How many 3900x silicon lottery losers are going to upgrade to 3950x?

RamonGTP

Supreme [H]ardness
Joined
Nov 9, 2005
Messages
8,150
Curious how many silicon lottery losers like myself are going to roll the dice on a 3950x hoping for better luck and an additional 4 cores when they are released? I was hoping the next bios update would fix some of my issues with boost and exceptionally high idle voltages but everything appears to be identical. I'm lucky to hit 4.3GHz on my 3900x and very lucky to hit 4.4GHz for about half a second. My plan is to see what the general consensus is with the 3950x and if I see a good number of users having better luck and closer advertised boost clocks with it. If that's the case, i'll likely upgrade.

Thoughts?
 
Does it do what you bought it for?

Yes, but that extra 400mhz of single core boost could be quite beneficial in certain games. Not really asking if I should or shouldn't, that's going to be dictated by how well the average 3950x boosts. I'm mainly curious if other 3900x owners are eyeing a 3950x for similar reasons.
 
You're asking what others will do.

Are they going to need more cores? if so, why would they buy a 3900x in the first place?:confused::confused::confused::confused::confused::confused::confused::confused::confused::confused::confused::confused::confused::confused::confused::confused::confused::confused::confused::confused::confused::confused::confused::confused::confused::confused::confused::confused::confused::confused::confused::confused::confused::confused::confused::confused::confused::confused::confused::confused::confused::confused::confused::confused::confused::confused::confused::confused:

I guess the thread title and everything I said before “AND an additional 4 cores” was too confusing. Let’s just simplify this and say this “complicated” question wasn’t directed at you.
 
Yes, but that extra 400mhz of single core boost could be quite beneficial in certain games. .

When people say games benefit from high single core performance, they don't mean on only one core. Modern games need high core performance on multiple cores, as soon as you start using multiple cores that single core boost is out the window.
 
I don't seem to be able to hit 900 on the fclock which is disappointing, but a 3950 is not in my future.
 
900 or 1900?

Sorry, 1900. I can hit 1866 with every single other setting on the mobo at default settings. However, I can't hit 1900 with SOC at 1.1.

So far, the highest boost I've seen on a cpu core is 4438 I believe.

I'm running an Asus Hero 8 and a 32 GB 2 x16 b-die memory kit. The board won't even boot with the memory at XMP settings (4000, 19-19-19-39 @ 1.35v), of course I'm looking for 3733 now based on my fclk.

None of the b-die presets on the mobo or what I can get from the Ryzen memory calculator work without more memory voltage than indicated. I'm going with recommended 3800 timings while I'm set to 3733 and I'm having to use 1.485v, where 1.45v is recommended.
 
Curious how many silicon lottery losers like myself are going to roll the dice on a 3950x hoping for better luck and an additional 4 cores when they are released? I was hoping the next bios update would fix some of my issues with boost and exceptionally high idle voltages but everything appears to be identical. I'm lucky to hit 4.3GHz on my 3900x and very lucky to hit 4.4GHz for about half a second. My plan is to see what the general consensus is with the 3950x and if I see a good number of users having better luck and closer advertised boost clocks with it. If that's the case, i'll likely upgrade.

Thoughts?

According to Siliconlottery.com's data, the more cores Ryzen 3000 has, the lesser the overclockability.
 

Attachments

  • Annotation 2019-08-19 002955.jpg
    Annotation 2019-08-19 002955.jpg
    85 KB · Views: 0
If you want to play that lottery for the extra 300-400mhz, you're probably better off buying an Intel or waiting for next year's 7nm+ which should *still* be AM4 compatible... unless you have a lot of money burning a hole in your pocket.

But in case you do, I'd like to buy 'leftover' 3900x's at 200$ each :D
 
I lost so badly on the silicon lottery that I didn't even get a silicon. Had to buy a 3700x after a month of waiting if I wanted a CPU in my computer.
 
According to Siliconlottery.com's data, the more cores Ryzen 3000 has, the lesser the overclockability.

Not really true. The 8 core 3800X clocks higher than the 6 core 3600X.

And the 12 core 3900X's lower all core OC clocks are kinda-sorta misleading. The 3900X has one golden chiplet, which typically OC's at least as well (if not better) than a 3800X, and a "shitlet" which clocks much lower. The shitlet holds back all-core OC. This binning was probably done to grant the 3900X one efficient chiplet to keep TDP limits in check AND provide a high single core boost, while keeping costs down (and volume up) with the second, shittier-binned chiplet.

The story of Zen 2 overclocking is much more complex than more cores = lower clocks.
 
Curious how many silicon lottery losers like myself are going to roll the dice on a 3950x hoping for better luck and an additional 4 cores when they are released? I was hoping the next bios update would fix some of my issues with boost and exceptionally high idle voltages but everything appears to be identical. I'm lucky to hit 4.3GHz on my 3900x and very lucky to hit 4.4GHz for about half a second. My plan is to see what the general consensus is with the 3950x and if I see a good number of users having better luck and closer advertised boost clocks with it. If that's the case, i'll likely upgrade.

Thoughts?

Play with the BIOS versions floating around out there right now. der8auer was getting 4.575 max boost out of his 3900X after a BIOS update that fixed some of the boosting behavior. Supposedly this update ALSO adjusted idle voltages.

I don't think you had a silicon lottery loss so much as you are experiencing the teething issues of the new uarch/platform.
 
Not really true. The 8 core 3800X clocks higher than the 6 core 3600X.

And the 12 core 3900X's lower all core OC clocks are kinda-sorta misleading. The 3900X has one golden chiplet, which typically OC's at least as well (if not better) than a 3800X, and a "shitlet" which clocks much lower. The shitlet holds back all-core OC. This binning was probably done to grant the 3900X one efficient chiplet to keep TDP limits in check AND provide a high single core boost, while keeping costs down (and volume up) with the second, shittier-binned chiplet.

The story of Zen 2 overclocking is much more complex than more cores = lower clocks.

And this is why I’m curious about the 3950x. With a $250 premium and a 2 month layover from the rest of the Ryzen lineup, I’m hoping it won’t have any “shitlets”
 
And this is why I’m curious about the 3950x. With a $250 premium and a 2 month layover from the rest of the Ryzen lineup, I’m hoping it won’t have any “shitlets”

My money is on the 3950X also having a shitlet. It's possible I'm wrong, of course, but I suspect you'll need to move up to Threadripper to get a pair of golden chiplets.
 
there are people who buy new cars and sell less than a year later. I'm sure there will be people who make similar poor financial decisions and buy a 3900x only to sell it and get a 3950x.

Honestly though, the customer for a 3950x (and even to an extent, the 3900x) begins to overlap with the threadripper. Other than the next gen TR's not being out yet, why would someone in the market for that many cores opt to not get the lower core count threadrippers? It offers quad channel ram, as well as many more pcie bus lanes and other benefits that makes sense when going massively parallel. Most of the numa issues are solved in zen2 it seems. Also, it seems unlikely flagship cpu buyers would be re-using old motherboards even though they're compatible, but I guess a couple will. But other than socket compatibility, there seems little reason to not be using a threadripper if you really need the cores and 12 isn't enough. Only impatience would force your hand otherwise. But then, we're talking about a thread about people who bought a 3900x when they really wanted a 3950 and couldn't wait. :-/
 
there are people who buy new cars and sell less than a year later. I'm sure there will be people who make similar poor financial decisions and buy a 3900x only to sell it and get a 3950x.

Honestly though, the customer for a 3950x (and even to an extent, the 3900x) begins to overlap with the threadripper. Other than the next gen TR's not being out yet, why would someone in the market for that many cores opt to not get the lower core count threadrippers? It offers quad channel ram, as well as many more pcie bus lanes and other benefits that makes sense when going massively parallel. Most of the numa issues are solved in zen2 it seems. Also, it seems unlikely flagship cpu buyers would be re-using old motherboards even though they're compatible, but I guess a couple will. But other than socket compatibility, there seems little reason to not be using a threadripper if you really need the cores and 12 isn't enough. Only impatience would force your hand otherwise. But then, we're talking about a thread about people who bought a 3900x when they really wanted a 3950 and couldn't wait. :-/

3900X is very interesting to me. I bought an X370 and 1700X originally. Swapped out for a 2700X (but kept everything else). If I knew Threadripper was coming out back when I first bought the 1700X, I'd have gone with that. But at the time, this was not known, and an 8 core Ryzen was about the best budget workstation deal around. Now, I can drop a 3900X in here and get Threadripper-like performance without having to pay the Threadripper price, or swap out the motherboard.

It's great!
 
Even with current bios my OC's to 4.516 routinely, 4 sticks of B-Die happily running at 3733mhz - Maybe I got a golden sample. I have not tried yet to manually OC to see my max. When all cores are active it is a little over 4.1ghz at 1.3v- 1.33v on HWinfo for Cinebench CPU. Bios for the CPU is basically default PO.

Update: Running Cinebench 20 Single Core, HWinfo has up to 3 different cores at 4516mhz at the same time, obviously HWinfo is not polling 1/1000sec but it does show I have at least 3 cores that will do 4516mhz. What is strange it is 3 cores in the same CCX, cores 0, 1, 2.
 
Last edited:
  • Like
Reactions: N4CR
like this
It's the first generation of a new architecture on a new process. Surely people know there will be yield problems and OC headroom will be minimum for the majority of products. Go buy some lootboxes if you need that gambler's high and let the technology mature a little.
 
It’s not even about over clocking it’s more about hitting anywhere near the advertised boost clocks.

noko I’d say you came up on the better side of the silicon lottery. The cores hitting those levels of boost makes perfect sense considering what DuronBurgerMan said about a Golden chiplet and a shitlet
 
  • Like
Reactions: noko
like this
Not really true. The 8 core 3800X clocks higher than the 6 core 3600X.

And the 12 core 3900X's lower all core OC clocks are kinda-sorta misleading. The 3900X has one golden chiplet, which typically OC's at least as well (if not better) than a 3800X, and a "shitlet" which clocks much lower. The shitlet holds back all-core OC. This binning was probably done to grant the 3900X one efficient chiplet to keep TDP limits in check AND provide a high single core boost, while keeping costs down (and volume up) with the second, shittier-binned chiplet.

The story of Zen 2 overclocking is much more complex than more cores = lower clocks.

Well said. This is exactly what I've seen out of my test CPU.

Play with the BIOS versions floating around out there right now. der8auer was getting 4.575 max boost out of his 3900X after a BIOS update that fixed some of the boosting behavior. Supposedly this update ALSO adjusted idle voltages.

I don't think you had a silicon lottery loss so much as you are experiencing the teething issues of the new uarch/platform.

This has been my exact experience as well. It also varies somewhat by motherboard. The MSI MEG X570 GODLIKE allows me to achieve the correct boost clocks after a UEFI BIOS update, but the ASUS Crosshair VIII Hero doesn't.
 
the frequencies are changing faster than software is sampling them. So even people who think they're not hitting the advertised frequencies may be ...assuming they're not thermal throttling already.

It may take many many tests for you to register hitting the max boost frequencies. Some entire test runs I wont see 4.5+ghz ...then I'll run them again or run different ones and see it.

Either way, the frequency you see is kind of beside the point if the measured performance is on par with or beating what's published from people who show it performing as advertised.

I'd only be concerned about what frequency you're hitting if you're seeing a problem with actual measured performance in relation to other people with similar hardware and software.
 
the frequencies are changing faster than software is sampling them. So even people who think they're not hitting the advertised frequencies may be ...assuming they're not thermal throttling already.

It may take many many tests for you to register hitting the max boost frequencies. Some entire test runs I wont see 4.5+ghz ...then I'll run them again or run different ones and see it.

Either way, the frequency you see is kind of beside the point if the measured performance is on par with or beating what's published from people who show it performing as advertised.

I'd only be concerned about what frequency you're hitting if you're seeing a problem with actual measured performance in relation to other people with similar hardware and software.

Where on Earth did you come up with that? When I don't see the boost clocks hitting their target frequencies, the benchmarks I'm running at the time show worse results than I see in cases where they do.
 
from amd. The frequency changes in microseconds, not milliseconds. If the frequency isn't being held for very long, when you poll the kernel/driver to see what the cpu frequency is, you'll get whatever it happens to be that moment or an average

a lot of what the recent bios changes made slowed down the ramp up from deep sleep states but did not alter how it behaves once it has woken up. zen2 still will adjust frequency with microsecond rates, allowing it to reach 4.5-4.6Ghz for microseconds at a time and some lower frequency the rest of the time as needed or capable. Monitoring software will only have what it was the moment it asked ...and how it deals with that number over the time it last sampled is up to it (whether it averages it or just takes it as is)

Depends on the test being done as to how long you would expect boost frequencies to be hit. Realworld benchmark tests would see much more erradic cpu use as real software tends to not hold a steady 100% cpu processing load since it's doing various tasks with the data it's computing. I'd expect back to back tests like that to sometimes show boost frequencies being hit and other times not even though the test is the same.

For the more synthetic variety, yea, you would expect to hold the boost frequencies well into the sampling rate of the monitoring software and so not seeing it usually means you're not reaching them at all.

Which I covered. If you're seeing performance lower than expected, then worry about your frequencies. But if your performance numbers are on par, then it can easily be monitoring issues making you think you're not reaching advertised frequencies
 
My money is on the 3950X also having a shitlet. It's possible I'm wrong, of course, but I suspect you'll need to move up to Threadripper to get a pair of golden chiplets.
Your title here should be changed to "Shitlet". ;)

allowing it to reach 4.5-4.6Ghz for microseconds at a time
Here's the deal: If the CPU is, in fact, clocking to 4.5 - 4.6 GHz for microseconds at a time but not long enough for any monitoring tool to actually capture it while it is happening and display it on the screen then that means the CPU isn't hitting really any sort of sustained boost clock near those values. And anyway, how much time does a CPU need to hit boost clocks to be considered "real"? 1 microsecond every second? Every minute? Overheard at AMD: "Engineer: I saw it boost to 4.6 GHz for 4 milliseconds every once in a while! Marketing: Put that on the box and ship it!"
 
Your title here should be changed to "Shitlet". ;)


Here's the deal: If the CPU is, in fact, clocking to 4.5 - 4.6 GHz for microseconds at a time but not long enough for any monitoring tool to actually capture it while it is happening and display it on the screen then that means the CPU isn't hitting really any sort of sustained boost clock near those values. And anyway, how much time does a CPU need to hit boost clocks to be considered "real"? 1 microsecond every second? Every minute? Overheard at AMD: "Engineer: I saw it boost to 4.6 GHz for 4 milliseconds every once in a while! Marketing: Put that on the box and ship it!"

Marketing is marketing. How many phone's are sold talking about how fast they are when we know for a fact that they all thermal throttle if you try to make use of that performance for more than a couple seconds. Or laptops that advertise they use xyz cpu which you'll find benchmarks of in various places all over but the particular laptop you're actually buying has crap cooling so they underclock the the crap out of it and it never has a chance to perform at anything like the levels that part could.

There's no established standard rules for saying how long something has to run at a given speed before you can advertise that it reaches that speed. So companies use whatever they want. Just like there's no established rule on advertising how thick your circuits are in your cpu when your cpu is actually a heterogeneous mix of circuits fabricated at various widths/sizes since cpu's have consumed more and more traditionally external systems. Do manufacturers give you an average or the smallest one that they feel is equivalent to historic measurements or maybe the smallest one they can get away with reporting? Do they advertise the max heat output of their cpu at the highest speeds they're advertising? or some average based on a workload only they care about defining (cough)intel(cough).

We all hate marketing.

it may hit 4.6Ghz for microseconds at a time. It might hit it only under very specific workloads. It's specifically marketed as an overclock speed that the cpu decides to boost to when certain criteria allow it to. In terms of marketese, that's pretty up front about things. There's tons of other things in the cpu hardware space that is far more of a stretch of the truth to be upset about.
 
  • Like
Reactions: N4CR
like this
How long does something need to happen before it's considered happening? Having it happen in an observable time frame at least once would be nice, to some people.

This sounds remarkably similar to when the Devil's Canyon chips came out and Intel claimed 5 GHz was possible, but only if you knew how to overclock it correctly. That shit didn't fly well then. Apologists don't need to explain away a deficiency on the AMD side either.
 
there's a far difference between the two.

I think the description is pretty clear. Some people are more upset than warranted.

Precision boost overdrive is an extremely aggressive opportunistic overclock that most people probably wouldn't be able to tune on the cpu and be stable at all. let alone utilize for a short burst of time. That's why it's marketed as an overclock that voids all warranties when used. https://community.amd.com/community...precision-boost-overdrive-in-three-easy-steps It's basically a free auto overclock that you would otherwise not have anyway.

Granted that article is in reference to threadripper's implementation, the same gist would hold true for zen2

You should expect that if your temps are good and your motherboard can feed it the voltage that you'll hit the speeds for short periods of time, but it's not guaranteed. As far as guarantee goes, you should really just expect regular precision boost speeds if your temps are good. But even that is an "up to" type of number, and the only thing really absolutely guaranteed is hitting the base clocks in a worst case scenario before thermal throttling kicks it down even lower than that.


It's pretty apparent that the cpu's running the field in benchmarks aren't golden samples running or running on ln2 or other exotic coolers. They'll be running the gamut in terms of silicone lottery just like ours are. The performance numbers are solid. If yours aren't on par, it's not because everyone in the media is working with amd in a conspiracy to make their product look better than it is. It's probably because your particular setup is not up to the task or your environment isn't comparable or maybe you just got that unlucky. But you can be pretty confident that you're in the minority or _everyone_ would be complaining ...3900x's and their friends have been selling faster than they can be stocked pretty much everywhere but apparently canada.
 
What's considered a decent 3900x overclock? Just got one a few days ago but i've not had time to really mess with it. I replaced a 2700x that sat at 4.3ghz without issue.
I do want a 3950x and figured I'd sell the 3900x and 2700x when they launch to get one without a lot of financial pain over what i've just dumped on the 3900x.
 
4.5ghz on 2 CCX seems to be a pretty "good" OC to me. Really though it's still quite obvious we're fighting for a tiny gain on maxed out chips in terms of real world performance (ignoring actual benefit from RAM OCing)

Honestly, I grabbed a 3700x to save a few bucks and kind of think a 3600/x would have been fine for now since this is just my gaming rig with the occasional VM running to do some work. 3700x brick walls at 4.3ghz but the performance is ample. I have no manual OC, just +200 PBO.. all cores stay at 4275 while gaming, no boosts to 4.4ghz happen ever. I want the 16 core model but I'm going to wait for the 4950x (if that happens - I hope, or whatever ends up being the very best thing for x570/am4). This platform needs a bit of time to mature IMO.
 
the cpu and motherboard can overclock it by itself to 4.2 all cores ...then there's the thread's subject of contention, that it can send a couple cores to 4.5-4.6Ghz


You can try to do better than the algorithm used by amd ... but those are the overclocks. Your goal is to do those speeds with less voltage to hold them for longer periods of time ....or maybe you want to force the cpu to hold those frequencies and ignore the warnings of operating at certain voltages/temps for too long.
 
from amd. The frequency changes in microseconds, not milliseconds. If the frequency isn't being held for very long, when you poll the kernel/driver to see what the cpu frequency is, you'll get whatever it happens to be that moment or an average

a lot of what the recent bios changes made slowed down the ramp up from deep sleep states but did not alter how it behaves once it has woken up. zen2 still will adjust frequency with microsecond rates, allowing it to reach 4.5-4.6Ghz for microseconds at a time and some lower frequency the rest of the time as needed or capable. Monitoring software will only have what it was the moment it asked ...and how it deals with that number over the time it last sampled is up to it (whether it averages it or just takes it as is)

Depends on the test being done as to how long you would expect boost frequencies to be hit. Realworld benchmark tests would see much more erradic cpu use as real software tends to not hold a steady 100% cpu processing load since it's doing various tasks with the data it's computing. I'd expect back to back tests like that to sometimes show boost frequencies being hit and other times not even though the test is the same.

For the more synthetic variety, yea, you would expect to hold the boost frequencies well into the sampling rate of the monitoring software and so not seeing it usually means you're not reaching them at all.

Which I covered. If you're seeing performance lower than expected, then worry about your frequencies. But if your performance numbers are on par, then it can easily be monitoring issues making you think you're not reaching advertised frequencies

Your explanation sounds like its from AMD's PR machine. While your explanation is technically true to a point, its actually not that hard to test your system to see if you can achieve the desired boost clocks. This isn't rocket surgery, all you really need to do is fire up the single CPU benchmark in Cinebench, POV-Ray or any number of tests to get the system to boost its clocks to the advertised frequency, or near it anyway. Unfortunately, this doesn't work in all cases or in all configurations. Many people are having issues with this and its something that may vary from processor to processor and even motherboard to motherboard. In most cases, a UEFI BIOS update will resolve the issue. However, I've seen situations where it doesn't work. And like I said, in cases where it doesn't work, the benchmarks back up what I'm seeing in Ryzen Master, which is as close to real time monitoring as you will get for these CPU's.

You bring up Precision Boost Overdrive, so lets talk about that. It doesn't really do anything with the Ryzen 3000 series. All it does is adjust PPT, EDC and TDC values. You can even input your values for it as those values override the CPU's presets and use the motherboard values instead. Even with PBO+AutoOC, boost clock behavior on the 3900X at least doesn't really change. Gamer's Nexus found that it essentially didn't work at all. I only tested it on the 3900X and so far, PBO+AutoOC doesn't do anything. In fact, it often hurts performance. I'm not the only one who experienced this either. It doesn't matter how aggressive the algorithm is, PPT, EDC and TDC values aren't what's holding these back from overclocking. The algorithm worked the same on 2nd generation Threadripper CPU's, but actually worked well. On Ryzen 3000, not so much.

I've literally been reviewing hardware and testing this stuff for well over a decade. I know all about what you said above. I have AMD's review guide, I participated in launch coverage of these CPU's and even did an entire article which covered the BIOS updates and boost clock behavior. I've seen boost behavior under controlled conditions with different motherboards and different Ryzen CPU's. Take the ASUS Crosshair VIII Hero I have here. Using my 3900X test CPU, it won't achieve the proper boost clocks under a single threaded work load. This is on a clean system with very little on it. I don't have applications which could be generating enough activity on the CPU to keep it under a multi-threaded workload. That same CPU will boost correctly on the MSI MEG X570 GODLIKE without a problem. It didn't during my initial review, but did after a BIOS update. Back on the Hero, the Ryzen 3700X works fine and boosts correctly with no software changes to that system using the same tests. Again, the 3900X doesn't.

Boost clock issues with these CPU's are a real thing. That said, DuronBurgerMan is right on the money. On a 3900X, you typically end up with one good chiplet and a shitlet. The shitlet holds back your all core overclock and most reviewers got about the same result. 4.3GHz for all cores. You can bring specific cores to 4.5GHz or 4.6GHz, but you can really only expect about 4.4GHz out of one CCX. The other chipset likely won't go that high. AMD even states in its reviewers guide that you'll only see boost clocks at 4.6GHz on a single core and that heavily multi-threaded tests will result in something around 4.1GHz or so on all cores.

What's considered a decent 3900x overclock? Just got one a few days ago but i've not had time to really mess with it. I replaced a 2700x that sat at 4.3ghz without issue.
I do want a 3950x and figured I'd sell the 3900x and 2700x when they launch to get one without a lot of financial pain over what i've just dumped on the 3900x.

4.3GHz is likely all you will see on a 3900X using all cores. However, you can clock a specific CCX or CCX's to 4.4GHz or so if your lucky but your mileage may vary considerably there.
 
the cpu and motherboard can overclock it by itself to 4.2 all cores ...then there's the thread's subject of contention, that it can send a couple cores to 4.5-4.6Ghz


You can try to do better than the algorithm used by amd ... but those are the overclocks. Your goal is to do those speeds with less voltage to hold them for longer periods of time ....or maybe you want to force the cpu to hold those frequencies and ignore the warnings of operating at certain voltages/temps for too long.

It can, and you can try, but PBO is virtually useless on the 3000 series in my experience so far. At least where the 3900X is concerned. I think some people have reported better results on other Ryzen 3000 series CPU's.
 
How long does something need to happen before it's considered happening? Having it happen in an observable time frame at least once would be nice, to some people.

This sounds remarkably similar to when the Devil's Canyon chips came out and Intel claimed 5 GHz was possible, but only if you knew how to overclock it correctly. That shit didn't fly well then. Apologists don't need to explain away a deficiency on the AMD side either.


There is quite a distinct difference from intel claiming you could OC to 5Ghz "on AIR!" when in reality the test they used for that claim involved the testers using an AIO and chilled intake air from LN2. That is about as far from "on AIR! cooling" as an apple is from an orange.

AMD claims the CPU will boost up to x.xx GHz if the thermal and electrical conditions allow it. They did not claim the cpu will do an Allcore OC to the boost speed, or that if you are running a single threaded application that it will sustain x.xx Ghz indefinitely (or for any certain amount of time)...


A lot of people are upset because they think they are being robbed of performance when in fact the CPU's in this series deliver an insane amount of performance (while sipping power at levels that Intel cannot begin to reach) at unheard levels of pricing prior to the launch of the Ryzen series. You do not have to be an AMD "fan" to know this is true. Even the diehard Intel Defense Force members should be the first to acknowledge this since it has allowed them to buy their vaunted Intel SKUs at cheaper prices then they would have been able to had Ryzen/TR not existed in the market.


I think that the 16 core SKUs will have 1 of the tightest binned 8 core chiplets and 1 average one unless AMD has reached new yield levels such that even the "average" chiplets are now on a par with the launch window's "golden" chiplets. If this happens great, but I think that AMD will launch with 1 great/1 average to above average and maybe refresh with a sku for $899 that has dual Golden chiplets allowing the CPU to boost to 4.5Ghz+ AC...That would be the fastest single socket CPU on the market in anything other then latency bound sub 1080P gaming. There is a fairly strong desire level for such a SKU should AMD be able to bin tight enough to release it. They could call that one the 50th AV performance edition which would line up with the other 50th AV SKUs.
 
  • Like
Reactions: N4CR
like this
I've literally been reviewing hardware and testing this stuff for well over a decade. I know all about what you said above. I have AMD's review guide, I participated in launch coverage of these CPU's and even did an entire article which covered the BIOS updates and boost clock behavior. I've seen boost behavior under controlled conditions with different motherboards and different Ryzen CPU's. Take the ASUS Crosshair VIII Hero I have here. Using my 3900X test CPU, it won't achieve the proper boost clocks under a single threaded work load. This is on a clean system with very little on it. I don't have applications which could be generating enough activity on the CPU to keep it under a multi-threaded workload. That same CPU will boost correctly on the MSI MEG X570 GODLIKE without a problem. It didn't during my initial review, but did after a BIOS update. Back on the Hero, the Ryzen 3700X works fine and boosts correctly with no software changes to that system using the same tests. Again, the 3900X doesn't.

Boost clock issues with these CPU's are a real thing. That said, DuronBurgerMan is right on the money. On a 3900X, you typically end up with one good chiplet and a shitlet. The shitlet holds back your all core overclock and most reviewers got about the same result. 4.3GHz for all cores. You can bring specific cores to 4.5GHz or 4.6GHz, but you can really only expect about 4.4GHz out of one CCX. The other chipset likely won't go that high. AMD even states in its reviewers guide that you'll only see boost clocks at 4.6GHz on a single core and that heavily multi-threaded tests will result in something around 4.1GHz or so on all cores.


You're contradicting yourself. Is it the motherboard holding back the cpu, or the cpu? You go into length talking about how the cpu couldn't hit the frequencies it theoretically could under one motherboard but switch it to a different one and it did. The cpu chiplets didn't magically get better. That implies that the VRM and setting related to it play a significant part in how the cpu can overclock itself.


The rest of what you stated just backs up the fact that amd isn't misleading people about what the cpu's can do. People are trying to do more than what's advertised or totally skipping the reading part and assuming things that aren't being stated. The idea that peak frequencies above the 4.2-4.3Ghz range were never in question about being limited to a single core since before they went to market even. Feeling gipped by that fact is kind of stupid.

Silicone lottery is real, dont get me wrong. I think the motherboard variables though play a bigger role than that. Even on high end boards, vrm's are tuned pretty conservatively. Upping the frequency of the vrm and giving it higher current limits may let your cpu spread it's wings better on the boards you're not seeing it hit the mark on.
 
There is quite a distinct difference from intel claiming you could OC to 5Ghz "on AIR!" when in reality the test they used for that claim involved the testers using an AIO and chilled intake air from LN2. That is about as far from "on AIR! cooling" as an apple is from an orange.

You're correct in that Intel said it could hit 5 GHz on air. No where on the box does it claim that it does—and I should know as my CPU in sig came in a box that never claimed it.
 
Back
Top