SW05 Benchmark is out

formatting link
Be warned, it's 137Mb.

Let the madness begin.

Reply to
Dale Dunn
Loading thread data ...

I tried to run it. Went to dinner,came back 45 minutes later and it would "loop" over and over. Had to use the Task Manager to shut it down. Best Regards, Devon T. Sowell

formatting link

Reply to
Devon T. Sowell

Hmm. I've run it sucessfully 3 times so far... I can't even guess why it would fail for you unless there is some system requirement not met.

Reply to
Dale Dunn

I see that the benchmark has more than one scoring method now. In addition to the time in seconds to complete the test, there is now a ratio that seems to compare the results to some base system. Has anyone been able to find out what that base system is?

And another thing: Am I alone in seeing pistons hanging out the bottom of the motorcycle engine?

Reply to
Dale Dunn

"Dale Dunn" a écrit dans le message de news:Xns95BA56BFDEC15daledunnatjamestoolc@65.24.7.50... . Has anyone been able to

From the SW2003 Benchmark page: The reference system for computing the normalized ratio is a 1.5GHZ Pentium4 with an Intel 850 chipset, 1GB PC800 ECC SDRAM, 40GB ATA/100 and NVIDIA Quadro2 Pro graphics.

From the SW2005 Benchmark page: The reference system for computing the normalized ratio is an Intel Xeon

2.4GHz, 2GB PC800 ECC RDRAM, 40GB ATA/100 HDD and NVIDIA Quadro4 700XGL graphics running Windows XP SP1.
Reply to
Jean Marc BRUN

Ah, thank you. I must have glossed over it.

What a lame system. Now I don't feel so great about having a score over 1. I suppose that is probably a fair average representaion of the age of machine most of us have. Anyhow, as arbitrary standards go, It'll serve as well as any other.

I wonder how it is that my QuadroFX 1100 with 128MB and AGP8x scored less than 1 against a Quadro 4 with 64MB and AGP4x. Something's not right here...

Reply to
Dale Dunn

After crashing it myself, I hope you (and everyone else) had used the copy settings wizard to make a backup of your settings. Unless you like SW to be bogged down as much as possible with shadows, and anti-aliasing, etc.

Reply to
Dale Dunn

I haven't done anything with the SW05 Benchmark, but on earlier versions the CPU/Graphics/IO splits are not very helpful. Changing the CPU will affect all three, changing the graphics will affect all three, changing the disk or memory will affect all three.

Jerry Steiger Tripod Data Systems "take the garbage out, dear"

Reply to
Jerry Steiger

I did find that you actually have to set your discplay resolution to 1280 x

1024 as the instructions state. The benchmark resizes the window, but it seems to be a percentage of the display resolution. Kinda dorky.

After doing that, the video score did go over one, but still not by as much as I expected. Something is still screwy. The video score seems to be limited by something else.

Reply to
Dale Dunn

Realview will also effect video score. It is not one of the settings that the benchmark controls and you do not know if it was turned on on their base machine. There maybe other settings like this.

I also noticed some rebuild error and not all the files were converted to 2005 format when I manually opened the files. I corrected these problems and defragged afterwords and my scores were better.

All of my numbers were above 1 except for the small case score and it is consistantly a one or below. Test machine was AMD 64 bit 3400+ with

2 GB PC3200 ram, FX1100 video card, and SATA harddrive. The only thing that I can think of causing mine to be slower than theirs is dual processor.

Wayne

Reply to
Wayne

I don't know if RealView gets switched on, but I did notice a lot of parts with textures applied. I always run with RealView turned off, so if the benchmark didn't turn it on, it wasn't on. I wonder how many people spend their days with RealView, shadows, and anti-aliasing turned on. Benchmarking in this state does not reflect a typical day for me. Maybe most people leave this stuff turned on. I dunno.

I noticed all of those errors too. For example, the pistons hanging out the bottom of the Suzuki engine. It looked like they didn't try very hard to clean it up for the new benchmark.

Your machine is almost identical to mine. I have half your RAM and twice the HD (2x RAID 0). Here are my scores (starting with the highest):

Resolution set to 1280, all extra processes shut down, and o/c to 2310MHz (105%):

sw2005.Result.score = 1.33 sw2005.Result.graphics = 1.09 sw2005.Result.cpu = 1.62 sw2005.Result.io = 1.59

Same as above with screen resolution set to 2048 x 1536 (normal for me):

sw2005.Result.score = 1.29 sw2005.Result.graphics = 1 sw2005.Result.cpu = 1.69 sw2005.Result.io = 1.56

Same as above not overclocked (2200MHz):

sw2005.Result.score = 1.23 sw2005.Result.graphics = 0.98 sw2005.Result.cpu = 1.6 sw2005.Result.io = 1.39

Same as above, after reboot (all normal processes loaded):

sw2005.Result.score = 1.22 sw2005.Result.graphics = 0.97 sw2005.Result.cpu = 1.58 sw2005.Result.io = 1.34

Same as above, not rebooted (after working all day, downloading and installing the benchmark):

sw2005.Result.score = 1.13 sw2005.Result.graphics = 0.92 sw2005.Result.cpu = 1.42 sw2005.Result.io = 1.24

I'm wondering if your graphics scores are significantly differnt from mine. I expected a much higher number than the Quadro4 could deliver. I'm running 53.03 drivers. I wish I had had the brains to set the screen resolution lower at the beginning of all this. That would have been more useful.

Reply to
Dale Dunn

Well, duh, do it again... :-)

WT

Reply to
Wayne Tiffany

Um, ok. I was planning to, actually, but I got distracted. I almost had a life for a few days there.

Reply to
Dale Dunn

Here are my results with the system as is during a typical day. It is probably slower than what I could get if I used faster RAM and the FX4000 graphics card.

sw2005.Result.score = 1.67 sw2005.Result.graphics = 1.55 sw2005.Result.cpu = 1.76 sw2005.Result.io = 1.93 AMD64 FX53 NVidia FX3000

2GB PC2600 RAM 1600x1200 resolution
Reply to
P.

Ok, I've been benchmarking for a while now. Time to post some results and chew on them a bit. Here's my system essentials:

MSI K8T8 FIS2R Neo (on-demand overclock disabled, Cool 'n Quiet enabled) Athlon64 3400+

2 Mushkin 512 PC3700 QuadroFX 1100 2x 7200rpm Seagate Barracuda 120MB, RAID 0 SolidWorks 05 Sp0.1 on all tests XP Pro Sp1 XP 64 Preview, build 1218

The evening after downloading the benchmark, I dove right in:

After working all day, installed and ran the benchmark:

sw2005.Result.score = 1.13 sw2005.Result.graphics = 0.92 sw2005.Result.cpu = 1.42 sw2005.Result.io = 1.24

After reboot:

sw2005.Result.score = 1.22 sw2005.Result.graphics = 0.97 sw2005.Result.cpu = 1.58 sw2005.Result.io = 1.34

Quite a jump. Must remember to reboot before every run...

Rebooted, stripped all extra processes, including AV and firewall (network disabled):

sw2005.Result.score = 1.23 sw2005.Result.graphics = 0.98 sw2005.Result.cpu = 1.6 sw2005.Result.io = 1.39

Is it really worth the bother to keep the system free of all those neat little utilities that make Windows tolerable? Yeah, I know some of them can cause crashes and other trouble. I say an occasional hiccup is tolerable in comparison to going without WinAmp, Google Desktop, GetRight, Trillian, StrokeIt, SpamPal, Handy Thing, PTFB, CapsWiz, Restart, Daemon Tools, Gmail Notifier, MouseImp and, of course, Anti- Virus.

Stripped extra processes, overclocked to 2310 (from 2200):

sw2005.Result.score = 1.29 sw2005.Result.graphics = 1 sw2005.Result.cpu = 1.69 sw2005.Result.io = 1.56

Quite a jump in IO performance there. Interesting that the increase in IO performance offsets the lesser increase in video performance to keep the overall perfomance increase almost linear WRT clock speed.

Stripped extra processes, overclocked to 2310, set display resolution to

1280 x 1024 like the instructions said, duh! (All previous results were run at my normal working resolution of 2048 x 1536. Why the benchmark doesn't resize the SW window to a specific size I don't know.):

sw2005.Result.score = 1.33 sw2005.Result.graphics = 1.09 sw2005.Result.cpu = 1.62 sw2005.Result.io = 1.59

We have to read those instructions, I suppose. Video and CPU both had a pretty good jump. I wonder what it is about a larger window that makes CPU tasks take longer. Unless of course, the CPU test is influenced significantly by video performance.

Normal clock, updated chipset and video drivers(VIA Hyperion 4.55 from I don't remember what, NVidia 67.22 from 53.03):

sw2005.Result.score = 1.38 sw2005.Result.graphics = 1.21 sw2005.Result.cpu = 1.6 sw2005.Result.io = 1.46

I should have updated these one at a time. What a difference 14 versions makes in video performance, even returning to normal clock speeds. I wonder why SW hasn't certified a new driver yet. IO performance is back to normal.

updated drivers, NView CAD profile loaded with SolidWorks settings:

sw2005.Result.score = 1.41 sw2005.Result.graphics = 1.26 sw2005.Result.cpu = 1.6 sw2005.Result.io = 1.46

Not an insignificant change.

As above, with themes service stopped (was not using it anyway):

sw2005.Result.score = 1.4 sw2005.Result.graphics = 1.28 sw2005.Result.cpu = 1.56 sw2005.Result.io = 1.4

This is so wierd I ran it twice. I can imagine why the video would improve, but why would eveything else get worse?

As above, a second install of XP Pro with very little usage or extra tools (used during 05 beta):

sw2005.Result.score = 1.42 sw2005.Result.graphics = 1.26 sw2005.Result.cpu = 1.63 sw2005.Result.io = 1.5

A very similar overall score, which is an encouraging sign for the state of the installation I work on. I'd like to figure out how to get the cpu score up that extra bit on the main installation.

As above, sound card drivers uninstalled:

sw2005.Result.score = 1.46 sw2005.Result.graphics = 1.33 sw2005.Result.cpu = 1.62 sw2005.Result.io = 1.49

I found it interesing that sound card drivers hold the system back a bit, even though the sound card was not in use during any tests. I should probably run through the BIOS and disable all the serial ports or whatever else I'm not using or going to use.

XP64 Preview Build 1218, latest beta chipset and video drivers (with CAD and SW settings), no audio drivers found (yet):

sw2005.Result.score = 1.46 sw2005.Result.graphics = 1.36 sw2005.Result.cpu = 1.57 sw2005.Result.io = 1.57

How very interesting. The 64 bit device drivers offset the loss in CPU performance. The IO in particular got a boost, but so did the graphics. By the way, XP 64 looks pretty polished now. I hear speculation that it will finally be released the first half of this year. Hopefully they're getting drivers written as fast as their little fingers can fly.

As above, with themes disabled (previous results in XP64 had the default XP theme):

sw2005.Result.score = 1.49 sw2005.Result.graphics = 1.38 sw2005.Result.cpu = 1.64 sw2005.Result.io = 1.52

Ok, I'd have to overclock 32-bit XP to get these results. This is looking promising for XP64. If SW was recompiled for 64-bit, I wonder if it would see a simliar boost to what the various drivers got. I don't know enough about the architecture to guess what benefit Sw might receive, and the previous sentence may prove it. It looks to me like XP 64 might be the OS of choice for those of us who need or want those last few percentange points of performance.

Follow the link and search for "Ship in a XP64" for the previous thread where I said, "There certainly won't be any reason to take SW to a 64-bit OS until SW is compiled to 64 bits."

formatting link

Ok, comparisons to other hardware: Of the 5 systems reporting results on the SPECapc site, Only one of the P4 systems can outrun me, and that's due to the Quadro 3400 video. These systems were tested late in 04, and can just keep pace with the AMD system I built last March.

formatting link
The two Opteron 250 systems (same clock speed as my system) absolutely spanked me, with composite scores of 1.91 and 2, and CPU scores of 1.78 and 1.88. The slower Opteron ran ECC memory, which could account for the lower scores. I'm trying to remember what architectural advantage the Opteron has over Athlon64. IIRC the Opteron and AthlonFX have dual channel memory, where the Athlon64 does not. Whatever the case, if I had had this information back in March, I might have built an Opteron system instead. Is Fujitsu Siemens available in the states?

Time for bed.

Reply to
Dale Dunn

Nice... Apparently the dual channel memory controller and maybe a higher clock speed are enough to keep you ahead of my Athlon64 with PC3200. What is the clock speed on the FX53?

What kind of storage system do you have? It's 30% faster than my pair of Barracudas in a striped array.

Reply to
Dale Dunn

I think the AMD64 FX53 runs at 2.4 GHz and O'Clocking doesn't work on this MOBO but would on the sister board ASUS makes.

It was faster when first built and got ship in a bottle down to 18 seconds with the appropriate settings. If I set it to 1280x1024 it would be faster by a bit too.

A single WD SATA drive at 72GB. It is still supposed to be one of the fastest around. But then this benchmark only measures the disk drive speed on the first iteration as after that everything is cached in memory. I'll have a look at the difference:

Test Number 1 Test Total = 460.28 Graphics = 112 CPU = 183.84 I/O = 164.44

Test Number 2 Test Total = 441.42 Graphics = 111.27 CPU = 180.86 I/O = 149.29

Test Number 3 Test Total = 440.4 Graphics = 114.61 CPU = 179.79 I/O = 146

So I/O dropped by about 15 seconds between the first and subsequent iterations. That is the hard drive. After that I/O is almost all memory. And I don't have the fastest memory in this thing either.

Reply to
P.

I agree that the IO test is a little weak. I wonder how much of your IO score has to do with the CPU's dual channel memory. Some of it is probably influenced by the amount of memory you have too. With more memory, Windows can allocate more memory to the disk cache. Those WD Raptor drives are fast at 10k rpm, but it's surprising that it scores faster than a pair of 7200 rpm drives.

If you ever get a chance to upgrade your memory, I'd love to see the effect it has on IO scores. I suppose I could underclock my memory to see what happens. I'll try to remember to do that tonight.

Reply to
Dale Dunn

The reason I got the Raptors was that there was a general consensus at Anandtech and Toms and a few other places that they were possibly faster than SCSI.

What really shocks me about this setup is that it is six months since it was built and it is still quite competetive. And if 64 bit SW comes out it will remain cost effective for another year. I can't remember when a system would remain near the top of the heap this long.

Dell was pretty smart sticking with Intel where they are still fighting with themselves to get more speed.

Reply to
P.

I remember reading that. My thinking when I built my system was that the speed increase was not enough to justify the cost and lack of capacity. IIRC the 72GB drives were announced, but not yet available.

You say "raptors" in the plural... You have more than one in your system then? In an array?

Reply to
Dale Dunn

PolyTech Forum website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.