Is my 5950x a dud?

[Heofz]

Hey guys,

So I took delivery of my 5950X yesterday. Out of the box I'm led to believe that the individual cores in this CPU are each (i.e. not all simultaneously) meant to be able to boost to 4.9GHz minimum if I'm not mistaken? If so, I think mine is pretty sucky in this department (see the "Maximum" recorded values):



As you can see, only 5 out of my 16 cores seem able to break past 4.9GHz. This is all after an hour of observation and running various combinations of Cinebench (single core and multi core tests) and doing various tasks, which surely will have triggered a max boost across each core by this time.

I've spent several hours fiddling with various PBO settings, Curve Optimization, upping/downing voltage offsets, LLC, even manual overclocking in Ryzen Master, and cannot for the life of me get those other 11 cores to take a load @ 4.9GHz without a black screen crash. I saw better boost speeds from my 5800X under the same conditions.

Overall I'm still super impressed by the CPU of course but I'm seeing other peeps here with all of their cores able to boost past 4.9-5ghz at the same voltages seemingly without breaking a sweat.

Performance wise, the CPU isn't terrible I don't think, managed a nice score on R20 with a manual OC @ 4.675GHz with 1.43v and tickling 88-89°C with my Lian Li Galahad AIO 360 running at full pelt...



See signature for specs, running the MSI 7C84v151(Beta) BIOS which I believe includes the AGESA 1.1.0.0 Patch C from the previous version.

So yeah, just a bit disappointed with those individual core speeds. Tempted to return the CPU and roll the dice on a new one, but was wondering what you guys thought?



Wondering if this batch just a bad pick of binned cores... or others have just been super lucky.

 

[Heofz]

4.9ghz boost is just for single core (which will be more like 5ghz from what I've seen). You should be able to hold between 4.3-4.5ghz multi-core depending on the program.

Yes, but each of my cores (I think?) should all individually be capable of hitting 4.9, but as you can see in my first pic of HwInfo, only 5 of them are able to hit that speed (if you look at the "Maximum" column).

 

[Heofz]

I don’t think there’s anything wrong there. It’s inevitable that some cores may not be as strong as others, and in most workloads it will obviously share around the burden. Of a couple of cores are out by a 100MHz or so, I really would not be worrying in the slightest.

hmm a couple cores sure, but 11 cores?

 

[Heofz]

Ahhh apologies! Gotcha. Well, the chips have the concept of 'preferred' cores, and from all the playing around I've seen regarding Curve Optimisation, everyone has at least a couple of cores that are stronger than all the others, which are used for the single threaded burst loads. I would be very surprised if any chips can hit max ST boost on every core. Happy to be corrected on that from someone more knowledgeable though.

hmm yeah see, here's a screenshot from @bongo78 on his thread:

His core #5 was over 5ghz on another screenshot of his too, so literally all of his cores are able to boost past 4.9 Ghz no problem it seems :/

 

[Heofz]

Everything OP posted reads like a humble brag flex to me. Your CPU is awesome, you're lucky - mine won't do above 4.3ghz all core using fixed voltage without hitting 100c on 360m AIO and yours does nearly 4.7ghz. 4.9ghz is the advertised clock for a SINGLE CORE, getting that speed on multiple is great and above the advertised speed

Definitely not a humble brag Just wondering what AMD’s policy is on core binning for these chips. Also your all core temp doesn’t sound right to me but depends on what that fixed voltage is that you’re hitting.

The key question is: are all the individual cores supposed to be able to boost to 4.9 (not at the same time obviously) or does only one or two of them need to be able to hit it before the chip is deemed acceptable for sale.

 

[Heofz]

This is maybe a bit of humble bragging! I'm not usually lucky when it comes to silicon lottery but (touch wood) I seem to have done well this time.

To OP, I would give it a bit more time and observation. Maybe check cooling - I originally had higher temps like you towards high 80's / low 90's. Gave my PC a good clean out and temps dropped significantly and then saw much better boosting.

Yep see, this is absolutely night & day compared to mine. For the price I was expecting all 5950X’s to be binned like this... definitely think mine is at the bottom end of the scale whereas yours is near the top.

Will monitor the max hit core speeds for a bit longer but there’s just no way my silicon is allowing individual core speeds at that level. The whole pc build is only a few weeks old too and everything is clean

 

[Heofz]

The 5950x is advertised as single core 4.9ghz and you want all 5.1? If you want a heavy binned chip you should be buying your CPUs from siliconlottery.com not OC UK

Not necessarily, i just want all 16 of my cores to be individually able to hit 4.9 at least

 

[Heofz]

To get the same picture* you need to run a single threaded benchmark, preferably not too cpu heavy and manually pin its affinity to each core one at a time.
Else windows scheduler will always make single threaded task run on preferred cores. Other cores look like they never boost high because windows never gives them a single threaded job light enough.

* minus AutoOC, so boosting up to 5050
your cpu is fine

Yep, tried this with Cinebench R20, set affinity to various different cores while the single core benchmark was running. Like I said, most of them simply will not boost to 4.9Ghz no matter what I do*. The picture in my first post is the result of that, hence my concern.

* Unless I manually set them to 4.9Ghz in Ryzen Master @ 1.5v then try to run a single core benchmark, at which point it black screen crashes immediately.

 

[Heofz]

OK, so you're chip isn't the greatest for overclocking. Realistically outside of benchmarking you'll never notice a difference. Just like any other chip whether that be amd or Intel, if you want a good clocker, you've got to buy in bulk, bin them all and most likely take a hit when selling the rejects. I used to go through 20+ chips on average easily for Intel back when I was benching.

If you ignore manual oc and starting playing with pbo / curve, you'll get that single core boost raised no problem and end up with a more efficient multicore.

yeah i guess my main concern is more with the single core boost rather than a multicore push as that will affect performance the most on a day to day basis. Not entirely sure how the boost algorithm works during a single core or single thread load (e.g. gaming), but I assume the workload gets passed from one core to the next “hot potato style” in order to keep the temps from going too high. If this is the case, my chip performance would suffer when the algo runs out of higher boosting cores to juggle with.

Or am I talking complete **** ?

 

[Heofz]

5950x straight in, 100% stock bios, 1 run of r20 multi and single. Just the same, not all core hit 4.9. Must be another dud...I'll show you clocks after I've had a play about with it tomorrow.

that's only after 12 minutes and you've already wiped the floor with mine Mine's been going for 14 hours and I still haven't hit even 4.8Ghz on any of the cores on my CCD2. Something is definitely not right:

Unusual.
Seems like CCD2 will need heroic negative amounts of Curve Optimizer to boost higher.

I suspect you are right I've done lots of fiddling in curve optimizer but no luck whatsoever.

 

[Heofz]

I had the same issue as you earlier today

A reinstall of windows helped, but also notice HWInfo is not reporting the correct boost clocks. After the install PBO CO worked again and all cores are going between 5050 - 5175 now although HWInfo is not reporting it properly

Very interesting.. what prompted you to reinstall windows?

 

[Heofz]

I was running a temporary install as I knew I’d be crashing a lot testing the breaking point of all my components. I had originally seen PBO with CO working fine and then it suddenly not. Others have reported similar cases of PBO with CO doing nothing for them too. After trying a Bios reset with no change, the next logical step was an OS reinstall which seemed to do the trick.

It all went downhill though after installing ICue, which I haven’t reinstalled yet. Will be interesting to see what happens when that’s installed.

I have iCue installed although I close it when running benchmarks. Seems very odd how applications can affect core speed measurement in this way.

 

[Heofz]

I haven’t installed it again yet, but I’ll let you know if it changes anything. Still testing stability of my PBO CO curve which is causes random reboots at idle. This has to be a bios thing as I’ve seen a lot about this.

Hey, any developments on this? Super curious to see if there's a software issue that messes with CCD2 core speed readings.

 

[Heofz]

Just tested now and after installing iCUE PBO with CO makes no difference again. I cant get past 5 ghz on any core no matter the CO settings ICUE definitely screws with it. Just going to try one more thing

Can't believe my eyes killed iCue and stopped all the Corsair services. Within 15 mins I'm seeing 5/8 of my cores on CCD2 over 4.9GHz... wtf


I did also update AMD chipset drivers to latest but I'm 99% sure that didn't make any difference here as the hwinfo core speeds didn't change at all. Deffo onto something I think, perhaps the iCue monitoring software has been messing with the HWiNFO64 monitors...

Managed to score 632 on Cinebench R20 single core as well which I think is the best I've seen yet. Very promising

Gonna play with my PBO and CO settings a bit more to see what I get now...

 

[Heofz]

So.... here I am now, wishing I never RMA'd my original 5950x, ended up waiting 3 months for a replacement (after waiting I somehow managed to get 3 at the same time), and after testing all three, I can tell that the newer 5950X's seem to be binned like absolute crap (at least in comparison to the first batches). Pretty sure AMD lowered their standards at the factory to increase their yield, which wouldn't be surprising given the recent shortage situation. Still seems a bit cheeky considering those "golden samples" are the ones the reviewers etc got their hands on and gave everyone a false sense of consistent chip quality.

The one I'm running now (best of the three) is still significantly worse than my first one. Damn shame tbh.

Tempted to get rid of the 5950X and switch to the 5900X. Do we know if the binning criteria for those are higher? I.e. is it still pretty common to see all the cores individually capable of PBO boosting to >5GHz (let alone the advertised 4.9Ghz)? Because these new 5950X's sure as hell can't, apart from like maybe 1 or 2 cores out of 16, which absolutely sucks.