@ hwbot inventor | please separeted championship for INTEL and AMD

i have an importent concerns!

i love to overclock AMD CPUs

but there is a big problem

INTEL is not AMD and so a benchmark can not compare INTEl and AMD

a realy good sample is superpi1m

a E6600 @ default use in superpi1m between 21 seconds and i use and AMD Windsor @ 3,8 ghz and has the same time! but in most of other benchmarks i am much faster than an E6600 and has the performance with this Windsor @ 3,8 ghz of an E8300 @ default

Superpi1m : AMD X2 6400+ @ 3750 MHz (21,87s) | E6600 @ 2400 MHz (21,00s)

WPrime32m : AMD X2 6400+ @ 3740 MHz (23,092s) | E6600 @ 2400 MHz (69,265s)

and this is an realy big problem. because i can not get global points!

so why could hwbot not make an seperate range for INTEL and a seperate for AMD like ripping.org?

hope the hwbot inventor can help me and the other AMD overclocker and chanche this from one championship to one championship for INTEL and a seperate for AMD user

nice greetings from Techtrancer

Being AMD ocer as well I vote for the same, ofc

i think it would be best, if there are more different championships with global points

one for AMD CPU benchmarks with global point

one for INTEL CPU benchmarks with global points

one for AMD + NVIDIA benchmarks with global points

one for AMD + ATI benchmarks with global points

one for INTEL + NVIDIA benchmarks with global pointspoints

one for INTEL + ATI benchmarks with global points

i think it would be best, if there are more different championships with global points

 

one for AMD CPU benchmarks with global point

one for INTEL CPU benchmarks with global points

one for AMD + NVIDIA benchmarks with global points

one for AMD + ATI benchmarks with global points

one for INTEL + NVIDIA benchmarks with global pointspoints

one for INTEL + ATI benchmarks with global points

That would make things too complicated, I'm afraid. The design has to be held as simple as possible, otherwise the point distribution service will slow down the site completely AND, more importantly, no one will understand how the hwboints actually work.

There are blueprints for a system that goes beyond the points-for-result concept, but the implementation requires us to clean up every page of code ... which would take months to complete. Frederik is doing all he can in the little spare time he has and since this still is a non-professional database (no one gets paid, fred is paying to keep it running) we have no funds to hire a coder at the same level of RB to do the cleaning for us.

We're still trying to find an alternative solution for this problem, though .

yes! may be it will be complicated!

but i think with this 6 different championships it will be the best distribution points for all and so the AMD bencher also can make global points

and the point that a benchmark can be INTEL ore AMD optimized is negligible

big thx that you are working on it!

Welll I personally like things how they are at the moment. Just push your Phenom II with some strong GPUs and you surely get some global points in 3D Also in wprime you can get some, still behind all i7 but still. So I don't think we need separation.

if it will not be different, an AMD overclocker will never have a chance to get global points in superpi!

and in some 3D benches, the AMD must have more than 1000 mhz more to have the same points like an INTEL

yes! may be it will be complicated!

 

but i think with this 6 different championships it will be the best distribution points for all and so the AMD bencher also can make global points

 

and the point that a benchmark can be INTEL ore AMD optimized is negligible

It's not 6 different championships, it's WAY more.

one for AMD CPU benchmarks with global point

one for INTEL CPU benchmarks with global points

one for AMD + NVIDIA benchmarks with global points

one for AMD + ATI benchmarks with global points

one for INTEL + NVIDIA benchmarks with global pointspoints

one for INTEL + ATI benchmarks with global points

2D benchmarks: AMD / Intel / other manuf.

3D: AMD / Intel / other + Ati / Nvidia / other + Single / multigpu

* 4 categories of cpu's

* 7 categories of 2D benchmarks

* 9 categories of gpu's

* 2 categories of gpu configuration

* 6 categories of 3D benchmarks

2D - 4 x 7 = 28 global categories

3D - 4 x 9 x 2 x 6 = 432 global categories

Total: 460 global rankings

If we add more benchmarks in the future, the categories will expand even further.

big thx that you are working on it!

No problem, but please understand that the current problem is not the theoretical design of a complete new system, but the time available to implent it. Don't expect drastic changes in the near future.

yes! the time is the biggest problem!

when do you think that it would be feasible

yes! the time is the biggest problem!

 

when do you think that it would be feasible

There simply is no date or timeframe.

As I explained, the new system requires a total clean-up of the current structure and code of the HWBot database, features and website. That clean-up takes roughly 3 months (estimated time by Frederik) ... full time; given that Frederik has a daily job, he can only work on hwbot maybe 1h a day (7h a week).

3 months at 8h a day = 736h

736h at 7h a week = 105 weeks ~ 2 years.

Note that these are fairly rough calculations and only to explain my point here: even if RB started the clean-up in his spare time, using all the time he has available, it would still take more than 2 years before the clean-up is completely done. That would also mean that he spends NO time on other features/bugs in those 2 years ...

In other words, in the current situation (1 coder, little time, no income) it's practically impossible to redesign the current hwboints system. Little changes are no problem, though, for instance: adding the sli function, adding benchmarks, changing cap limits, ...

I think hwbot don't need this!

AMD must build CPU wich are strong enought to beat intel and not hwbot must be changed...

ok for SuperPi is will be OK, but why for 3D's?

Oh, forgot to mention this. At the moment we have no plans to split up 3D rankings into AMD/Intel. 3D = vga card benchmark .

It would be nice to split SPi 1/32M and perhaps PiFast since these are the benchmarks that cannot be compared directly.

Well, it was worth the shot Then how about adding the checkbox when searching e.g. SuperPI 1M benchmark results to return Intel or AMD only?

Adding a checkbox would be usefull, I totaly agre. Separating the categories is not.

I don't think splitting categories for cpu's is a good idea top total needs to always be the best performer.

But on the other hand I don't see why a sub category standing for each chip maker could not work?

something like:

each chip > best of each brand > global best.

Yeah I also agree that the rankings not be split up. If amd makes a cpu that can compete and get global points fine. No reason to complicate things much more by adding in additional rankings. A search function to only return amd cpus results should be enough.

With Deneb I think it's no real need to separate the ranks. There was a screenie at 7.1ghz or so, I guess 6.9 pistable is possible (at least), I bet the score will be sub 10 at that speed. Which makes it comparable to the Conroe's, actually.

It would, however, be nice to have access to some list of benchmarks done with a specific core, for example. You can't fight for AMD pi WR's by using 90nm, or even 65nm, these days anyway;)

@ benchbros

superpi is not the only INTEL optimized benchmark

superpi is only the benchmark where everyone knows that it is INTEL optimized

i dont make an comparative in other benchmarks. but i know in cinebench it is the same problem!

and maybe it is in other benchmarks like 3dmark ore so the same problem

@ TheKarmakazi

in superpi intel wil be every time faster, because it is an old application wich AMD dont evolve again

when AMD K8 was hammering Intel P4 it was Intel owners disregarding the other side

this IS a fair comparison. why penalise a much better performing procesor

if you really allow AMD to be equal in points and SO much slower in real life are we then going to give trameta crusoe or via CPUs a fair go

are we going to then have separate Nvidia and ATI and make equal points even thought one dominates a certain benchmark compared to other like for example 3DMARK Vantage.....ATI is absolutely shit at it.....so are we going to now give them equal points because they are shit at it

this makes no sense

the current system works fine in my opinion and anything suggested by OP would be absolutely unfair to a much superiour tech of the day whether that is going to be Intel or AMD today or tomorrow it doesnt matter

have you seen the benchmarks in my first post?

there ist an E 6600 @ default in superpi1 a litle bit faster than a Windsor @ 3,75 ghz

and in wprime the same CPUs with the same setting and the windsor use the half time!

and in every other benchmark the windsor @ 3,75 ghz is also much faster than the E6600 @ default because @ 3,75 ghz the windsor is rating an E8300 default

thats right

Intel is just a lot better at those benchmarks

it should not be penalised otherwise you open a pandora's box

what AMD needs to do is improve its processor to compete not HWBOT adjusting to compensate for their poor performance

as i said before we had a situation in the past where AMD K8 was FAAAAAAAR better than P4...it was even a bigger difference that this

i think you dont understand what i want

many banchmarks are INTEL optimiced

and if you bench there an INTEL who has the same power like an AMD, the INTEL is in the banchmark faster

you could se it in the screens in the first post

my windsor has the double power of an E6600 @ default in wprime, but in superpi the E6600 is at the same setting faster because the benchmark is INTEL optimiced

i do understand what you're saying

you cant really compare across benchmarks to determine how a memory based benchmark compares to raw MHz computation power based benchmark

wprime is purely CPU MHz

superPi is RAM efficiency

what you really need to do is email AMD if you think they are not optimising properly for superpi to resolve it not HWBOT to make adjustments

if you RAM wprime on old hyperthreaded P4 against K8 at such different frequencies you would come up with exactly the same conclusion >> K8 absolutely smashing P4 in Pi but losing wprime >> is K8 better than P4 >> WHO CARES...thats not the point at all..........[edit]looked up wprime hard to say but its roughly about 5% faster with same speed differential)

Yes I agree with what dinos is saying. When AMD was on top and beating Intel bloody, many AMD users were very smug and quite happy with how things were. Now the tables are turned, so why should AMD get special treatment? They shouldnt. They should just design a more efficient processor...