Jump to content
HWBOT Community Forums

G. SKILL OCWC 2018 benchmark concernings


Dancop

IGP and full out - is it a mem or a cpu competition?   

11 members have voted

  1. 1. IGP - FULL OUT OR FOCUSING MEM INSTEAD OF CPU?

    • Leave the benchmarks like they are?
      8
    • Eliminate igp and limit the freqs?
      2
    • Eliminate igp, keep full out
      0
    • Keep igp, limit freqs
      1

This poll is closed to new votes


Recommended Posts

Hey fellas,
Maybe some of you are already laughing about my post on Facebook, but I'm serious!

I really like the competition and the organization from year to year. But another igp stage is just too much, to stay calm.

For this competition u need to bin the following:

Igp 8600k: at least 3 chips to find one decent

Imc 8600k: at least 3 chips to find one great

Core 8600k: at least 5 chips to find one great

RGB mem for max freq and normal for the rest!

Open freq=core binning

Igp=igp binning

Max mem=imc binning

 

Limit the freq to 6g/6g and eliminate igp, instead choose 3d11 physx

Limited freq shouldn't be any issue with Futuremark, geekbench, x265, xtu!

Im a huge fan of 32m...so why not running a video with 32m??? On the other hand... With 8600k u have 6 cores where one might be great if u don't like the video idea...

 

What do you guys think? Leave it or discuss it?

 

Link to comment
Share on other sites

No idea how to... Who cares... Dicisions are made. I reported the 3d01 igp issue together with the 3d11 issue the first year. They took out the 3d01...noone else had the ballz to do it! I could have just used it to enter the finals... I didn't! Verification link in 3d11 wasn't mandatory as as u bench in win7... I proved that ist has to be a must, otherwise igp can give u easily very high scores...

From that moment on and after noticing the huge quality difference between chip and chip, I really try to avoid benching this BS

Link to comment
Share on other sites

Looking at things, we already have:
1) Mandatory Full-out DDR4 Mhz
2) Full-out CPU Multicore (Geekbench 3)
3) Full out CPU Single-Core (32M)

If full-out IGP wasn't desirable, I agree that limited-clock CPU stage on a bench that relies heavily on memory could be a substitute, as long as we have a reliable monitoring system that's not really complicated (needing a video would be slightly too much), Futuremark Sysinfo worked for Galax GOC limited clock I guess? 

So that's left us with 3DMark11 Physics, Time Spy Physics, or Time Spy Extreme Physics. (Fire Strike Physics doesn't really scale much with mem IMHO). If I have to suggest anything, I'd do Time Spy / Time Spy Extreme Physics since it was less used compared to 3D11 Physics.

  • Thanks 1
Link to comment
Share on other sites

If you have a solid 3dm11 physics score and not from bruteforcing cpu core then graphics score will follow. Nearly every intel igpu can do around 1900 mhz even on ss, and unless you get an extra crap igpu you'll be within 100-200 mhz of anyone else. Also I've found very little core scaling from 1600 up to 1900 on hd graphics on 7700k. So it's mainly mem OC on igpu. Tertiary and secondary timings as well as high cpu cache clock are way more import primary timings. Also you have to use the right driver, sometimes as much as 10-30% difference from using best vs worst driver on intel igpu. I see no reason why you would need to bin igpu if you have a good imc. But of course your average hwbot member doesn't want to go outside their comfort zone and learn a new way of OCing so perhaps we should throw it out. Cause new things are evil.

  • Like 2
  • Thanks 2
  • Haha 1
Link to comment
Share on other sites

1. Benchmark is 3dmark11 performance not physx!

2. 1900 igp on ss??? Kk im a Noob!

3. 1900 on every igp??? Kk Noob again!

4. Telling me things about mem settings... Maybe you are right, who knows.... BUT!!!

DO YOU REALLY THINK I'M NOT OPEN FOR NEW STUFF??? I did igp on a competitive way at least 5 times, how often u did?

me for instance, I'm not afraid of loosing, just afraid of spending too much money in different chips, cause one thing is for sure... 1900 almost every igp is just a guess! Proof needed from your side man!

Link to comment
Share on other sites

1 hour ago, yosarianilives said:

If you have a solid 3dm11 physics score and not from bruteforcing cpu core then graphics score will follow. Nearly every intel igpu can do around 1900 mhz even on ss, and unless you get an extra crap igpu you'll be within 100-200 mhz of anyone else. Also I've found very little core scaling from 1600 up to 1900 on hd graphics on 7700k. So it's mainly mem OC on igpu. Tertiary and secondary timings as well as high cpu cache clock are way more import primary timings. Also you have to use the right driver, sometimes as much as 10-30% difference from using best vs worst driver on intel igpu. I see no reason why you would need to bin igpu if you have a good imc. But of course your average hwbot member doesn't want to go outside their comfort zone and learn a new way of OCing so perhaps we should throw it out. Cause new things are evil.

So you are saying that last years people didn't find proper driver and everyone failed at igp benching :D

 

  • Thanks 1
Link to comment
Share on other sites

7 minutes ago, Alex@ro said:

So you are saying that last years people didn't find proper driver and everyone failed at igp benching :D

 

Hopefully we'll find out soon.  Yos has a 7700K that did 1900mhz+ IGP in 3Dmark03 so it should be no problem at all for him to smoke all the scores made on 7350K last year.  I do believe him that it's more about memory than IGP core, but the proof is in the benching.

Edited by mickulty
Link to comment
Share on other sites

Maybe the major concern here was benching the IGP hard has a potential to kill more chips than benching CPU-only? (Not sure if this still happened on CFL, also this is a pro-class competition, so dying chips was normal risk us overclockers have to take).
I don't mind IGP benches, did one time back in 2016 with 6700Ks. Would've been fine for me if the rules was unchanged, but I also open to discussions and concerns regarding the benchmark/rule and whether it was necessary to be changed in the end. 

Waiting for decision then :)

Link to comment
Share on other sites

7 minutes ago, Alex@ro said:

It is 7700k remember not 7350K, different IGP and also 7700K clocked better on average than 7350K, so this is apples vs oranges....

Amusingly it's the same category on hwbot but yeah the 7700K's bigger cache is a fairly big advantage.  Maybe we'll get lucky and he'll have a 7350K.

Link to comment
Share on other sites

  • Administrators

 On the 630, sure cpu will make a difference even though igp itself is same at specs, core count and cache will influence physics and combined.

On the poll and discussion, I can tell you that the choice of benchmarks was done by G.Skill, it is same benchmarks we had last year for the qualifier on which all worked quite OK. Not sure you are asking the right people to change it...

Link to comment
Share on other sites

So I've been hearing in other forums a bit of confusion what I mean by "most intel igpu should do 1900 mhz". I'm talking i5 8600k ofc. I've seen this every gen, top end i7 clocks igpu best, mid tier cpu second best, and bottom tier has lowest clocks. Now most gens it's not so far, look at sandy for example where my 2600k does 2260 on igp, but my pentium g870 only does 1860. The top mhz for intel igp has gone down a bit each gen. So now on 7700k I'd say top is probably around 2000 mhz for a good chip, however if you go down to i3 you'll find hot garbage. I think the fact that the cpu clocks core and memory generally lower than average 7700k should give a good indication of what to expect. 

On SS my 7700k can run 5.8/5.6 core/cache with the igp at 1926 mhz with 1.7v and 1.75v respectively. However on same cooling my 7350k does like 5.3/5.1 core/cache at 1.6-1.7v and 1530 mhz igp at 1.6v+ and you can forget 4000 mhz mem. I hope you don't misunderstand me to say that any issue with your core clocks on 7350k was cause you didn't know how to clock igp, I'm more saying that you'll see far more scaling from mem than igp core. For example just testing on the 7350k today I was scoring around 1800 graphics score. Increasing to 1500 mhz boosted me to 2100. However increasing mem to 3930 (best this crap IMC could do) boosted me to over 2500 pts. You can see that if an igp was 100 mhz behind you could easily make it up with more mem, especially in 8600k which has a much better imc. Main takeaway is that 8600k is gonna be nothing like 7350k was, it should be a much better experience and much less hot garbage. Shame I don't have/want z370 cause you're all making me feel like I'd almost be competitive on ss with your attitude towards this stage. Unless intel decided that now i5 get's binned like i3...

Link to comment
Share on other sites

Have u tried more then one 7700k igpu? Maybe you are lucky and yours is decent. I tried 12x 7350ks and the max clock range was massive. Its more you need to bin. Do u think in a memory competition people will all be pushing the mem as hard as possible also so gains from pushing mem because your igp sucks is kind of a non issue imo. 

Score wise are you using the same driver that rsan is? Maybe a newer one also scores better. Make sure to compare apples to apples.

I appreciate your passion for igp. Maybe you have found your niche. :D

  • Haha 1
Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

×
×
  • Create New...