Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

Processor speed

RAM speed

CPU Cache size

Disk speed (Maybe for loading? was a suggestion)

[Fleeting Frames]

Not likely. I think 64 bit did give some performance improvements due better low-level optimization, but nothing "bring back forts from fps grave" level. It'd be slightly less bad but still not so good if upgraded.

[Nagidal]

Added my line, although I'm not sure if I tested it in the proper conditions. I downloaded the linked LNP, but it wouldn't start DF (DF then always crashed). I had to start DF by manually executing Dwarf Fortress.exe. And I did not change any settings, except for FPS:YES. I then let it run for about a minute in the saved position and observed the FPS counter.

Also, I don't know where do I find Processor Priority. Do you mean the process priority of Dwarf Fortress.exe by any chance? (I did not find that either.)

Also I found the window size a bit unusual (too wide), and it looks like it uses a graphics pack and TWBT, and DF hack. why don't we test things in a vanilla environment, or vanilla+DF hack?

[Zachst]

Added my line, although I'm not sure if I tested it in the proper conditions. I downloaded the linked LNP, but it wouldn't start DF (DF then always crashed). I had to start DF by manually executing Dwarf Fortress.exe. And I did not change any settings, except for FPS:YES. I then let it run for about a minute in the saved position and observed the FPS counter.

Also, I don't know where do I find Processor Priority. Do you mean the process priority of Dwarf Fortress.exe by any chance? (I did not find that either.)

Also I found the window size a bit unusual (too wide), and it looks like it uses a graphics pack and TWBT, and DF hack. why don't we test things in a vanilla environment, or vanilla+DF hack?

Processor priory is in Task Manager on windows (Yes DwarfFortress.exe) The game doesn't have an INIT setting for it.  Your FPS should be fine as its the same save. By default TEMP and WEATHER are on in the INIT file, so ill update your entry on the spreadsheet.

[em312s0n]

The highest one so far is from a system with DDR4 memory. Now from what I know someone on the forums actually did testing on CAS latency and found that it was significantly better if the memory's latency was lower.

I dont exactly know how to calculate for the exact latency from the specs posted on the document but the latency on those should be largely slower than a DDR3 right? Can someone please calculate for the latency on those?

EDIT: I just realized that these things run on quad channel instead of dual. Now if that's the case, going quad channel has a HUGE impact on performance despite the high latency with DDR4.

[BauxiteProcessor]

The highest one so far is from a system with DDR4 memory. Now from what I know someone on the forums actually did testing on CAS latency and found that it was significantly better if the memory's latency was lower.

I dont exactly know how to calculate for the exact latency from the specs posted on the document but the latency on those should be largely slower than a DDR3 right? Can someone please calculate for the latency on those?

EDIT: I just realized that these things run on quad channel instead of dual. Now if that's the case, going quad channel has a HUGE impact on performance despite the high latency with DDR4.

Could you track down that testing on the effects of latency? I'd like to see it.

You calculate ram speed by dividing latency by frequency (lower is better).

The 6700k is dual channel, not quad channel.

[Devast]

Just did a quick test. Installed the game onto a Samsung 950. I had EVE online in the background (docked of cause) and was getting between 75-100 fps

[wierd]

The highest one so far is from a system with DDR4 memory. Now from what I know someone on the forums actually did testing on CAS latency and found that it was significantly better if the memory's latency was lower.

I dont exactly know how to calculate for the exact latency from the specs posted on the document but the latency on those should be largely slower than a DDR3 right? Can someone please calculate for the latency on those?

EDIT: I just realized that these things run on quad channel instead of dual. Now if that's the case, going quad channel has a HUGE impact on performance despite the high latency with DDR4.

Could you track down that testing on the effects of latency? I'd like to see it.

You calculate ram speed by dividing latency by frequency (lower is better).

The 6700k is dual channel, not quad channel.

Could be an artifact of better cache management technology between processor releases.

[em312s0n]

The highest one so far is from a system with DDR4 memory. Now from what I know someone on the forums actually did testing on CAS latency and found that it was significantly better if the memory's latency was lower.

I dont exactly know how to calculate for the exact latency from the specs posted on the document but the latency on those should be largely slower than a DDR3 right? Can someone please calculate for the latency on those?

EDIT: I just realized that these things run on quad channel instead of dual. Now if that's the case, going quad channel has a HUGE impact on performance despite the high latency with DDR4.

Could you track down that testing on the effects of latency? I'd like to see it.

You calculate ram speed by dividing latency by frequency (lower is better).

The 6700k is dual channel, not quad channel.

It was this one I believe. http://www.bay12forums.com/smf/index.php?topic=119474.msg3812737#msg3812737

Also, isnt it dependent on the motherboard if it was quad or dual channel?

[BauxiteProcessor]

The highest one so far is from a system with DDR4 memory. Now from what I know someone on the forums actually did testing on CAS latency and found that it was significantly better if the memory's latency was lower.

I dont exactly know how to calculate for the exact latency from the specs posted on the document but the latency on those should be largely slower than a DDR3 right? Can someone please calculate for the latency on those?

EDIT: I just realized that these things run on quad channel instead of dual. Now if that's the case, going quad channel has a HUGE impact on performance despite the high latency with DDR4.

Could you track down that testing on the effects of latency? I'd like to see it.

You calculate ram speed by dividing latency by frequency (lower is better).

The 6700k is dual channel, not quad channel.

It was this one I believe. http://www.bay12forums.com/smf/index.php?topic=119474.msg3812737#msg3812737

Also, isnt it dependent on the motherboard if it was quad or dual channel?

Yes it does but the 6700k fits into LGA 1151 sockets and no motherboard with a 1151 socket has quad channel memory. The 6700k is a high end mainstream CPU. Those do not get quad channel ram. You need to get an enthusiast CPU like the 6850k or a Xeon if you want quad channel ram.

Which produces the only real question for DF performance that isn't basically trivia or a matter of performance per dollar; which is better, the high end mainstream which has the best single core clockspeed or the enthusiast which has quad channel ram and more cache. Ideally Skylake-x will have the increased overclockability of Kaby Lake in addition to it's improvements (since Kaby Lake basically _is_ Skylake) which will simplify DF computer building for a while at least.

[em312s0n]

The highest one so far is from a system with DDR4 memory. Now from what I know someone on the forums actually did testing on CAS latency and found that it was significantly better if the memory's latency was lower.

I dont exactly know how to calculate for the exact latency from the specs posted on the document but the latency on those should be largely slower than a DDR3 right? Can someone please calculate for the latency on those?

EDIT: I just realized that these things run on quad channel instead of dual. Now if that's the case, going quad channel has a HUGE impact on performance despite the high latency with DDR4.

Could you track down that testing on the effects of latency? I'd like to see it.

You calculate ram speed by dividing latency by frequency (lower is better).

The 6700k is dual channel, not quad channel.

It was this one I believe. http://www.bay12forums.com/smf/index.php?topic=119474.msg3812737#msg3812737

Also, isnt it dependent on the motherboard if it was quad or dual channel?

Yes it does but the 6700k fits into LGA 1151 sockets and no motherboard with a 1151 socket has quad channel memory. The 6700k is a high end mainstream CPU. Those do not get quad channel ram. You need to get an enthusiast CPU like the 6850k or a Xeon if you want quad channel ram.

Which produces the only real question for DF performance that isn't basically trivia or a matter of performance per dollar; which is better, the high end mainstream which has the best single core clockspeed or the enthusiast which has quad channel ram and more cache. Ideally Skylake-x will have the increased overclockability of Kaby Lake in addition to it's improvements (since Kaby Lake basically _is_ Skylake) which will simplify DF computer building for a while at least.

I see. So you think the results from the DDR4 system on the doc was largely due to the CPU then?

[Zachst]

Zachst: based on the spreadsheet, it seems you're making some headway for higher FPS.
Will you be publishing your results and methods in the future?

Hopefully ill be making more graphs like this for RAM, CPU Cache and other metrics- I just need more data

[Zachst]

The highest one so far is from a system with DDR4 memory. Now from what I know someone on the forums actually did testing on CAS latency and found that it was significantly better if the memory's latency was lower.

I dont exactly know how to calculate for the exact latency from the specs posted on the document but the latency on those should be largely slower than a DDR3 right? Can someone please calculate for the latency on those?

EDIT: I just realized that these things run on quad channel instead of dual. Now if that's the case, going quad channel has a HUGE impact on performance despite the high latency with DDR4.

Could you track down that testing on the effects of latency? I'd like to see it.

You calculate ram speed by dividing latency by frequency (lower is better).

The 6700k is dual channel, not quad channel.

It was this one I believe. http://www.bay12forums.com/smf/index.php?topic=119474.msg3812737#msg3812737

Also, isnt it dependent on the motherboard if it was quad or dual channel?

Yes it does but the 6700k fits into LGA 1151 sockets and no motherboard with a 1151 socket has quad channel memory. The 6700k is a high end mainstream CPU. Those do not get quad channel ram. You need to get an enthusiast CPU like the 6850k or a Xeon if you want quad channel ram.

Which produces the only real question for DF performance that isn't basically trivia or a matter of performance per dollar; which is better, the high end mainstream which has the best single core clockspeed or the enthusiast which has quad channel ram and more cache. Ideally Skylake-x will have the increased overclockability of Kaby Lake in addition to it's improvements (since Kaby Lake basically _is_ Skylake) which will simplify DF computer building for a while at least.

I see. So you think the results from the DDR4 system on the doc was largely due to the CPU then?

DDR4 isn't some revolutionary new technology. Literally all it is is DDR3 that uses less power and can have higher clock speed. if you were to clock modern DDR3 and DDR4 at the same speed the performance would be more or less the same. (However DDR4 currently has faster clock speeds, but much higher latency)

I think the higher Cas latency of DDR4 will off-set any benefits of higher clock speeds in an application like dwarf fortress
I think the higher FPS on the DDR4 comes down to the fact that they're both running the best possible Skylake processors with the 5th best per-thread performance of any CPU (Of course we need more data to confirm my speculation (Thats the whole point of this thread)

Spoiler (click to show/hide)

https://www.cpubenchmark.net/singleThread.html

[BauxiteProcessor]

Yeah and the 4790k is comparably placed while also having fairly close performance to a 6700k which supports that.