POLL: CPUZ Feelz, let your opinion be known.

[Matsglobetrotter]

15 minutes ago, Leeghoofd said:

Proposal 2 has CPUZ limitations as mentioned in other posts, so that might be something for the future

Regarding the memory proposal we have already tried it out in Country and Team Cup and indeed it is more challenging, but again CPUZ is not 100% bulletproof.

bullet proof hehe. we are working on the edge of all hardware so no wonder the software gets.. flaky hehe

[Leeghoofd]

Try e.g. hotplugging dimms, you might be surprised what readouts one gets

[Midnight703]

I am another ambient bencher who this likely wont make a huge difference for either way. That being said, having 2 categories one for all core and one for single core seems to be the best answer to both sides. Just seems to make the most sense. 

[zeropluszero]

18 hours ago, Leeghoofd said:

Ah so you are giving up so easily? ever thought about providing good arguments or even better maybe alternatives instead of feeding the troll machine or spicing up urban myths about HWBOT.

Nah, not really.

[ground]

23 hours ago, Leeghoofd said:

Try e.g. hotplugging dimms, you might be surprised what readouts one gets

oh there are also platforms where you can drop channels in such a way all tools but taskmanager fail to detect how many channels are currently populated (making a certain tc stage a couple years ago quite pointless...). Can see in Taskmanager(due to capacity) and bios only one out of 3 sticks is detected, yet cpu-tweaker and cpuz detect all 3 channels as populated and running at high frequency...

[Leeghoofd]

Yeah one of the reasons we never can apply this for eg for online qualifiers of G.Skill or such... some people are too creative ?

[Strunkenbold]

22 hours ago, Leeghoofd said:

How about an added column in the CPUZ ranking? Is that a feasable solution for all parties ?

 

I think this is going in the right direction. But that still would have the problem of not knowing if someone downclocked his cores. If I got it right, this discussion is about showing if someone overclocked all of his available cores to a specific maximum frequency, right?  So what if Tim codes a script, which reads out the page from the provided CPU-Z verification link and IF 100 % of x cores are active AND frequency between them is within a reasonable deviation mark them as all core max frequency and call it a day. This way you could filter them too like in your above example.

Im twisted about the possibility of separated ranking. I mean CPU max frequency is not really a bench. Its simply the take whatever it takes to maximize your clock game. If you want to have all of your cores at the same level, youre simply not playing it right. 

[Matsglobetrotter]

it makes sense  to measure max frequency and so far CPUz has done the best job of doing so since eons ago. Though i sometimes hate the crashes in validation ?

To simply say that we would not be turning off cores though does not make sense since most world records are based on adjusting the same which is a simple change in the bios. probably there since the first dual core was introduced.. I am quite sure some of the 3D benches would not have the records they have without some cores being turned off also, afterall some of them are not multithreaded.

Since now intel and i guess AMD also allow to turn off selected Cores and maybe some specific hyper threads on any core why would we not allow people to play around with it to max it out.

Meanwhile if it is possible to confirm that all cores are on ( i thought CPUZ was better at such confirmation but maybe not i do consider we need two categories. 

Am just now playing with my 3930K and it seems most CPUZ  validations and Superpi subs are with 2 cores and even 1 core at times. why would I then limit myself to 6 ?? and 300 Mhz slower CPU? and non competitive results on CPUz, Superpi, Pifast?

[Cautilus]

On 4/6/2021 at 12:41 AM, _mat_ said:

Maybe this is the right time to discuss that the validation of a fictional number like the CPU frequency is not a credible basis for competitive ranking.

I'm thinking a small workload would be more suitable for this kind of task (idea by @mllrkllr88). Something completely independent of memory, but big enough to show the scaling (in comparison to other CPUs of this generation). This workload could even come in multiple flavors like single-threaded, multi-threaded, AVX*, each separated into their own categories here on HWBOT.

Not an easy task all in all but a big improvement for the moderation and therefor the credibility of suicide runs.

Edit: It also makes a lot of sense to use effective clocks instead of the fictional numbers. Much more reliable (although harder to comprehend for the overclocker when benching).

Adding my two cents. I'm fine with max freq as is and I like mat's suggestion (above) of a new way to do "max freq valid" using some kind of very light workload instead of a CPU-Z valid.

I also like the suggestion of creating a new benchmark for all-core + HT on CPU-Z valids.

[Matsglobetrotter]

I like seeing new benchmarks being developed however i also realize that software seems to be a commodity that have short expiry date. The code is propriety and linked to a specific developer. The moment the developer get tired of the project the software somehow dies off. I liked Hwprime, Catzilla (when it worked) , XTU.  I also like CPUZ as its the one single longest lasting benchmark/measurement tools that seems to surpass all the generations of CPU's and motherboards. i guess because its the so far most robust tool for the same (despite some flaws). I wonder what the lifespan has been for anything else in relation to that. What is great about CPUz is that it has its own score board also thus has its own ecosystem irrespective of the bot. Meanwhile I only validate to put it here on the bot.  Fundamentally i think it would be dangerous for the longevity of the bot if it was dependent on inhouse developed benchers/tools for everything.  CPUz is used by all communities not just hwbot. Its also the standard for most reviews in most magazines globally. Meanwhile scores Mhz, Ghz wont mean anything if we suddenly change validation to lighter loads. Meanwhile not even HWinfo, Ohm, Sysinfo, none of them seems to be able to confirm cpu core speeds relevantly across all platforms. This was proven in some of the challenges lately where CPUz was showing one value, HW info another sampling every xx milliseconds and then futuremark made some scores disqualified as sysinfo showed spikes 500 mhz beyond what we could even get any single core to do in reality. 

I am non believer in creating yet a new solution if there is an issue in a software. Try fix the software first ? especially if its a very active one.

Meanwhile ofcourse CPUz is also the golden standard all hwbot submissions is demanding for frequency validations same as GPUz for gpuz.

[Splave]

Any ruling on this. Probs should know by the 4th ??

[Leeghoofd]

Nothing has changed Allen on the CPU part, you can still use XOC mode and only clock up one CPU core to reach maximum frequency.

For memory frequency the score needs to be checked by CPUZ website, so no more XOC mode there...

Fingers crossed for 8.0.1GHz

[jessec0626]

On 4/4/2021 at 6:26 PM, der8auer said:

The good thing is that we don't need a poll to decide this. I already talked to Albrecht last week to tell him we will not allow CPU-Z submissions anymore with disabled cores. We are currently thinking of when exactly we are going to enforce this rule. If already starting from this Gen or next one. You can give opinion on this if you want.

But for the future Only full cores + HT will be accepted. Everything else is just nonsense on a technical level.

this is booty right here. 

[Splave]

3 hours ago, Leeghoofd said:

Nothing has changed Allen on the CPU part, you can still use XOC mode and only clock up one CPU core to reach maximum frequency.

For memory frequency the score needs to be checked by CPUZ website, so no more XOC mode there...

Fingers crossed for 8.0.1GHz

thanks for the confirmation so no butts are hurt 

 

 

[I.nfraR.ed]

It would have been strange if you enforce "full cores + HT" for cpuz valids when you allow disabling "E" cores for other benchmarks. I think it is logical to say we need cpuz categories for P and E cores, what do you think?

After all, we should maximize the fun from a single processor and push it in every ranking we possibly can. 

Self-checkmate (if that's a term) 

[speed.fastest]

What next? only Prime95 48 hours stable with all cores and ht enabled stable for CPU-Z Validation maybe?

[Splave]

2 hours ago, speed.fastest said:

What next? only Prime95 48 hours stable with all cores and ht enabled stable for CPU-Z Validation maybe?

yes and give is 200 points, see who wants run ln2 that long

[Massman]

[FUGGER]

my bad

Edited November 2, 2021 by FUGGER

[I.nfraR.ed]

I have another suggestion.

Leave the P categories only, there aren't many globals with full cores, so they are pointless, IMO. Most people are not really interested in benching the full cores anyway it seems. Leave the 10x (P+E) category though for 12600K(F).

Super Pi is not interesting either, so maybe disable global points for it. Replace it with CPU-Z bench single thread instead and give that globals.

What do you think? :p

On a more serious note, have you thought about single-threaded benchmarks. Browsing Pi 32M scores (1 score), I can only see them in 8P category, but nothing in normal "I don't care about how many enabled cores are there" 12900K.

Is there a way to consolidate them or are people supposed to submit the same score twice (and get globals for both)? Sorry if that specific thing have been discussed before - if so, I've missed it.

PS: Btw, that 8GHz valid is also in my local forum as a news and I bet noone have seen the updated articles. Damage is done.

Edited November 4, 2021 by I.nfraR.ed