Summary of Benchmark Results: 3900X vs 9900k

RamonGTP said:
I have the game and never played it. What’s the issue? I’ll try it out tonight.
Click to expand...

The issue is that Destiny 2 can't run on any system with a Ryzen 3000 series CPU installed in it. The game will appear to launch, but won't actually start. You'll look into task manager and see the process. It will consume little CPU and about 150MB of RAM. That's all it ever does. I explained why this is above.
 
Well I'm downloading the nearly 100GB game now since curiosity is getting the best of me. I do have the latest beta bios from ASUS installed but don't know if that's the one with the "fix"
 
Sure enough... Game won't run. Glad it's not one I play but hopefully they hurry with the fix. That said, it's already been a good while since these CPU's have been released and I doubt this would have happened with Intel or nVidia.
 
RamonGTP said:
Well I'm downloading the nearly 100GB game now since curiosity is getting the best of me. I do have the latest beta bios from ASUS installed but don't know if that's the one with the "fix"
Click to expand...

It isn't. They pulled the one that had the fix from its website as soon as they discovered it caused other issues.

RamonGTP said:
Sure enough... Game won't run. Glad it's not one I play but hopefully they hurry with the fix. That said, it's already been a good while since these CPU's have been released and I doubt this would have happened with Intel or nVidia.
Click to expand...

I told you. I actually tried to add the game to our review and couldn't make it work. And your right, it almost certainly wouldn't have happened with an Intel CPU.
 
DuronBurgerMan said:
Let the desktop users be unofficial beta testers
Click to expand...
Dan_D said:
it almost certainly wouldn't have happened with an Intel CPU
Click to expand...
The thing that gets me is: Is it really possible AMD did not try any of the latest Linux distros on Zen 2 before release and not notice the problems with systemd and its use of the RdRAND instruction? I can understand not testing Destiny 2. But Linux? I find that... shocking.
Unless they knew about it but decided to release anyway. Either way, it's a little bit scary.
It pushes up my back-of-the-mind worry that there is a lurking hardware design issue that may have been missed because these CPUs are so complex.
 
Azrak said:
The thing that gets me is: Is it really possible AMD did not try any of the latest Linux distros on Zen 2 before release and not notice the problems with systemd and its use of the RdRAND instruction? I can understand not testing Destiny 2. But Linux? I find that... shocking.
Unless they knew about it but decided to release anyway. Either way, it's a little bit scary.
It pushes up my back-of-the-mind worry that there is a lurking hardware design issue that may have been missed because these CPUs are so complex.
Click to expand...

This gets me too. This is why I have been so vocal about why I am not going to AMD this round.
 
Azrak said:
The thing that gets me is: Is it really possible AMD did not try any of the latest Linux distros on Zen 2 before release and not notice the problems with systemd and its use of the RdRAND instruction? I can understand not testing Destiny 2. But Linux? I find that... shocking.
Unless they knew about it but decided to release anyway. Either way, it's a little bit scary.
It pushes up my back-of-the-mind worry that there is a lurking hardware design issue that may have been missed because these CPUs are so complex.
Click to expand...

Given how many distros there are is anyone really surprised they could have missed something like that? Even the Destiny 2 thing surprises me. It's popular enough to have warranted testing. However, AMD probably couldn't have imagined a game would use the RdRAND instruction like that.
 
Dan_D said:
I thought some distros worked with it. I play Destiny 2, I do not run Linux these days.
Click to expand...

Yeah some distros work because they use different versions of the kernel.
 
Last edited:
If it's a systemd thing there are very few mainstream distros that have not switched. Even testing the top four should have caught something.

Considering the first gen Ryzen had linux issues I'm very surprised they got hit with another major linux bug. (The linux issue is what kept me off Ryzen to begin with.)
 
https://bugs.launchpad.net/ubuntu/+source/systemd/+bug/1835809

* The fix itself is very small, it ignores known to be faulty random values returned by the rdrand instruction and use a different random source. Those values can still be returned by a properly working rdrand implementation in 2 in 2^32 cases on 32 bit arches and in 2 in 2^64 cases on 64 bit arches, but the fallback to the other random source ensures that in those rare occasions a random number can be generated."
Click to expand...

A bigger problem is that some things (eg pfsense/openSSL) can use RDRAND to generate random numbers for security purposes. While it is not a “true” random source, it definitely is close to it. I wonder how they can fix this in the AGESA...

Heck I wonder if browsers use it for ssl.. I would think so.. could even affect the random number generation hardware for other parts of the cpu..

This explains how it could be a much bigger sec threat... Forget spectre and meltdown...

https://linuxreviews.org/AMD_Ryzen_3000_series_CPUs_can't_do_Random_on_boot_causing_Boot_Failure_on_newer_Linux_distributions
 
Last edited:
  • Like
Reactions: STEM
like this
Algrim said:
If it's a systemd thing there are very few mainstream distros that have not switched. Even testing the top four should have caught something.

Considering the first gen Ryzen had linux issues I'm very surprised they got hit with another major linux bug. (The linux issue is what kept me off Ryzen to begin with.)
Click to expand...

That's a fair point.
 
Wow, the RDRAND bug is a major one. The good news is that Ryzen 1000 and 2000 don't have it. The bad news for AMD is that they probably already shipped Rome CPUs with it. But ... But ... If they use the same chiplets for Enterprise class CPUs, wouldn't they find the bug during testing / validation / qualification?

The good news is that they can fix it at the CCX level, then produce new chiplets with the bug fixed.

Man, I'm sure happy I didn't jump on the Ryzen 3000 bandwagon.

Early Adopters = Beta Testers
 
STEM said:
The good news is that they can fix it at the CCX level, then produce new chiplets with the bug fixed.
Click to expand...


The question I have now is "Will AMD own up to the problem?"

When the other Linux compile bug showed up, they fixed it in the CCX - the problem however is that consumers didn't know which units had the bug, and which units didn't. AMD did a "if this affects you, then send the chip back and we'll replace it"... but that was months after it was discovered, and at that time retail channels still had old stock.

This is bad form IMO as then they're selling faulty stock.

The other question I have is "If they fix it in AGESA, how can we know it's as random as it is supposed to be and that they are not just doing pseudo random?"


Do keep in mind that user-space programs may also be affected since Ryzen 3000 series CPUs will always fail to produce random data when RDRAND is called. User-space software will typically use /dev/urandom but there could be some rare pieces of software which do call RDRAND directly. It is interesting to note that OpenSSL used to make frequent calls to RDRAND. This was changed years ago since someone with a buggy AMD APU happened to notice a problem using openssl after a system suspend.
Click to expand...

It is interesting that I haven't seen this reported on tech sites....
 
Last edited:
  • Like
Reactions: STEM
like this
Keljian said:
The question I have now is "Will AMD own up to the problem?"
Click to expand...

Keljian said:
AMD did a "if this affects you, then send the chip back and we'll replace it"... but that was months after it was discovered, and at that time retail channels still had old stock.
Click to expand...

I'm going to put on my conspiracy hat here for a second and say this:

What if AMD discovered the bug internally when it was already too late, and a bunch of chiplets had already been manufactured? What if they already fixed the problem and it doesn't show up in Rome CPUs? I couldn't dig up any info if Rome CPUs are affected or not by this. Anyway, if that's the case, then AMD turned really dark, which means that they are knowingly selling bad stock to get rid of it. They know early adopters will gobble up the new CPUs, and they also know that only a fraction of consumers will send in their flawed Ryzen 3000 CPU for replacement.

The bottom line is that you can no longer trust any corporation. AMD still benefits from this underdog image of a company that wants to sell you a better product at a lower price, but dammit, if only Intel wouldn't sabotage them all the time, right? So consumers will put up with the flaws because for the most part, they get a cheaper product. Well, at least they're not like NVIDIA to charge $1200 for a Space Invaders riddled GPU and then shut up about it, never admitting to anything being wrong. I wonder how long it will take AMD to get there...
 
The thing is, the USA is a litigious society. If they admit to a faulty product, they can be sued/subject to a class action.

Call me old fashioned, but I reckon they should just own up and make amends, none of this lawsuit stuff.

“We released a chip that is faulty, if you are affected by this or want to exchange it for one that isn’t, contact us and we will swap it, no questions asked”
 
Keljian said:
The thing is, the USA is a litigious society. If they admit to a faulty product, they can be sued/subject to a class action.

Call me old fashioned, but I reckon they should just own up and make amends, none of this lawsuit stuff.

“We released a chip that is faulty, if you are affected by this or want to exchange it for one that isn’t, contact us and we will swap it, no questions asked”
Click to expand...

Do you mean like Intel never owned up to their ginormous clusterfuck?

Anyway, with Ryzen 3000 they're not worried about any kind of lawsuit right now because not enough people have bought one yet. What they're concerned about, and the reason why AMD will never own up to it is stock price. As soon as AMD admits to anything, some Wallstreet analyst will recommend a downgrade to their stock price. Then investors follow like sheep. If people had a clue in what they're investing then two things would happen: 1) analysts would be out of a job - and - 2) "American Greed" would be out of material.

No worries, after YouTube "tech experts" reviewed and recommend Ryzen 3000, everyone followed and bought one. It's always out of stock. None of these "tech experts" will talk about the bug because they get freebies to review from tech companies. There is no longer such a thing as an "NVIDIA" or "Intel" only shill. Now they will shill for anyone willing to throw them a bone.
 
I think some of y'all are overreacting a bit. Yes, it was stupid and par-for-the-course with an AMD release. However, there was a microcode update that fixed the problem. The only reason it was pulled, AFAIK, is that it contained other bugs.

So a fix exists. It's a rare use case outside of Linux. And in a few weeks or a month or so I bet there is a BIOS fix available. In fact, I am skeptical that this is really a hardware problem at all, or else surely it would have come up in Epyc Rome testing. My money is on this scenario:

1. AMD rushes release. Is pushing a ton of different Agesa versions out in prep for the embargo drop.
2. Somebody failed to Q/A these damn things properly, and accidentally created this bug in one of the Agesa versions, and it persisted in the following releases because AMD was probably focused on fixing the pile of other BIOS issues.
3. Q/A didn't test Destiny 2, or a scratch Linux install (instead used an existing install).
4. Now we have problem.
 
DuronBurgerMan said:
I think some of y'all are overreacting a bit. Yes, it was stupid and par-for-the-course with an AMD release. However, there was a microcode update that fixed the problem. The only reason it was pulled, AFAIK, is that it contained other bugs.

So a fix exists. It's a rare use case outside of Linux. And in a few weeks or a month or so I bet there is a BIOS fix available. In fact, I am skeptical that this is really a hardware problem at all, or else surely it would have come up in Epyc Rome testing. My money is on this scenario:

1. AMD rushes release. Is pushing a ton of different Agesa versions out in prep for the embargo drop.
2. Somebody failed to Q/A these damn things properly, and accidentally created this bug in one of the Agesa versions, and it persisted in the following releases because AMD was probably focused on fixing the pile of other BIOS issues.
3. Q/A didn't test Destiny 2, or a scratch Linux install (instead used an existing install).
4. Now we have problem.
Click to expand...

I hope you're right, but they seem to be taking an awfully long time with updates. Latest BIOS on ASUS site is from 7/5 and it's a beta.
 
RamonGTP said:
I hope you're right, but they seem to be taking an awfully long time with updates. Latest BIOS on ASUS site is from 7/5 and it's a beta.
Click to expand...

They had one ready, and pulled it at the last minute. This happened about a week ago. So they aren't sitting on their asses. Maybe someone just grew enough of a brain to realize fixing one bug, and then creating several others, was not a good way to solve this.

They need to take the time they need and do the next BIOS update right.
 
If the fix is in or around the corner - ? what is the question? Speaking of Intel and all their security problems for chips going back some time . . . Never mind. Is this the only bug known? Isn't that rather on the good side in the scheme of things, I think so but more could show up.
 
DuronBurgerMan said:
They had one ready, and pulled it at the last minute. This happened about a week ago. So they aren't sitting on their asses. Maybe someone just grew enough of a brain to realize fixing one bug, and then creating several others, was not a good way to solve this.

They need to take the time they need and do the next BIOS update right.
Click to expand...

Had to have been a lot longer than a week ago. I’ve had this system for 2 weeks and the first thing I did after installing Windows was update the bios. I’ve checked for updates every day since.
 
STEM said:
Do you mean like Intel never owned up to their ginormous clusterfuck?

Anyway, with Ryzen 3000 they're not worried about any kind of lawsuit right now because not enough people have bought one yet. What they're concerned about, and the reason why AMD will never own up to it is stock price. As soon as AMD admits to anything, some Wallstreet analyst will recommend a downgrade to their stock price. Then investors follow like sheep. If people had a clue in what they're investing then two things would happen: 1) analysts would be out of a job - and - 2) "American Greed" would be out of material.

No worries, after YouTube "tech experts" reviewed and recommend Ryzen 3000, everyone followed and bought one. It's always out of stock. None of these "tech experts" will talk about the bug because they get freebies to review from tech companies. There is no longer such a thing as an "NVIDIA" or "Intel" only shill. Now they will shill for anyone willing to throw them a bone.
Click to expand...

So much for your conspiracy theory, tin foil hat fiction you were trying to spin because now you are spinning it as fact. I would not be concerned, AMD will take care of it's customers and to be honest, if I wanted to, I bet they would replace my R7 1700 that I bought at release and has the Linux bug that was mentioned. However, I do not use Linux on that machine and therefore, have not had any issues with that particular machine, which is why I have not bothered to have it replaced.

The YouTube reviewers are accurate in their assessments, regardless of the fact that you clearly do not like the results. Was this a mess up by AMD with the RDRAND thing? Yes. Is it the end of the world? No, they will fix it but, it has not even been a month since release. Also, take a look at the seriousness with which motherboard manufacturers are taking with the X570 chipset boards now, that is a clear indication that the Ryzen 3000 series are the real deal.
 
DuronBurgerMan said:
You have two separate claims here.

1. "9900k is still the better overall CPU".

No. This is false. For "overall" or "general" results, we must look at all applications/use-cases. The 3900X wins a majority of them no matter how you slice it and dice it, and where it loses, the loss margin is small. Zen 2 has caught up enough in single thread performance that the remaining deficit no longer outweighs both raw core count, and superior SMT yield. In some throughput cases, AMD even wins outright in single thread performance. As an "overall" CPU, the 3900X is clearly superior. However in specific use cases, gaming being the most obvious, the 9900k is superior. Some video applications can benefit from QuickSync or high single thread performance, and this allows Intel a few victories in some video editing scenarios, but the 3900X's raw core count usually wins the day here as well.

I agree, however, that the 9900k is not a fuck up, or a P4 situation. It's a great CPU. If you're building the ultimate gaming rig (and don't intend to do much else with it) you should buy it. If the teething issues of new launch/platform bother you, then sure, the 9900k is a great alternative there too.

2. "Could crush AMD without effort."

Perhaps. Perhaps not. I've mentioned before that AMD and Intel have very different approaches from a business level. Intel has a lot of money, but it also has a lot of baggage. AMD's advantage is that some smart wizard in the basement can come up with some hare-brained idea and run with it, resulting in occasional flares of unexpected brilliance at AMD. AMD's weakness is exactly the same thing. Look at Bulldozer as a hare-brained idea that turned to crap. Look at Zen 2's chiplet design as a hare-brained idea that turned to gold. AMD is, at its core, inconsistent, but not stupid and not easy to get rid of on a whim.



Correct. The HEDT parts can't just come down to the mainstream because of power requirements and a host of other things. This is where AMD's CCX and later, CCD/chiplet concept proved excellent. This allowed for rapid, cheap, and (relatively) high clock core scaling. It was a great idea. Intel doesn't have an equivalent yet, and has continued to rely on the fact that they've really perfected their 14nm process, and really perfected Skylake. That and AMD's single thread weakness kept AMD from really running away with a victory. But the single thread weakness is almost gone - and is now almost solely due to clockspeed. And I doubt we see much higher - if any higher - clocks from Skylake derivatives. If Intel drops a 10 core - and it seems like they will - AMD has a 16 core answer. The status quo is preserved. The status quo being, as per the first part of this reply, at almost any given price tier Zen 2 competes in, it's faster generally (but not necessarily specifically).
Click to expand...

For a Ultimate gaming rig where you are not doing much else, I would say the 9700K would be the best option. The 9900K and 9700K typically match on the performance charts, as far as gaming is concerned.
 
ManofGod said:
For a Ultimate gaming rig where you are not doing much else, I would say the 9700K would be the best option. The 9900K and 9700K typically match on the performance charts, as far as gaming is concerned.
Click to expand...

I'd disagree. The additional threads of the 9900k aren't all that useful for gaming today. But I suspect they might be in the future.
 
RamonGTP said:
Had to have been a lot longer than a week ago. I’ve had this system for 2 weeks and the first thing I did after installing Windows was update the bios. I’ve checked for updates every day since.
Click to expand...

The update was pulled before it was available to the public, but after it got sent to the motherboard vendors. Although I've read some reports that some folks actually got it (but grain of salt - don't believe everything you hear on the Internet).
 
RamonGTP said:
I hope you're right, but they seem to be taking an awfully long time with updates. Latest BIOS on ASUS site is from 7/5 and it's a beta.
Click to expand...

You can go to the ROG forum and you can find beta bioses if you want to try them. I am using a beta bios for my x370 board and it works good.
 
DuronBurgerMan said:
FDIV Pentium bug did. So did the 1.13 Ghz Pentium 3. Intel has had some fuckups too, they are just less common than AMD's.
Click to expand...

To be fair, you are talking around 20-26years ago, that is ancient history in computer terms.

That said, the N3710 (and lower) atom processors have major Linux issues to this day which result in random crashing due to a hardware bug in sleep states...

https://askubuntu.com/questions/803640/system-freezes-completely-with-intel-bay-trail/803649#803649
 
Last edited:
Keljian said:
To be fair, you are talking around 20-26years ago, that is ancient history in computer terms.

That said, the N3710 (and lower) atom processors have major Linux issues to this day which result in random crashing due to a hardware bug in sleep states...

https://askubuntu.com/questions/803640/system-freezes-completely-with-intel-bay-trail/803649#803649
Click to expand...
This is a really a major issue, as many ATOM families are affected and plenty used in NAS. If you use one of those it will not crash but permanently die after running for 6 months and two years depending on the CPU. No possibile recovery if you encrypted your data and that kind of things. And everything on proprietary Linux. And mind that plenty are Raid 1 because people were scared of the hard drive failing while the rest of the hardware was supposed to work forever. Intel should have done something about but they didn't. They failed to inform, looked into little excuses and made BIOS patches that solved nothing and proved it was pure CPu hardware failure. That even didn't stop them from selling those CPU and make new ones with the exact same bug. One could wonder if they did it on purpose as for planned obsolescence.
 
Dan_D said:
I wouldn't bet on that. All Ryzen's pretty much hit the same wall.
Click to expand...

I purchased a brand new R5 1600 from Microcenter back in March of this year and it can hit 4.0 Ghz without issue. The original 1600's, not so much. My 1700 is one of the originals and I can only get 3.7 to 3.8 GHz out of it but, a new 1700 would probably get better than that, since it would be a much more recently produced one. I never got the 3.9 to 4.0 that many were able to obtain.
 
ManofGod said:
I purchased a brand new R5 1600 from Microcenter back in March of this year and it can hit 4.0 Ghz without issue. The original 1600's, not so much. My 1700 is one of the originals and I can only get 3.7 to 3.8 GHz out of it but, a new 1700 would probably get better than that, since it would be a much more recently produced one. I never got the 3.9 to 4.0 that many were able to obtain.
Click to expand...

I agree. Look at what 14+++++++++ is for intel. Skylake couldn't hit 5, but 3 years later 9900k can. December 2019 ryzen may well pick up 100mhz.
 
Randall Stephens said:
I agree. Look at what 14+++++++++ is for intel. Skylake couldn't hit 5, but 3 years later 9900k can. December 2019 ryzen may well pick up 100mhz.
Click to expand...

In September it should. The Ryzen 9 3950X is supposed to have a boost clock of 4.7GHz. That won't be on all cores of course, but there's your 100MHz. :)
 
Dan_D said:
In September it should. The Ryzen 9 3950X is supposed to have a boost clock of 4.7GHz. That won't be on all cores of course, but there's your 100MHz. :)
Click to expand...

LOL. Yes, it'll be the additional 100mhz, but that's just binning. I think once the process matures, we'll get a tiny itty bitty little bit more mhz out of the chiplets not related to straight binning :)
 
You must log in or register to reply here.
Back
Top