notbOh. This is interesting.
Nevertheless, no AM4 Zen CPU to date is ECC-certified.

IMO we'll see AM4 EPYC CPUs - competitors for Xeon E (LGA1151). There was some talk about EPYC APU as well.
That would be rather nice.

Actually all Ryzens support ECC. The only difference is, Ryzen Pro is validated for ECC (certified if you wish) and is an OEM. Regular Ryzen CPUs aren't validated but still support ECC.

Vlada011Intel harder and harder compete with new versions of AMD Zen Core.

When you add price on top really become ugly investing in Intel 10 core example, or even i9-9900K.

Why so gloomy? This is when the fun starts, when Intel and AMD are close enough to create real competition.
When Zen 2 launches, Intel will adjust their prices, and will soon replace the Coffee Lake (2) with Comet Lake, which may not be a shiny new architecture, but still a very good contender in the market. Intel have really good margins on i7-9700K/i9-9900K, and their production capacity reserved for 14nm CPUs is the highest yet, so we can look forward to an autumn with price drops for good 8-cores and good supplies.

ratirtActually all Ryzens support ECC. The only difference is, Ryzen Pro is validated for ECC (certified if you wish) and is an OEM. Regular Ryzen CPUs aren't validated but still support ECC.

Pretty much all x86 CPUs support ECC, i.e. you can force them to work in ECC mode. People successfully booted Pentiums in ECC (despite them being officially locked).

ECC has to be verified to make sense for a corporate user. Ryzens aren't verified (PRO neither - someone lied to you :-)).

The fact is Ryzen & TR work with ECC ram, with certain boards. You can't dance around it, heck Intel disables many feature in their unlocked CPUs - how cheap is that?

You might be able to get a memory controller to run a configuration it's not certified for, but that doesn't mean you are guaranteed to make it work or that it will remain working over time. There are even examples of motherboards which lets Skylake-X CPUs run registered memory. When a CPU have a feature that is not enabled, it doesn't mean it wouldn't work, it means it's not tested, so you don't know if it will work and work reliably throughout the lifetime of the product. ECC is a feature to increase reliability, and also requires more expensive memory. Running ECC memory on an "unsupported" system makes no sense, even if you can sort-of get it working.

efikkanbut that doesn't mean you are guaranteed to make it work or that it will remain working over time.

Rubbish. When Intel/AMD/ARM/Etc. state that a memory type will work it's because they have internally certified that will work with 100% reliability. Otherwise they physically remove the ability for it to run in ECC mode.

notbPretty much all x86 CPUs support ECC, i.e. you can force them to work in ECC mode. People successfully booted Pentiums in ECC (despite them being officially locked).

ECC has to be verified to make sense for a corporate user. Ryzens aren't verified (PRO neither - someone lied to you :)).

Intel has ECC Disabled for desktop CPUs. AMD never did this. for Intel, ECC is exclusive only for server CPUs for example for which you have to pay a lot more than a desktop CPU.

here's some stuff you may want to read.
www.hardwarecanucks.com/forum/hardware-canucks-reviews/75030-ecc-memory-amds-ryzen-deep-dive.html
www.overclock3d.net/news/cpu_mainboard/amd_confirms_that_ryzen_supports_ecc_memory/1

efikkanYou might be able to get a memory controller to run a configuration it's not certified for, but that doesn't mean you are guaranteed to make it work or that it will remain working over time. There are even examples of motherboards which lets Skylake-X CPUs run registered memory. When a CPU have a feature that is not enabled, it doesn't mean it wouldn't work, it means it's not tested, so you don't know if it will work and work reliably throughout the lifetime of the product. ECC is a feature to increase reliability, and also requires more expensive memory. Running ECC memory on an "unsupported" system makes no sense, even if you can sort-of get it working.

Skylake-X doesn't support ECC only Xeons do.
www.pugetsystems.com/labs/hpc/Intel-Skylake-X-vs-Skylake-W-1082/

lexluthermiesterRubbish. When Intel/AMD/ARM/Etc. state that a memory type will work it's because they have internally certified that will work with 100% reliability. Otherwise they physically remove the ability for it to run in ECC mode.

You're wrong. While the design supports it, they haven't validated the sample.
It may still fully work, but it's not guaranteed.

ratirtSkylake-X doesn't support ECC only Xeons do.

And I never said they did.;)
ECC and registered memory are two different things.
Who knows, perhaps even ECC can be enabled if the BIOS wants to.

notbOne could think that people on a "computer enthusiast forum" would know how to make a screenshot.

Not sure if you are talking to me about that, but if you are, then clearly you can’t read and understand a graph ‍♂️

Berfs1Not sure if you are talking to me about that, but if you are, then clearly you can’t read and understand a graph ‍♂

He was poking at the circa 1990s screencap you posted as opposed to a print screen which, unlike that screencap, is clear and easier to read.

R0H1TNo bitcoins, Paypal? Just to confirm ~ no 3xxx SKU will have a base clock of 4GHz or above :toast:

Here we go:

wccftech.com/amd-ryzen-3000-16-core-7nm-zen-2-cpu-leak-up-to-4-2-ghz-es/

Ryzen eng. sample, 16-core, 3.3Ghz BASE CLOCK. The leaker is legit, so the eng. sample is real. Now, remember the numpty Adored was claiming the 16-core Ryzen 3000 has a BASE CLOCK of 4.3Ghz.

So according to him, from this eng. sample, another....1Ghz frequency is going to be added to retail chips, right!??! Madness. Now, at best, we'll get another 300Mhz on top of this sample's base clock. Please have your £10 paypal payment ready to send me once this is confirmed. Cheers :clap:

I really wouldn't read too much in to clock speeds of engineering samples, not unless we have some additional context about the quality of it, as engineering samples can be all over the place, ranging from very low clocks to golden samples, all depending on their intended usage.
When you see them demonstrate benchmarks in public, like on CES, it usually represents their target performance (in the cherry-picked benchmark), but final clocks are only set after the final stepping arrives, so there can be some deviation in either direction.

Shatun_BearHere we go:

wccftech.com/amd-ryzen-3000-16-core-7nm-zen-2-cpu-leak-up-to-4-2-ghz-es/

Ryzen eng. sample, 16-core, 3.3Ghz BASE CLOCK. The leaker is legit, so the eng. sample is real. Now, remember the numpty Adored was claiming the 16-core Ryzen 3000 has a BASE CLOCK of 4.3Ghz.

So according to him, from this eng. sample, another....1Ghz frequency is going to be added to retail chips, right!??! Madness. Now, at best, we'll get another 300Mhz on top of this sample's base clock. Please have your £10 paypal payment ready to send me once this is confirmed. Cheers :clap:

That's just one chip, I thought we had a bet on all 3xxx SKU :confused:

efikkanYou're wrong. While the design supports it, they haven't validated the sample.
It may still fully work, but it's not guaranteed.

Validation & certifications are costly, without them it doesn't mean that ECC (memory) don't work or it stops working randomly ~
www.gsmarena.com/oneplus_7_and_7_pro_wont_have_an_ip_rating_for_water_and_dust_resistance_because_its_expensive-news-36889.php

Also AMD isn't known to block out unlocked features, unlike their two competitors ~
www.guru3d.com/articles_pages/amd_athlon_220ge_and_240ge_review,27.html

R0H1TThat's just one chip, I thought we had a bet on all 3xxx SKU :confused:
Validation & certifications are costly, without them it doesn't mean that ECC (memory) don't work or it stops working randomly ~
www.gsmarena.com/oneplus_7_and_7_pro_wont_have_an_ip_rating_for_water_and_dust_resistance_because_its_expensive-news-36889.php

Also AMD isn't known to block out unlocked features, unlike their two competitors ~
www.guru3d.com/articles_pages/amd_athlon_220ge_and_240ge_review,27.html

Yes, none of them will have a base clock of 4Ghz or above. You need to realise eng. samples are indicative of retail silicon, and tend to only deviate by about ~+300Mhz~ from final. Release of this chip is only about 4-5 months away.

This real leak here is important as the frequencies in this TPU news story come from that fraudster Adored. If he was wrong about this chip, then the rest of that horse crap SKU list with all the clocks is wrong, and by extension hopes of base clocks over 4Ghz are wrong. Anyway this will be revealed soon.

You remember the earliest zen ES leaks? The 1800x exceeded the boost clocks by 700~800Mhz & base clocks by a similar margin IIRC. You might well be right but let's wait for the entire lineup to be revealed first.

Also it can be the low clock 16C and not the top SKU.

R0H1TYou remember the earliest zen ES leaks? The 1800x exceeded the boost clocks by 700~800Mhz & base clocks by a similar margin IIRC. You might well be right but let's wait for the entire lineup to be revealed first.

I don't remember there being such a gap. And we're too close to release for anything more than 400Mhz to be added. At latest 16-core release October.

HwGeekAlso it can be the low clock 16C and not the top SKU.

Unlikely. These are about the numbers you'd expect. But to be clear, my issue all along is fake base clock numbers. I expect boost up to 4.7Ghz.

Here's one from TPU, you'll find more if you look closer ~ AMD "Summit Ridge" ZEN CPU at 2.80 GHz Beats 3.40 GHz Core i5-4670K

efikkanYou're wrong. While the design supports it, they haven't validated the sample.
It may still fully work, but it's not guaranteed.

Incorrect. When a manufacturer makes a product with support for a certain function and that function is made available to the public, you can bet your life on the fact that they have tested it completely.

*cough, cough*
www.techpowerup.com/255386/amd-ryzen-9-3000-is-a-16-core-socket-am4-beast

lexluthermiesterRubbish. When Intel/AMD/ARM/Etc. state that a memory type will work it's because they have internally certified that will with 100% reliability. Otherwise they physically remove the ability for it to run.

And AMD does not state Ryzens support ECC, does it?

ratirtIntel has ECC Disabled for desktop CPUs. AMD never did this. for Intel, ECC is exclusive only for server CPUs for example for which you have to pay a lot more than a desktop CPU.

here's some stuff you may want to read.

AMD doesn't disable ECC, but it doesn't mean it works as it should. It's just there.

lexluthermiesterIncorrect. When a manufacturer makes a product with support for a certain function and that function is made available to the public, you can bet your life on the fact that they have tested it completely.

Like the RNG in Excavator?

You don't understand enterprise computing - I've told you that many times.
ECC has to be validated to make sense. Just like helmets and ropes are certified/attested/rated to be used in a construction zone. It doesn't mean non-rated helmets don't protect your head.
What's the point of an untested security feature?

R0H1TThat's just one chip, I thought we had a bet on all 3xxx SKU :confused:
Validation & certifications are costly, without them it doesn't mean that ECC (memory) don't work or it stops working randomly ~

Without them it means you can't really expect ECC to work. And no one is liable when it stops.
That's the whole point of certification. It is important not in the 99% of time a feature works, but in the 1% time it doesn't.
Certified ECC systems sometimes don't work properly, just like a certified airbag doesn't always save your life in a crash. But until someone gives you a guarantee that an airbag should work in a particular way, it's just a small bomb with a baloon. WTF would you willingly put a bomb in your car?

Berfs1Not sure if you are talking to me about that, but if you are, then clearly you can’t read and understand a graph ‍♂

First of all: they aren't graphs. A graph is a graphical representation of information, for that you need things like properly described axes. There's no horizontal axis in your case - maybe you assumed/checked it is time - the viewer doesn't know. In one of the photos you've missed the vertical axis as well.
Second: I was talking about the way you've shared this - as photos made with a smartphone. Why?
Also this:
And that's all pretty objective and obvious. I could now start making fun of your forecasting, but as you see - I don't. I'm in "nice mode" today. But it may change tomorrow, so weight your words carefully. ;-)

R0H1TThe fact is Ryzen & TR work with ECC ram, with certain boards. You can't dance around it, heck Intel disables many feature in their unlocked CPUs - how cheap is that?

Intel has a very enterprise-oriented approach, with all features being pretty well documented. They ride on an image of being a solid enterprise partner. They can't afford to put ECC in CPUs that may not support it properly.
AMD has a different target client structure and they can afford not to properly describe what features a product has. We had a nice discussion about this lately in the NVENC thread (with AMD you don't know what features are supported by GPU, there's no documentation).

Honestly, I think you guys even like it. I think you like being forced to test and search and ask on forums instead of just checking in the datasheet.
But assuming AMD is hoping for a larger share of business clients, they'll have to really focus on more than just performance.

lexluthermiesterIncorrect. When a manufacturer makes a product with support for a certain function and that function is made available to the public, you can bet your life on the fact that they have tested it completely.

I think the same. Besides Ryzen TR's are validated. It is even mentioned on AMD's webpage. So if TR is validated then sure as hell other chips will work as if they are created to work with ECC. Tested for sure validated not necessarily.

notbAMD doesn't disable ECC, but it doesn't mean it works as it should. It's just there.

It does support and works the problem is finding motherboard which will allow using ECC ram modules.

notbAnd AMD does not state Ryzens support ECC, does it?

Actually AMD stated that all ECC features on Ryzen CPUs are working fine and never been disabled.
www.overclock3d.net/news/cpu_mainboard/amd_confirms_that_ryzen_supports_ecc_memory/1
www.hardwarecanucks.com/forum/hardware-canucks-reviews/75030-ecc-memory-amds-ryzen-deep-dive.html

notbYou don't understand enterprise computing - I've told you that many times.

You are welcome to your opinion. Doesn't mean you are correct.

notbWhat's the point of an untested security feature?

Without them it means you can't really expect ECC to work. And no one is liable when it stops.
That's the whole point of certification. It is important not in the 99% of time a feature works, but in the 1% time it doesn't.
Certified ECC systems sometimes don't work properly, just like a certified airbag doesn't always save your life in a crash. But until someone gives you a guarantee that an airbag should work in a particular way, it's just a small bomb with a baloon. WTF would you willingly put a bomb in your car?

First off, ECC isn't a security feature, it's a data integrity feature, unless everything I've ever read about it is wrong. It is for when the electronics make a mistake and mangle a bit, it can then be corrected instead of possibly corrupting something. Unless you mean security from hardware failure??

Second off, comparing it with a safety device (airbags, also helmets) makes absolutely no sense. That's not even apples and oranges. It's apples and mutton. An airbag protects you in an accident, ECC corrects hardware data errors should they occur.

I'm pretty sure ECC will work properly on Ryzen if a motherboard supports it, and not if it doesn't. I'm guessing AMD didn't spec Ryzen as supporting ECC to avoid the inevitable lawsuit when someone snags a non-ECC or no-brand "ECC" mobo and it doesn't work right, even though the Ryzen would operate according to ECC spec (that's just my guess, maybe it actually has a flaw or something?). Pretty sure you could find out by checking what the specs are on the Ryzen memory controller. *shrug*

ECC is normally only used in applications where downtime is unacceptable, like data centers, mission critical (like maybe military, medical, perhaps ATC and possibly AI driving?) applications, as it's more expensive than non-ECC, even though once upon a time, ECC was the norm (if I remember my computer history right).

en.wikipedia.org/wiki/ECC_memory

0x6A7232First off, ECC isn't a security feature, it's a data integrity feature, unless everything I've ever read about it is wrong.

Well, maybe it's a broader definition of "security" than people here may be used to. Sorry. Risk management comes out once in a while.

Unless you mean security from hardware failure??

Yeah, I meant "security" as the general idea of mitigating risk.

An airbag protects you in an accident, ECC corrects hardware data errors should they occur.

What if airbags are designed on a computer without ECC and we're so unlucky that the simulation data got corrupted or lost?
I know this may sound funny, but that's what risk management is usually about: mitigating the risk of very rare but significant events.

Yes, corruption/instability stemming from a class of RAM errors that ECC targets is very rare. But it can happen and we have a technology that makes it few orders of magnitude less likely. So this technology became a standard in production systems.

I'm pretty sure ECC will work properly on Ryzen if a motherboard supports it, and not if it doesn't.

OK, you may be sure. I'm pretty sure it won't. These are just opinions.

I was speaking from a PoV of an enterprise, so the party actually interested in ECC memory.
The system has to officially support ECC, i.e. someone has to take responsibility. And that's the whole point: responsibility.

Maybe Ryzen today can work in ECC mode, we don't know that. And honestly, do we really know whether Threadripper, EPYC or Xeon support ECC properly? No, we don't.
But one CPU has an "ECC validated" sticker and one doesn't. And that sticker changes everything.

ECC is normally only used in applications where downtime is unacceptable, like data centers, mission critical (like maybe military, medical, perhaps ATC and possibly AI driving?) applications, as it's more expensive than non-ECC, even though once upon a time, ECC was the norm (if I remember my computer history right).

No, ECC is required in virtually all production systems in large enterprises.

Also, you have a very military understanding of something being "mission critical" (gaming much? ;-)).
A mission critical system is any system essential for an organization to perform its core tasks.
For example the system responsible for selling products is "mission critical", because selling is the most important activity in a company. The database that holds client or sales data is critical as well.
If your company designs fans on CAD workstations, they may also be considered "production" and "mission critical", i.e. it's very unlikely this job will be given to ordinary office desktops - even if they're fast enough.