OSiBS season 4 design thread

[websmile]

Soon it will be time for old school is best school season 4. The competition series had good feedback over last years, so of course we decided to keep it alive and go for another season.

 

This thread is meant organize the five (5) rounds of season 4, our plan is to give the participants the chance to create their own rounds and thereby take in experiences of last years plus the knowledge on old hardware accumulated in community. It might take some work and discussion, but we want each round designed by another bencher, so we start early to have some time to get this done.

 

What do we need and want:

1. 5 rounds of 3 or 4 stages focused on same hardware (socket/cpus/vgas). Mixed rounds only raise cost and make things more difficult.

2. One or maximum of two rounds might include or focus on old server cpus which can be bought cheap. No professional cards after last years experience and please make sure to define hardware requirements precisely (socket, cpu, adapters allowed or not, and so on)

3. Please take in the possible cost of the round and hardware. I am not happy and others aren´t as well when you need to spend hundreds of euros to get rare unicorn hardware.

4. Please understand that we prefer 5 different members designing one round each this time, this doesn´t mean you cannot propose 2 or 3 different round concepts, some proposals might use similar hardware so we can merge for example. I would also preer different teams involved.

 

 

I will tag some guys here, this doesn´t mean they are the only ones allowed or preferred, but I want them to spread the news as well^^

 

@GRIFF @zafiropo @lanbonden @Mr\.Scott

 

So let´s start, feel free to post your proposals in this thread

Edited September 19, 2017 by websmile

[havli]

It is great to see another season of old school.

 

I'm sure there will be some very interesting rounds and stages. Maybe I can start with this proposal?

 

Platform: Pentium MMX + EDO RAM + any 3dfx Voodoo GPU (single or SLI)

Stage 1 = SuperPI 1M

Stage 2 = 3DMark 99

Stage 3 = Aida64 Memory Read

(optional) Stage 4 = CPU Frequency

 

3Dmark is CPU limited anyway, so Voodoo2 or Banshee should be more than enough, others are allowed just to extend the range of suitable HW. To make things more interesting, EDO memory must be used, no SDRAM. Feel free to use any board based on i430 series chipsets, VIA, SiS or others. Even Super 7 boards are fine as long as they have SIMM slots for EDO.

 

 

As for the server HW - perhaps some socket 603/4 action would be nice. Prestonia and Gallatin Xeons are cheap and easy to find. Boards are not so difficult to get either. Gallatin DP (socket 604) has 1 or 2 MB of L3 cache, Gallatin MP (socket 603) has 2 or 4 MB L3 cache... both are compatible with s604 boards and can be overclocked. Running something like SuperPI 32M could be fun.

[websmile]

Good start, I think this can be one of the rounds

[TASOS]

Tualatin socket 370 (non Intel chipset)

 

Reference clock

 

Memory clock

 

Aida64 memory read

 

superpi 1M

 

wprime 32M

 

 

//

 

 

Socket 423 (sdram only - no adapters allowed)

 

Reference clock

 

Memory clock

 

Aida64 memory read

 

superpi 1M

 

Cinebench 2003

 

 

//

 

 

Socket 939 (non Nforce chipset)

 

Reference clock

 

Memory clock

 

Aida64 memory read

 

superpi 1M

 

3DMark11 Physics

 

 

//

 

 

Fastest AGP (on single cpu socket only)

 

a) on Intel cpu systems

 

Stage 1 = 3DMark03

 

Stage 2 = 3DMark Cloud Gate

 

 

b) on AMD cpu systems

 

Stage 3 = 3DMark03

 

Stage 4 = 3DMark Cloud Gate

 

 

//

[Dead Things]

I like what havli & TASOS have suggested so far! Can I suggest 3DM05 instead of '03? It seems comps don't let us run it anymore. I like the Cloud Gate suggestion, though - don't get to play with that much either. Why non-Intel chipsets only for S370? I think a few of us are fond of the i815 chipset for Tualatin. S939 we already did this year, and it's also part of the TC. Can we maybe do S754 instead? I'd also love a chance to play with some Socket 8 or 604 stuff - those boxes have remained closed for far too long!

[lanbonden]

Sounds great with a new round of OSIBS, as discussed in the other thread Id say we take it a tad chill with requiring to special hardware. I would say @TASOS is on the right way but removing all the decent motherboards just feels wrong to me, a beter approach would be to choise some other core than the go to that everyone got binned already as CPUs are much easier to buy than new motherboards.

 

This year was:

Slot1

socket 462

Socket 478

Socket 939

Socket 771

 

For this year Ill check into it more this weekend but Id love a round of Pentium2 atleast. More to come

[TASOS]

I like what havli & TASOS have suggested so far! Can I suggest 3DM05 instead of '03? It seems comps don't let us run it anymore. I like the Cloud Gate suggestion, though - don't get to play with that much either. Why non-Intel chipsets only for S370? I think a few of us are fond of the i815 chipset for Tualatin. S939 we already did this year, and it's also part of the TC. Can we maybe do S754 instead? I'd also love a chance to play with some Socket 8 or 604 stuff - those boxes have remained closed for far too long!

 

Sounds great with a new round of OSIBS, as discussed in the other thread Id say we take it a tad chill with requiring to special hardware. I would say @TASOS is on the right way but removing all the decent motherboards just feels wrong to me, a beter approach would be to choise some other core than the go to that everyone got binned already as CPUs are much easier to buy than new motherboards.

 

This year was:

Slot1

socket 462

Socket 478

Socket 939

Socket 771

 

For this year Ill check into it more this weekend but Id love a round of Pentium2 atleast. More to come

 

My initial thoughts regarding the hardware i proposed , is to try out and see what we can achieve with less common used hardware.

 

We all know , what we can score with a Tualatin on BX or 815 chipset.

Lets try and see what's the case on SiS and VIA.

 

Same applies for my socket 939 choices.

No more Nforce ... let's try something new.

Plenty of VIA , SiS , ATI boards out there.

 

Wont be easy for anyone.

Let's all start fresh.

 

It would be a challenge !!!

[Dead Things]

It's a neat idea, but I honestly don't know how interested I would be in going out and spending more money on platforms I already have better equipment for, you know? I'm not one of those guys who has an endless supply of boards - I've got pretty much one per socket in most cases. So this sorta design would probably rule me out. And that's totally fine if it's what the majority would prefer to do!

Edited October 3, 2017 by Dead Things

[Bones]

Same here except it's usually the GPUs I have a problem getting.

Unless I get lucky and they specify something I already have I'll have to sit out the 3D part, esp nowadays with things the way they are here.

[TAGG]

 

Socket 423 (sdram only - no adapters allowed)

 

//

 

You mean RDRAM right? Because I'm pretty sure that there weren't any SDRAM Boards for that socket

 

One Thing I'd really like to see is one non Intel/AMD stage, something like CYRIX Socket 5/7 for example... However finding 5 benches that run could be really hard here...

 

The Fastest AGP would go nicely together with a stage that is limited by OS (like the win 98 stage from 2 years ago) imo. This would be great, because i think a wider choice of HW would increase participation more than overly specific stages(like S478 with SDRAM for example)... A good mix would be great, for example have two rounds with verry common HW so that everyone can compete and three with restrictions like TASSOS mentioned (ie. taking away the best boards/mems) for that stages...

 

Hope you can do something with that input

Edited October 3, 2017 by TAGG

[phobosq]

My ideas:

- extend MMX to Socket 7, so that K6 can compete as well, at some benches competition might be close

- 370 looks good, I'd exchange Aida to 2000/2001 with GPU capped to GF4xxx/R9600 (excluding 9500)

- have a round for 754 as well, 939 is well-exploited but 754 is a bit forgotten

- maybe also a round of something completely crazy, like legacy multi-CPU or Voodoo madness (with 1 stage for Voodoo2, 2 for 3 and 2 for 4&5)?

[havli]

My intention was to avoid K6... because this inevitably leads to >600 MHz K6-2+/3+. Which we already had in several competitions. On the other hand P MMX is always getting overlooked.

[mr.paco]

You mean RDRAM right? Because I'm pretty sure that there weren't any SDRAM Boards for that socket

 

For s423? Sure there are SDRAM boards.

Abit BW7, MSI 845 PRO (MS-6529), Gigabyte GA-8IDX3. Just to name 3

[yosarianilives]

I know nothing of Old School OC, the oldest I've OC'd was a a64 6400+ which is AM2. But I just wanted to mention that I look forward to what ever this ends up being and what platform I get a good excuse to learn and struggle with for a month or two. I think that something "exotic" like the 2p celeron board would be interesting, but I have a fondness for 2p systems. However I don't know how rare that board is and for comps you definitely need to pick hw that most people have reasonable access to something that qualifies.

[I.nfraR.ed]

Exploring obscure combinations of hardware sounds interesting, but that would probably rule me out, too.

Whenever I bench for some competitions, I'd like to combine this with hw points for the rankings and this almost always means the best possible mb-cpu-gpu combo.

So buying some motherboard with lesser performance is not the ideal scenario for me and would probably skip it.

[mr.paco]

I really do like TASOS line up but as mentioned with the proposed limitations it will definitely limit the number of participants because not many will have multiple variants of a platform they will usually have the optimum components only needed for a given platform.

Personally I dont have that issue .

But it would be so much more of a fun comp with as many participants as possible.

Not seeing many folks spending $$$ for Ol' skool parts they wont use again.

I realise the popular HW has been used n proven and the idea behind the limitations. Maybe lighter/looser limitations somewhere to attract more participants?

 

Just an opinion...

[Mr.Scott]

The object is 'old school'.

This will limit participation right off the bat, so that shouldn't even be a concern.

People that want to bench this comp will buy what they don't already have anyway. People that don't have or won't buy, are out. Simple as that.

Cater this comp to the true 'old schoolers' and quit trying to involve everybody. 'Old school' rarely sparks new interest in competitive benching as a whole.

[Dead Things]

I don't think we have to go sub-optimal to come up with stages that we don't all feel have been benched to death. Think about it... When was the last time you played with your Socket 4 gear? When was your last DP Socket 8 sub? Heck, even 386 can do SuperPi and we do have a whole month! I think there are still lots of different/unusual avenues left to explore without having to enforce sub-optimal setups.

 

edit - For example, I take no issue with benches like mem freq and maxmemm read on things like EDO and Rambus - but to force us to use those on benches for which they would be sub-optimal seems less interesting to me.

Edited October 4, 2017 by Dead Things

[websmile]

Good brainstorming here - I hope for a result of the five rounds that can be agreed on next days, no need to hurry that much but the goal needs to be achieved and I would definitely prefer this over me having to chose in the end

[lanbonden]

I promised my idea for the stages last weekend but real life hits hard sometime.. Anyways here it is somewhat inspired from @TASOS but with my view on how to limit it based on CPU family instead of using bunnyextractionty motherboards so we can still get good scores out of the runs, but on somewhat forgotten hardware. It’s also limited to 3 stages per round if it collides with other competitions.

This is hardware I’ve never seen used in a competition so should be an even race for everyone.

Pentium II + Riva TNT

superpi 1M

wprime 32M

3dmark01

 

Socket A Duron (spitfire) + geforce2 MX

HWBOT Prime

Cinebench 2003

Aquamark

Pentium 4 (Willamette) + Radeon 8500

superpi 1M

GPUPI for CPU - 100M

3dmark03

 

Socket 604

SuperPi - 32M

wPrime - 1024m

GPUPI for CPU - 100M

 

Fastest AGP (on single cpu socket only) (I liked this one)

 

a) on Intel cpu systems

Stage 1 = 3DMark03

Stage 2 = 3DMark Cloud Gate

 

b) on AMD cpu systems

Stage 3 = 3DMark03

Stage 4 = 3DMark Cloud Gate

 

 

Im absolutly open for discussing the choise of what benchmarks to use as this is just a crude draft but this is my suggestion on hardware to use. Maybe change 604 to something more accessible but I think it looks like a fun socket that Ive always wanted to take a bite of.

Edited October 5, 2017 by lanbonden

[Aleslammer]

Fastest AGP, going to fix a platform or anything goes?

[lanbonden]

Fastest AGP, going to fix a platform or anything goes?

 

In my example it meens that you need to run the benchmarks with any AMD and Intel system that supports AGP.

 

I think we all know it means an Radeon HD3850 with the Asrock AM2nf3-vsta and 4coredual-vsta motherboards but it will be fun to finally see whats the best choice for us AGP overclockers.

[TASOS]

I promised my idea for the stages last weekend but real life hits hard sometime.. Anyways here it is somewhat inspired from @TASOS but with my view on how to limit it based on CPU family instead of using bunnyextractionty motherboards so we can still get good scores out of the runs, but on somewhat forgotten hardware. It’s also limited to 3 stages per round if it collides with other competitions.

This is hardware I’ve never seen used in a competition so should be an even race for everyone.

Pentium II + Riva TNT

superpi 1M

wprime 32M

3dmark01

 

Socket A Duron (spitfire) + geforce2 MX

HWBOT Prime

Cinebench 2003

Aquamark

Pentium 4 (Willamette) + Radeon 8500

superpi 1M

GPUPI for CPU - 100M

3dmark03

 

Socket 604

SuperPi - 32M

wPrime - 1024m

GPUPI for CPU - 100M

 

Fastest AGP (on single cpu socket only) (I liked this one)

 

a) on Intel cpu systems

Stage 1 = 3DMark03

Stage 2 = 3DMark Cloud Gate

 

b) on AMD cpu systems

Stage 3 = 3DMark03

Stage 4 = 3DMark Cloud Gate

 

 

Im absolutly open for discussing the choise of what benchmarks to use as this is just a crude draft but this is my suggestion on hardware to use. Maybe change 604 to something more accessible but I think it looks like a fun socket that Ive always wanted to take a bite of.

 

I chose only one (double stage) with 3D , to reduce the total cost for people that dont have the hardware.

Assuming that most of us would use the same VGA for both stages.

 

No need for special VGA's , on all other stages.

 

 

... and i really like to see the following (2) stages in action.

 

 

Tualatin socket 370 (non Intel chipset)

 

Reference clock

Memory clock

Aida64 memory read

superpi 1M

wprime 32M

 

//

 

 

Socket 939 (non Nforce chipset)

 

Reference clock

Memory clock

Aida64 memory read

superpi 1M

3DMark11 Physics

 

//

[TASOS]

You mean RDRAM right? Because I'm pretty sure that there weren't any SDRAM Boards for that socket

 

Just got back.

I was away for work this week.

 

I saw Mr Paco , already answered your question

[TAGG]

Just got back.

I was away for work this week.

 

I saw Mr Paco , already answered your question

 

It really suprised me, i only have seen ones with RDRAM before... Plus 423 was a really shortlived platform, personally i'd preffer S478 Willamette for the competition (may be paired with SDRAM only) because the chips and boards are verry common So basically combining your idea with lanbondens...