RyZen 3000 Boost Issue: What's your take?

[silkstone]

If I touch my CPU Voltages at all, my system won't boot, so there's that...

Mine neither. I set an offset voltage slightly lower, and I only notice boosts being reduced and instability.
When I dial the voltage down further. It refuses to boot and comes up with a continuous load-defaults bios pop-up.

 

[Nordic]

I will tell once I get mine. 480 mm rad placed outside the case should give decent cooling. The rig have standing ready for at least five weeks now, well except the darn CPU. Ordered on the 7th of July.

I am eagerly waiting to see your results.

Then try the advice I offered earlier, lower your voltage to no higher than 1.3v and your boost clocks will be fine.

Alright. I finally tried this. I set voltage to a fixed 1.3v. It booted but CPUz reports that I am still getting voltage up to 1.45v. I will need to investigate further.

 

[xtreemchaos]

ive just watch this I think this guys take is interesting.

 

[tabascosauz]

I am eagerly waiting to see your results.


Alright. I finally tried this. I set voltage to a fixed 1.3v. It booted but CPUz reports that I am still getting voltage up to 1.45v. I will need to investigate further.

It's been a couple of years since I used an ASRock BIOS (Z77 Extreme3, babyyy) but on my B450I Aorus it won't care what I put in the voltage boxes, with the lone exception of Vcore. And manual Vcore still only works if CPB is disabled (and therefore boost). Hence my comment about Ryzen taking matters way too far into its own hands and not giving a singular shit about what anything or anyone else thinks it should be doing.

I know that in theory, once you lower Vcore, Ryzen will adjust to the max frequency that it can stably sustain at those volts. In reality, it either doesn't care or doesn't boot.

 

[Nordic]

Did you not see that posted 2 posts sooner than yours?

 

[RealNeil]

It’s worse on the Intel side, unless you have really good cooling forget about consistent boost clocks unless of course you enable All-Core Boost.

That really good cooling is key with my Intel chips. It doesn't work right without good cooling combined with good case airflow.

 

[TheoneandonlyMrK]

So im sitting with my 2600X staring at a 3800X to buy, is now the time , or is the shit show to great on an x470(crosshair7) board to bother.

 

[Sybaris_Caesar]

So im sitting with my 2600X staring at a 3800X to buy, is now the time , or is the shit show to great on an x470(crosshair7) board to bother.

Had no problem so far but if I hadn't sold my PC I'd have waited until B550 boards to show up or buy during holiday reason for discount.

 

[silkstone]

So im sitting with my 2600X staring at a 3800X to buy, is now the time , or is the shit show to great on an x470(crosshair7) board to bother.

No reason not to go for it, just so long as you know that you might not get the boost speeds advertised. It's still a great CPU for the price, especially if you compare it to the cost of switching platform.

 

[xtreemchaos]

to tell the truth Guys im finding it hard not to buy the 3900x but im trying to wait for the 16core the moneys burning a hole in my pocket and im scared the misses might get wind of it pray for me.

 

[TheLostSwede]

ive just watch this I think this guys take is interesting.

It's pretty much my experience as well, early UEFI's were not great, but by now, it seems everything is "working".
As I've mentioned elsewhere, Gigabyte delivered the first "final" F5 UEFI for my board and I now have working boost, my RAM runs at 3800MHz 1:1 at good timings for the RAM (better than spec) and I'm not seeing any issues. Some of the previous beta UEFI versions have had various issues that have prevent everything from working at once, except maybe F5l, but it was a bit buggy still for me.
I'm sure someone will say my boost Voltages are too high, but as AMD has said these are ok Voltages for boost, it's covered by their warranty, so I'm not worried about it.
I also think I've been, at least so far, one of the most frustrated and opinionated people here about the fact that my 3800X didn't behave as expected. So in all fairness, it seems like the board makers, or at least Gigabyte in this case, has fixed AMD's problems.

Not sure how many nano seconds my CPU has been over 4,500MHz though...

 

[xtreemchaos]

to be fair to the mobo makers thay have been playing catch up, the last round 2000s my 2700x was under performing until the f40 bios update but now it gets a little over the 4.3 boost when the weather on the cooler side. its still early days.

 

[TheLostSwede]

to be fair to the mobo makers thay have been playing catch up, the last round 2000s my 2700x was under performing until the f40 bios update but now it gets a little over the 4.3 boost when the weather on the cooler side. its still early days.

Well, AMD works VERY differently to Intel when it comes to building the motherboard ecosystem.

Intel: Hands out reference boards with multiple ES CPUs and with full documentation, has local FAE people, works closely with the board makers months ahead of launches to try to iron out most bugs that haven't already been found by Intel (not talking CPU bugs here...) and they also work closely with the BIOS/UEFI vendors and even provides full UEFI reference designs for its platforms.

AMD: Here, have some chipsets and some documentation, please make us some boards. Oh, right, here's a very early alpha UEFI/AGESA that will get you going, along with some very slow, alpha silicon CPUs. Let us know if you have any problems and btw, we're launching in two months.

At least this is the impression I've gotten from talking to the board makers. In all fairness, AMD has a lot less budget than Intel, but this is now how you go about making a great product launch experience for end users. So there's no wonder it has taken some time to fix all the issues and AMD can only do so much to fix them, as they need help from the board makers.
That said, due to how the AGESA works, AMD apparently dictates a lot of the UEFI settings and even naming. I guess to some degree, this has something to do with making Ryzen Master work across all boards as well.

 

[TheoneandonlyMrK]

No reason not to go for it, just so long as you know that you might not get the boost speeds advertised. It's still a great CPU for the price, especially if you compare it to the cost of switching platform.

Investigation of this imminent, r7 3700X on the way, i can't justify £50 for the 3800x and 100 mhz personally.
Can't really justify any purchase im homeless but their are more important things then homes lol.

I've taken all the likes the wrong way and upgraded to the 3800X :/

 

[Darmok N Jalad]

To me, AMD is not misrepresenting if board makers can work through the issues and get things to as-advertised. It’s the downside to using the same socket for 3 generations, especially considering how much AMD has iterated over those three product lines. They are all on different manufacturing nodes, there have been architectural changes, support for new technologies, and even changes to how the chip is physically laid out on the package. You don’t see that with Intel, as they won’t push one socket across 3 generations. It sounds like board makers need time to get it refined, and in 6 months, this will all be behind us.

 

[Xzibit]

AMD: Here, have some chipsets and some documentation, please make us some boards. Oh, right, here's a very early alpha UEFI/AGESA that will get you going, along with some very slow, alpha silicon CPUs. Let us know if you have any problems and btw, we're launching in two months.

Wasn't Gigabyte the one who leaked the Computex X570 launch last year (7 months early) even the naming

 

[Nordic]

95% of 3900x owners in his survey were not reaching boost clocks.

 

[xkm1948]

95% of 3900x owners in his survey were not reaching boost clocks.

 

[phanbuey]

i mean... yes, amd marketing fail.... BUT we're talking about maybe 2% situational performance loss on a $500 12-core CPU that trades blows with Intel's $1070 9920x, and comes out on top most of the time.

It's just so hard to get upset about that.

 

[xkm1948]

i mean... yes, amd marketing fail.... BUT we're talking about maybe 2% situational performance loss on a $500 12-core CPU that trades blows with Intel's $1070 9920x, and comes out on top most of the time.

It's just so hard to get upset about that.

False advertising is false advertising. Do not sugar coat wrong doings simply because "meh underdog AMD" As a multi-million for profit company AMD had every chance to decide what to put on their spec sheet regarding final clock speed. They choose to lie, and that is unacceptable.

This is why we have laws. As much as I hate to say, lawyers need to get on this for another class action law suit. False advertising should never be tolerated.

 

[phanbuey]

False advertising is false advertising. Do not sugar coat wrong doings simply because "meh underdog AMD" As a multi-million for profit company AMD had every chance to decide what to put on their spec sheet regarding final clock speed. They choose to lie, and that is unacceptable.

This is why we have laws. As much as I hate to say, lawyers need to get on this for another class action law suit. False advertising should never be tolerated.

In principle I absolutely agree with you. We should have manufacturers advertise boost clocks that the CPU hits.

But the two things I think the lawyers would have an issue with is 1) intent, and 2) actual damages to consumer.

1) Intent because of the way the AGESA patches rolled through, and because zen 2 is a brand new designed and AMD is known for getting caught with it's pants down on multitudes of issues in past releases (like not being able to use certain ram, not being able to run certain programs, instruction bugs etc.) So it would be hard to say "AMD DELIBERATELY ADVERTISED FALSE CLOCKS" because they're a bunch of keystone cops over there -- Did they intend to make chips that could turbo at 4.6Ghz? Probably. Did they deliberately manufacture a bunch of chips that can't hit turbo? most likely not.

2) Actual damages when you take into account the Boost clocks and the cost of the chip, how the actually behave, what is the % variance between chips? It's a $499 chip, that performs 98% to spec when turbo'd so the WORST CASE scenario for damages is that they ripped you off by what? $10?

So on top of the fact that intent is unclear (which is a afaik is lesser degree of false advertising) and that the actual damages can be considered within margin of error (1-2% is typically margin of error in tech testing). I'm really not sure this is the biggest deal.

I would be much more pissed about my RAM not posting or Destiny 2 not running, or the whole host of WHEA errors that are happening with NVIDIA cards than my 3600x hitting 4350 instead of 4400.

 

[TheoneandonlyMrK]

False advertising is false advertising. Do not sugar coat wrong doings simply because "meh underdog AMD" As a multi-million for profit company AMD had every chance to decide what to put on their spec sheet regarding final clock speed. They choose to lie, and that is unacceptable.

This is why we have laws. As much as I hate to say, lawyers need to get on this for another class action law suit. False advertising should never be tolerated.

you do realise GPu's have boost values rated in this way and Nvidia started it, I am not keen myself but hang them all or none for it, many describe that max Turbo/boost game clock whatever exactly the same but at the end of the day what do they all do, boost within your systems power and cooling envelope to the max If it can and drop when the going gets tough, who else ISNT making chips this way.
The Marketing sucks again not the only company passing off nonesense. IMHO.

 

[phanbuey]

you do realise GPu's have boost values rated in this way and Nvidia started it, I am not keen myself but hang them all or none for it, many describe that max Turbo/boost game clock whatever exactly the same but at the end of the day what do they all do, boost within your systems power and cooling envelope to the max If it can and drop when the going gets tough, who else ISNT making chips this way.
The Marketing sucks again not the only company passing off nonesense. IMHO.

Agreed - just talk to ANYONE who bought a macbook pro in the last 5 years lol. Those don't even run base clocks (on cpu OR gpu) if you push them.

Don't even get me started about phone battery/laptop battery advertisements.

 

[Darmok N Jalad]

False advertising is false advertising. Do not sugar coat wrong doings simply because "meh underdog AMD" As a multi-million for profit company AMD had every chance to decide what to put on their spec sheet regarding final clock speed. They choose to lie, and that is unacceptable.

This is why we have laws. As much as I hate to say, lawyers need to get on this for another class action law suit. False advertising should never be tolerated.

I doubt a class action lawsuit would hold water. Specs list the top frequency as a “max boost clock,” or the most the chip is capable of reaching. There is a base clock that you will get guaranteed, and I can see that argument holding. If it doesn’t, the Intel should also get sued for TDP ratings. AMD and Intel keep going back to clockspeeds as a major spec, but the reality is, workloads on multi core CPUs are so variable that a guaranteed max clock is hard to obtain or probably even measure.

It’s not new either. My DD has a W3690 Xeon that is rated at 3.73GHz top clock. I have never seen it hit that—ever. One time it came close, hitting 3.68GHz on one core. The rest of the time, it reads at 3.6GHz, which it will happily do across all 6 cores. The W3690 is a 10 year old CPU. I think the reality is, these CPUs hit these max clocks at very brief intervals, maybe beyond the point of being accurately measured. What would matter to me is if I’m getting the performance results in real life, and/or if that matches my expectations for the money. Right now, I wouldn’t buy a 3000 Zen, but that is because I think the platform is immature, and I wouldn’t want to deal with the bugs as board vendors and AMD get their stuff together.

 

[Vayra86]

I hate all forms of "boost" with a vengance. So, let me just say that I won't be saying this is an AMD exclusive issue. It's even present in the GPU scene. Heck, it practically started there.

I must say that I do like the way Nvidia approaches boost; the way they market it, its like it just keeps on giving, royally boosting above spec.

With most others, boost feels like a weird trick of the mind indeed.