did you not read what i said about SSE3 earlier?

they are mainly media extensions, as in, decoding video's.

they aren't much use for video games

Then I guess he's wasting his time?

Tried it before using SB, then dropped it. Who knows, maybe in conjunction with SB it might have a useful effect. Will see.

of course it's not the case, dude. I've already written more than a couple of times that I'm running with Vsync enabled. Always have, always will, I'm not even interested in the presentinterval variable.

As per above it might be worth trying SB together with a fps limiter. Right now I'm playing total vanilla and I think I want to stop experimenting for a while and enjoy the game. I'm keeping a tight eye on this thtread, though.

Aight, been using the acceleration layer for a while and saw this thing mentioned in one of their threads but seeing as its gotta little more presence decided to give it a try, wow.

Dragonreach->Tree view, beyond ultra settings at 5922x1080 resolution (the extra is bezel compensation):
Without either: 32 FPS
With Accelereration layer: 42 FPS
With Skyboost R3: 50 FPS

Incredible, excellent work. I am now a fan for sure.

So anyone thinks Oblivion would benefit from some optimizations like on Skyrim? wonder is anyone would do that.

It would. But why bother at this point (and this coming from a OB modding old timer)?

Definite improvements with r3, great work, I may actually start playing without fraps now as I'm not obsessed about low fps

Dude, stop spreading this bs. Just so you know both TESVAL and SkyBoost are already using SSE3. An important addition to SSE3 was so called horizontal SIMD instructions, and Arisu's original assembly code is making use of them in dot product and visibility calculations ( in a form of haddps instruction to perform sum phase of dot product calc ). And those are very relevant to games.

Umm and looking at those functions - they can be redone in SSE4.1, for example:

VOIDASMPROC dot_product()
{
__asm
{
mov eax, [esp+4]
movups xmm0, [ecx]
movups xmm1, [eax]
pslldq xmm0, 1*4
pslldq xmm1, 1*4

dpps xmm0, xmm1, 241

movss [esp-4], xmm0
fld dword ptr [esp-4]
retn 4
}
}

same can be done with dot_product_minus_unknown() and determine_visiblity() is calculating dot product as well.

Problem is that this code is either ~same speed or slower - the bulk of work is beeing down in those function setup/teardown, unaligned mem accesses and shifts. Real optimization would require patching call sites with:

movups xmm0, mem1
movups xmm1, mem2
pslldq xmm0, 1*4
pslldq xmm1, 1*4

dpps xmm0, xmm1, 241

movss mem1, xmm0
fld dword ptr mem1

But that is way too many bytes and larger than call sites.

:rollseyes:

its not bull****, the dot product optimisations are basically useless. they provided little improvement to K8/10 (Its been proven that SSE3 should be avoided on AMD even when hand written btw) and the SSSE3 Core 2 extensions provide a far rewarding improvement to performance on the cpu's that support it.

you're right about the 4.x dot product optimisations, though they will be more noticed on nahalem than on penryn.

I'm running on SSE1 processor Athlon xp.
OK, I'm a rookie in this, but is it possible to create injection which will operate memory and Cpu threads like the Unreal engine 3 does ?
UE3 games runs excelent on Single core cpu's (12 years old), not to mention Dual core ones. Is this possible to achieve ?

Also, I don't know about this Creation engine. Isn't this updated buildt of Oblivion engine ?

It would be pointless. Just about anyone with ether an ATI/AMD or Nvidia GPU can easily run Oblivion way beyond it's highest setting.

Yeah FPS stopped being a problem in Oblivion years ago, it was those loading stutters no one could ever quite fix, New Vegas could still do with a boost when it comes to NPC's causing lag though.

just wodering if this is compatible with simple borderless window.....

anyone running both?

I use Simple Borderless Windows and Skyboost r3 and my improvements on top of dragonreach stairs is:

On Windows 8 x64 (ati catalyst 11.11c modded with orca for compatibility on windows 8):

Vanilla: 24fps

Skyboost r3: 31-32 fps

BUT in Windows 7 x64 (ati catalyst 12.1) no improvements:

Vanilla: 19-20 fps

Skyboost r3: 19-20 fps

I have texture mods injected on original .bsa files with BSAopt (split in 2 textures .bsa: Skyrim - Textures1.bsa and Skyrim - Textures2.bsa) and iccc-enb mod 1.58d.

My rig is: Q9300 2,6Ghz @ 3,2Ghz, ATI HD5850 with ASUS bios unlock OC 880/1170, 4Gb ddr3 1600Mhz, PSU OCZ 650W and 1680x1050 screen resolution. AA without iccc-enb is x4 and with iccc-enb is x2, AF x16

DariusD
The point of sse3 and sse4 using is ambiguous , because :
- Timing is like on every other bottleneck , those people who has very fast cpus with sse4 support just don't need such code improvements
- SSSE3 is like 5% speedup from SSE2 on huge calculations , we don't have such here , 1% impact will do nothing
- In order to generate SSE3+ from high level code icc must be running on intel cpu of 6th series like c2d etc ( many cpuid checks inside of the compiler) and anyway because I'm using amd I wont be able to test something more than SSE3 (it has zero improvement comparing to SSE2) even if I'll find the way to force icc use SSE3+
Currently i have all dot code rewritten to high level because sse realization from Arisu on amd runs 5x times slower than pure x87 or sse1 generated by compiler

Xetrill, newt111
As i wrote before SSE3 has zero improvement comparing to SSE2 , SSSE3 has 5% gain top with huge calculations . Comparing intel to msvc compilation intel has 1 more fps on SSE2 SkyBoost version on my amd cpu . Currently the code path is one , size is huge cuz i need to check every byte to be from original game

Squall Leonhart
Not only dispatcher , compiler capabilities are limited while using it on amd

vejn
How much fps in skyrim do u have on vanilla game ?

So, what can we look forward to with r4? Assuming there will be an r4?

As soon as I will have +5 fps comparing to r3 that build will be named r4

5 more fps? daamn i dont know Skyrim has reach its limit on optimizations, but if you can make it happend we will be waiting after NY.

Just so I can properly be floored by the awesome, how much FPS does r3 give you over vanilla? Any ETA for this r4 release? Hours? Days? Weeks?

headbomb
The biggest boost i saw was only in few very rare cases in some places when vanilla was 30 fps r3 was 60 . No eta .

Wow amazing, with next version r4 and next version of ENB with optimizations for ATI users Skyrim will rocks! Now i have yet a few micro stuttering problems because enb is not good optimized yet. Thx for this incredible mod

Thank you for all your efforts. I am looking forward to the next version.

I know it is best if people test this in as many different locations as possible, but I still can't resist asking if there is any specific test location that you yourself would feel like recommending.