K+Z370 system.

Well, I have some cheap "shit" Patriot Viper 3600 modules that run at 3800 CL16 so...

 

[nguyen]

Well, I have some cheap "shit" Patriot Viper 3600 modules that run at 3800 CL16 so...

Haha yeah I bet it runs fine on Intel too.
Let just say I didn't have very nice PC building experience with x570 platform.

Just build a new 12400F+B660 and it took me 30min from building the PC to starting a game, not gonna happen on x570

 

[HenrySomeone]

Well, I have some cheap "shit" Patriot Viper 3600 modules that run at 3800 CL16 so...

It doesn't matter "cheap shit" or not, it matters whether the particular ram was(is) on the approved and confirmed-to-be-working list. If you failed to select proper ram, especially in the first couple years, you were likely in for a whole lot of issues that most often required change of modules or serious compromise on the frequency. Things are somewhat better nowadays (it'd be really bad if they weren't, X570 is out over 3 years now), but it's really pointless to deny the problems, there are literally thousands of documented cases on different forums.

Haha yeah I bet it runs fine on Intel too.
Let just say I didn't have very nice PC building experience with x570 platform.

Just build a new 12400F+B660 and it took me 30min from building the PC to starting a game, not gonna happen on x570

I actually know of a guy who awhile ago set out to build 10 systems with 3900x-s for cpu mining and wasn't paying enough attention to ram. Now apparently you want pretty decent specs for that and he got I think 3600 cl 16 sets, but when it came time to assemble the rigs, he couldn't get a single one working right. With some he couldn't get them to boot at anything over 2400 and some wouldn't boot at all, lmao! Needless to say, he was pretty pissed...

 

[docnorth]

Actually the 'rumors' about reviews starting 13 Sep and sales 15 Sep 2022 were supposedly based on AMD's presentations, not just leaks. This isn't fair, if indeed the hard launch will be delayed, how are we gonna bash Intel for the numerous delays++++?

 

[TheLostSwede]

Haha yeah I bet it runs fine on Intel too.
Let just say I didn't have very nice PC building experience with x570 platform.

Just build a new 12400F+B660 and it took me 30min from building the PC to starting a game, not gonna happen on x570

Whatever dude, this is just your opinion. I've built PCs since the 1990's and things are way much better these days and one experience doesn't mean the entire platform is shit.
My mate went with Alder Lake and Z690, can get his 4400 B-dies to work properly. Doesn't mean the entire platform is shit again.

It doesn't matter "cheap shit" or not, it matters whether the particular ram was(is) on the approved and confirmed-to-be-working list. If you failed to select proper ram, especially in the first couple years, you were likely in for a whole lot of issues that most often required change of modules or serious compromise on the frequency. Things are somewhat better nowadays (it'd be really bad if they weren't, X570 is out over 3 years now), but it's really pointless to deny the problems, there are literally thousands of documented cases on different forums.

No, not at all. I keep having to repeat this, as people keep praying coming up with this same BS.
The motherboard makers just test with whatever RAM they have laying around or what they get sent. They don't actively go out and purchase modules to test with. As such, what the QVL's contain means very little, beyond the fact that some junior engineer managed to boot the board with that RAM.

My modules aren't on the QVL or any X570 QVL that I've seen for that matter. They still work just fine.

I'm not denying there were problems, as I wrote, the first six months there were plenty issues, but after that, I've had none.

 

[nguyen]

Whatever dude, this is just your opinion. I've built PCs since the 1990's and things are way much better these days and one experience doesn't mean the entire platform is shit.
My mate went with Alder Lake and Z690, can get his 4400 B-dies to work properly. Doesn't mean the entire platform is shit again.

Whatever dude, just sharing my own experience, no need to get defensive LOL.

Looks to me like you had issue in the first 6 months of owning X570 too, apparently it was fine for you, but it's not for me.

 

[TheLostSwede]

Whatever dude, just sharing my own experience, no need to get defensive LOL.

Looks to me like you had issue in the first 6 months of owning X570 too, apparently it was fine for you, but it's not for me.

No need to spread FUD either.

 

[nguyen]

No need to spread FUD either.

You are spreading your own FUD

I'm not denying there were problems, as I wrote, the first six months there were plenty issues, but after that, I've had none.

 

[Crackong]

Yeah I don't believe you, 32GB 3600mhz working with 1800x? haha good joke

I had to dig through my old Z: to find these 2018 photos
Here is proof
I was running this thing in 1.3v 3.9GHz 32GB RAM 3600 16-19-19-39 until my 3700x arrived

And it was a pretty good 1800x runs up to 4.05GHz all core

 

[ThrashZone]

Hi,
Early adopters usually find issues first is this really surprising

 

[ratirt]

Changing release dates to Intels product release. AMD must feel confident with the product and want to steal the fanfares from Intel maybe. Comparing 2 new products at the same time of release should be interesting.

 

[Assimilator]

Just build a new 12400F+B660 and it took me 30min from building the PC to starting a game, not gonna happen on x570

I replaced my X470 board + 5800X + 32GB 3000CL15 memory with an X570 board + 5900X + 32GB 3600CL18 memory this weekend, reusing my current Windows 10 install. The only thing I had to do was set XMP in the BIOS and copy the MediaTek Wi-Fi driver from another PC, as Win10 doesn't include it by default and Gigabyte is still shipping drivers on CDs in 2022. That bit of setup took ~10 mins and I was playing games as normal.

 

[ratirt]

I replaced my X470 board + 5800X + 32GB 3000CL15 memory with an X570 board + 5900X + 32GB 3600CL18 memory this weekend, reusing my current Windows 10 install. The only thing I had to do was set XMP in the BIOS and copy the MediaTek Wi-Fi driver from another PC, as Win10 doesn't include it by default. That bit of setup took ~10 mins and I was playing games as normal.

Had the same thing with my 2700x and x470 when I swapped it with a 5800x and x570. The difference was, I didnt have to update WiFi since I didnt have one with either boards.

 

[trsttte]

I won't call this a 'push' since these are all RUMOR announcement and availability dates
You can't just put 2 rumored dates in your head and say AMD pushed it.

Well yes but if the shoe fits... It's a marketing game, wouldn't be surprised if Intel now tried to delay their release and/or review NDA to be able to do the same. Smart buyers will wait for proper comparisons between zen4 vs RaptorLake anyway regardless of whatever bullshit both companies try to pull to scew initial comparisons

 

[DeathtoGnomes]

Well yes but if the shoe fits... It's a marketing game, wouldn't be surprised if Intel now tried to delay their release and/or review NDA to be able to do the same. Smart buyers will wait for proper comparisons between zen4 vs RaptorLake anyway regardless of whatever bullshit both companies try to pull to scew initial comparisons

This.

I think this is a smart move for AMD, Intel has a history of using any tactic in their PRs to trash AMD. From glue comments to the selective performance tests with graphs out of proportion bars to the pricing competions. We also know that AMD is not innocent when it comes to PRs either, so this new launch campaign from both sides should be worthy of a bucket of extra buttery popcorn and Snow-caps.

 

[Shocktruppen]

Would it not look worse for AMD's image if Intel compares RL to Z3 and shows big wins (50% or more) and wins in everything?
Compared to comparing RL to Z4 and even if Intel wins it will not be in everything and will likely be single digit percentages in most cases.
I mean think about what would look worse for AMD:

Headline 1: Intel Raptor Lake is 50% faster than AMD Zen3.
Headline 2: Intel Raptor Lake is 10% faster than AMD Zen4.

At the same time AMD can say that yes Intel is slightly faster than us because they have pushed power consumption to 11 and have slight core count advantage.
I mean the second option makes AMD look much better in my opinion if their latest and greatest is only slightly slower but much more efficient and they have X3D model(s) coming that further improve gaming performance.

You left out a third headline, AMD Zen 4 is 30% faster than Intel Raptor Lake and uses
25%less power doing it.

 

[HenrySomeone]

You left out a third headline, AMD Zen 4 is 30% faster than Intel Raptor Lake and uses
25%less power doing it.

That's because this is Techpowerup, not Fairytale Central and Little RED Riding Hood doesn't win just because her adoring fans wish it to be so

 

[Crackong]

Well yes but if the shoe fits... It's a marketing game, wouldn't be surprised if Intel now tried to delay their release and/or review NDA to be able to do the same. Smart buyers will wait for proper comparisons between zen4 vs RaptorLake anyway regardless of whatever bullshit both companies try to pull to scew initial comparisons

Now the AMD announcement date is confirmed on 29 Aug, we will see.

 

[Smartcom5]

Feels like a very blatant move to prevent comparison with Raptor lake on Intel launch.

It actually for sure is – A blatant move to prevent Intel to compare themselves with anything AMD.
Though it is highly likely not for the very reasons you might think it would be. It's likely not for the reasons to hide a subpar product from AMD, but to prevent Intel to a) hide their lacklustre Raptor Lake through shady marketing stunts and b) to prevent AMD being ridiculed again by Intel like in any past through last-minute highly questionable manoeuvres.

Remember the Threadripper-launch of the Threadripper 3970X/3960X back then in 2019? That was when Intel just hours prior to the NDA-disclosure and official announcement of the Core i9 10980XE on the very same day shadily pulled the official embargo-lift a few hours, just to prevent being blasted away by AMD's utmost superior and outright outclassing new Threadripper-class CPUs for steal the well-deserved limelight from AMD for their outstanding achievements.

Since Intel's less-than-lukewarm Skylake-X 2.0 (Skylake-X miserly rehashed as the 'new' Cascade Lake-X) – which even had to be halved in price, to make the obvious bloodbath a little less painful – was so far behind, it wasn't even funny anymore. Intel knew what was coming and pulled the trigger to fool the public in a utterly pathetic move driven by desperation for damage-control to take cover for their own blame of Skylake-X 2.0 aka Cascade Lake-X.

Blatant yet honest, announced almost two weeks in advance …

AMD has learned their lessons the hard way in dealing with Intel's shonky benchmarks that most of the media are all too eager to promote with no questions asked. Smart move IMO.

Exactly. Most outlets still trying way to hard to picture anything Intel as way superior (or as at least even), as not as bad after all despite the huge drawbacks (heat-dissipation, power-draw, security-relates issues et al.) or generally downplay Intel's downsides a lot while overtly overemphasise the few given brownie points Intel still has left – Often up to the point that it becomes outright lying. At the same time many of those boys blow up any minor flaws on AMD's side eagerly and downplay many benefits.

Just look how many of the recent security-flaws didn't even were worth any mentioning, in regards to Intel-CPUs. These flaws are often dismissed, even when you message the editors in question. They're swept under the rug and they bring none whatsoever report on these flaws.

AMD knows that Intel still has a gazillion of paid shills claqueurs in their pockets on the media front, which always make sure that AMD gets the short end of the stick whenever possible … That's why it's so refreshing to come here after having just looked on other outlets. The die-hard factor for many, many in the media is sickening. One good turn deserves another, right?

… and I still predict since years, that a bunch of bigger OEMs, media-outlets, so-called 'independent' distributors and ODMs (who are, for some not-so-dubious reason magically bring everything Intel-only whenever possible) will actually die along the way as soon as the Intel-money dries up for when Intel can no longer afford these de-facto bribing. Many of them will die rather quickly, as they not only count on these payments but just rely on them to live at all and actually need them.

Mark my word, it's going to almost shock many enthusiasts to see those usual suspects quickly faltering and have trouble financing themselves (financially independent) to stay afloat, as soon as Intel can't afford nor wont hand out the usual payments … Though that's the price to pay for being dependent and with that have a rather comfortable living: You rise with your master – and die as well.

Changing release dates to Intels product release. AMD must feel confident with the product and want to steal the fanfares from Intel maybe. Comparing 2 new products at the same time of release should be interesting.

I think they're very confident. I mean, it's still Intel's infamous 10nm™ Raptor Lake is build upon, right? It's 10nm Enhanced SuperFin (Intel 7) against TSMC's 5nm.

A power-hog with well-fabricated benchmarks being dropped here and there and the usual FUD to prevent consumers from buying anything AMD in noble hope to secure some sales on one burning hand … Against the since years undisputed, quite cool delivered and chiplet-driven efficiency-king on the other hand, but now being fabbed on some even more superior process, which will unavoidably widen the gap between both even further.

---

Keep in mind, it's no coincidence that all these planted smoke-screenish leaks diversionary manoeuvres are of all things overwhelmingly from Geekbench. They're nice distraction tactics, to hide the fact that the artificially pumped numbers do not really represent reality nor any realistic workloads.

Since Intel favours such hugely staggering benchmarking numbers from GB as Intel heavily benefits from it. It perfectly meets Intel's turbo-mechanics which greatly benefit from jump-starting high-clocking sprint-like Turbo-modes for a couple of seconds, just to falter due to massive heat-dissipation quickly after. That is, when the heat-spreader has amassed too much heat – It can't be cooled down quickly enough and thus the heat removal is too limited to keep the die cool enough to prevent the CPU from throttling as a result from it (Thermal inertia is a nasty thing to play with on semiconductors!). Well, apart from the fact that Intel loves inflated numbers with AVX512 or other useless AI-mechanics virtually no home-user needs anyway …

If we look on leaks with other programs like Blender, Cinebench or other benchmarks (which, in stark contrast to Geekbench) taxing the CPU for a way longer period of time (for overcoming the sprint-like Turbo-modes' massive influence) and more thorough, the picture is quite clear: Intel's Raptor Lake looks like it's around ~8-10% faster at the core of it than Alderlake!

If we look at other benchmarks, it's always around 8–10% and tops out at around +12% (Alder Lake vs Raptor Lake). Keep in mind, thats often even benched with Alderlake on DDR4 while Raptor Lake being already on DDR5, so the actually real performance-uplift coming from the actual cores of Raptor Lake, is even smaller since backed with DDR5-throughput and -speed (and latency?).

Here are some benchmarks …

Spoiler: Make yourself a picture.

---

Furthermore, Raptor Lake comes on a socket which is already DOA (LGA1700) and a dead-end in terms of any future upgradability (Intel at it again). Meanwhile, AMD after years on the same socket brings a completely new socket for once (AM5), which is largely the only major drawback with Zen4 for customers. Alas, it seems to be the only path forward, especially technology- & connectivity-wise.

Speaking of which, AM4 was designed while having already Zen 1-4 in mind and thinking through the needs and reflect about future conceivable and expectable performance-impacts while on newer techniques – like AMD always does on sockets.

AMD virtually never really brought anything else apart from LTS-sockets (Long-term support) ever since. AMD always had LTS-sockets like AM4 or the AM3/+, replacing AM2/+, replacing 939 and so forth.

For sure PCi-Express 4.0 (and possibly even 5.0) – as obvious as its get – was also perfectly in mind on them when designing AM4 back then. And thus, AMD definitely did make sure it won't limit PCi-E 4.0 either when given chipsets would feature it anyway.

In the past, AMD always defined a socket more like in terms of some mechanical socket itself and some IC-socket, rather than speaking of the interconnect, protocols and chipsets which needs to be attached to it (that's the way Intel seems to see a socket; as soon as it needs to advance, it's limiting). AMD's understanding of a socket always was more like a universal and independent mechanical linking interconnection, which protocols and linked-through chipsets and data-busses could be changed at any time whenever the given need for it would arise. Like a lamp-socket, which light-bulb one day you run using AC, and the other day using DC-current. Then the bulb is switched against a higher-wattage rated one, and as a result the power is increased. Still, the socket stays the same since it just can.

Putting their north-bridge virtually out of their CPUs back to the motherboard again (on Ryzen 3xxx-boards), should pretty tell what I'm pointing at – It's a perfect example of how AMD's understanding of a socket didn't even changed a bit throughout the decades.

Even archiving backward-compatibility on AM4 with Zen2 was some incredible stunt AMD pulled. They've changed the process thrice, switched from one node to another, threw over the whole way Ryzen 3xxx is connected – and still maintained the identical socket.

AMD on the other hand, you can't really blame them here like at all, objectively speaking …
Since they managed to go from PCI-Express 2.0 over PCi-E 3.0 to PCi-E 4.0, kinda shifted the North-bridge in and out again and went from a low-power Dual-core up to a quite thirsty Hexa(kai)deca-core CPU (16-Core) over two completely different µArchs (Bulldozer-class Bristol Ridge, Zen) spanning many generations over several years – all on the very same socket, which was finalised already in the middle of 2015.

Heck, AM4 was specced out in 2015, was released at the IFA 2016 in Berlin, Germany with their Bristol Ridge-APUs in mid 2016, and AM4-boards could be bought in greater numbers already in fall 2016 – The later upcoming Ryzen SKUs in January 2017 taking its seats, when Ryzen 1st Gen was released. AM4 is now at least seven years old, just shifting into its eighth year. In fact, AM4 is that old, that it debuted with AMD's Bulldozer-descendant Bristol Ridge, whereas its Excavator-architecture's cores/module-assembly is some direct offspring of Bulldozer and its follow-ups.

Meanwhile AMD did basically everything what they could to make AM4 a real and true LTS-socket and make it last as long as possible. Well, … contrary to those backstabbing OEMs, who sabotaged most of of AMD's moves on longevity for their customers to make the socket last as short-termed as possible – 'Accidentally' forgetting using higher-sized flash-ROMs for BIOSs when those were not only supposed but told to do so or were always slow on purpose to bring any BIOS-updates for older boards to eventually kill just months-old boards as quick as possibly (in order to sell you new ones).

Having said all of the above, it's highly questionable that AM4 could be that twisted (again) with, to allow anything PCI-Express 5.0 and DDR5 (and now even Graphics atop again!) – likely too many changes and too few pins to overcome the socket's limitations.

tl;dr: AM4 debuted already with Bristol Ridge, Excavator was a Dozer-class CPU-arch. Zen is already the 2nd µArch running on AM4.

I think this is a smart move for AMD, Intel has a history of using any tactic in their PRs to trash AMD. From glue comments to the selective performance tests with graphs out of proportion bars to the pricing competions. We also know that AMD is not innocent when it comes to PRs either, so this new launch campaign from both sides should be worthy of a bucket of extra buttery popcorn and Snow-caps.

Just look what Intel did back then when Threadripper 3xxx launched. Intel pulled their embargo just hours before AMD's release, so that reviewers around the globe could not release given AMD-benchmarks in comparison – Intel willfully tricked the public and even the media to pull a stunt for them, in order to avoid the blasting competition.

Intel they knew it would be a bloody disaster … so what had to come, finally came. Jacking up cores, The Ripper™ came for a third (and final) time and most definitely killed the heart of Intel's HEDT pretty fast. It's just that Intel couldn't decide whether or not taking some narcotics for the sake of numbness beforehand, or actually really wanted to attend TR3's bloodbath in person – seems they just took the pill with that price-cut – to make it less bloody.

Now the AMD announcement date is confirmed on 29 Aug, we will see.

Yup, we'll se.. Blatantly delayed yet honest, announced almost two weeks in advance. Not like the desperately lame embargo-trick Intel pulled back then with Threadripper.
Also, I think it should do it for AMD logistically. Being able to amass given SKUs in securing proper volume for the launch afterwards, to prevent short supply of Ryzen 7000. Since we all know that the actual demand will be there without doubt. And since AMD supplies a good chunk of the market's volume by now, well …