VOGONS


First post, by gustep

User metadata
Rank Newbie
Rank
Newbie

TL;DR: Can you get a better Tom2D score, and if yes, how did you do it?

Long version:

I have a Windows XP box set up - not connected to the network - for using antiquated CAD software. The CAD software can run under Windows 7 and later, however, it will have 4x - 10x slower screen redraw speeds, which makes it much less usable. GUI responsiveness and rendering speed matter a lot when you're actually trying to do something .

The reason is that the CAD software, like many other programs from the Windows XP and earlier times, uses the GDI interface for rendering the screen. GDI was hardware accelerated in many older graphics cards under Windows XP, where 2D performance mattered. Later on, the focus on 3D meant, unfortunately, that 2D performance got worse again.

You probably know all this already, from the various references to the Tom2D GDI benchmark on this forum.

So let me ask: What's the fastest (honest, no cheating) 2D GDI benchmark that you can achieve, and what are the settings to get it?

For me, I found the following the best so far:

- Windows XP SP3
- Cleartype=Off (or it will be a little slower)
- Nvidia Geforce GTX660
- i5-2300 CPU

With that setup, I get around 6787 marks in Tom2D. See images below for system specs and details.

Please post your specs and how-to if you get a significantly better score.

Thank you!

Attachments

  • XP-box-Tom2D.png
    Filename
    XP-box-Tom2D.png
    File size
    20.56 KiB
    Views
    834 views
    File comment
    6787 Tom2D score
    File license
    Fair use/fair dealing exception
  • XP-box-HWINFO.png
    Filename
    XP-box-HWINFO.png
    File size
    43.55 KiB
    Views
    834 views
    File comment
    Detailed system specs with HWINFO
    File license
    Fair use/fair dealing exception
  • XP-box-CPU.png
    Filename
    XP-box-CPU.png
    File size
    18.55 KiB
    Views
    834 views
    File comment
    CPU-Z system info
    File license
    Fair use/fair dealing exception
  • XP-box-GPU.png
    Filename
    XP-box-GPU.png
    File size
    28.21 KiB
    Views
    834 views
    File comment
    GPU-Z system info
    File license
    Fair use/fair dealing exception

Reply 1 of 12, by auron

User metadata
Rank Oldbie
Rank
Oldbie

that's an interesting topic, that tom2d benchmark seems almost forgotten these days. i don't know what constitutes cheating for you but i managed to beat your score with a p3 1100/gf4ti4200 setup under 98se:

tom2d.png
Filename
tom2d.png
File size
9.11 KiB
Views
782 views
File license
Public domain

on a more serious note, your upgrade options are pretty obvious - an overclocked K-cpu with ddr3-2133. out of curiosity, do you get a different score when you start the benchmark on just 1 thread? i believe that's what was said in that original benchmark article.

Reply 2 of 12, by cyclone3d

User metadata
Rank l33t++
Rank
l33t++

As far as XP, what about also overclocking the PCIe bus a bit?

Faster RAM and an overclocked CPU with more cache would probably help as well.

And wouldn't the 2D score be higher on say a 680 or even a 780Ti ? Or did they kill performance after the 6 series?

I don't have any GTX 6 or 7 series so I can't actually test it myself.

For all we know, as is evidenced by the other reply is that older cards may very well beat the 2D score of the newer cards such as yours.

Does this CAD software use any 3D acceleration at all for the actual rendering of the drawings or is it all CPU and the only thing that matters as far as graphics cards go is the GDI speed?

What exact CAD software is it?

Yamaha modified setupds and drivers
Yamaha XG repository
YMF7x4 Guide
Aopen AW744L II SB-LINK

Reply 3 of 12, by Standard Def Steve

User metadata
Rank Oldbie
Rank
Oldbie

Athlon 64 3800+ with a GeForce 7800GS scores 6530.
Slotting the above video card into an Athlon XP 2800+ nets 5082.
PIII-S (@ 1.63GHz) with a 6800GT gets a 3109.

My main setup is only scoring 3865. Spanked by an Athlon XP - wonderful.

94 MHz NEC VR4300 | SGI Reality CoPro | 8MB RDRAM | Each game gets its own SSD - nooice!

Reply 5 of 12, by pete8475

User metadata
Rank Oldbie
Rank
Oldbie

Interesting, just ran it on my XP machine here are the results and specs.

Tom 2d score - 12166

Specs:
Asus P8P67 LE
Intel Xeon E3-1245 V2
4GB DDR3-1333
GTX970 4GB

EDIT - I have done nothing special to get this score, I generally use this machine to play some older XP era games. Just recently upgraded the CPU to a Xeon and the graphics card to a GTX970 after I put a 980TI in my Win 10 computer. The 970 isn't "officially" supported in XP but with a simple driver inf edit it works perfectly.

Attachments

  • tom2d.PNG
    Filename
    tom2d.PNG
    File size
    18.79 KiB
    Views
    675 views
    File license
    Public domain
  • gpuz.PNG
    Filename
    gpuz.PNG
    File size
    34 KiB
    Views
    675 views
    File license
    Public domain
  • cpuz.PNG
    Filename
    cpuz.PNG
    File size
    22.02 KiB
    Views
    675 views
    File license
    Public domain

Reply 6 of 12, by gustep

User metadata
Rank Newbie
Rank
Newbie

Great, thank you all for the excellent replies!

It all is consistent with what I surmised: Get the fastest GPU possible for XP (GTX970 or GTX980), and the highest clocked CPU.

Pete did everything right, and the results are great.

I'll get a GTX970 / GTX980 / GTX780Ti when they are more affordable (if ever). In the meantime, I am upgrading from my i5-2300 to a Xeon E3-1245 and will report back with that soon.

The CAD program uses almost zero 3D, it is the PCB freebie program "PCB123" from Sunstone:

https://www.sunstone.com/pcb123-lp

The program is free and *easy to use* (much easier than Eagle or Kicad - just plop down pins and footprints and wire them up, if you want), but the output files are proprietary, so you can order your PCBs only from Sunstone, or pay them $100 to get Gerber files for your design. PCB123 was updated regularly from XP times, but they never updated their graphics engine, so it runs badly in Windows 7+. Just zoom in and out and watch the snap grid on the monitor: In Windows 7+, when you zoom out of the page (especially just before the snap disappears altogether), the snap grid will fill up the screen slowly from bottom to top. In Windows XP, it's instantaneous. It gets worse if you edit complex real circuit boards. If you like to cruise and zoom around a lot while you work - like when wiring up distant components - XP is the way to go. Everything newer, you have to accept the stuttering.

And finally, the highest Tom2D score that I could find was on the Russian site iXBT, from user iceb0n, showing a score of 16800. However, I can't reach him to ask what his configuration is, and I can't post on that forum either (2 week wait). See here:

https://forum.ixbt.com/topic.cgi?id=77:22344-3

It also seems that the Tom2D score from iceb0n, while higher overall in the final score, is a little less convincing than that of pete8475. Pete has higher marks in several individual sections. And thank you, pete, for explaining your config.

Here's the Tom2D result from iceb0n without config explanation:

Attachments

  • iXBT-Tom2D-iceb0n-GDI.png
    Filename
    iXBT-Tom2D-iceb0n-GDI.png
    File size
    12.98 KiB
    Views
    655 views
    File comment
    Tom2D result 16808 posted by iceb0n on iXBT but without sys config
    File license
    Fair use/fair dealing exception

Reply 7 of 12, by pete8475

User metadata
Rank Oldbie
Rank
Oldbie
gustep wrote on 2021-03-31, 23:24:

Pete did everything right, and the results are great.
I'll get a GTX970 / GTX980 / GTX780Ti when they are more affordable (if ever). In the meantime, I am upgrading from my i5-2300 to a Xeon E3-1245 and will report back with that soon.

Cool!

I've never seen this program before today, I still have my GTX770 2GB if you want me to test that in the same machine just let me know.

Also make sure the Xeon CPU is supported by your board, the list of supported Xeon models was very short on my Asus board.

Reply 8 of 12, by pete8475

User metadata
Rank Oldbie
Rank
Oldbie

So I've run this program a few more times and have changed NOTHING on my PC and now get much higher scores.

Second run - 14779
Third run - 16259
Fourth run - 14729

No idea why there is such variance. I have attached pics of the results.

Attachments

  • 16259.PNG
    Filename
    16259.PNG
    File size
    19.06 KiB
    Views
    622 views
    File license
    Public domain
  • 14779.PNG
    Filename
    14779.PNG
    File size
    18.78 KiB
    Views
    622 views
    File license
    Public domain
  • 14729.PNG
    Filename
    14729.PNG
    File size
    19.02 KiB
    Views
    622 views
    File license
    Public domain

Reply 9 of 12, by gustep

User metadata
Rank Newbie
Rank
Newbie

Hi Pete,

Yes, same here: When I re-run the benchmark, it scores higher on subsequent runs. I think it may have to do with SpeedStep and other dynamic CPU / GPU clock adjustments. The first run starts with idle clock rates, and takes a moment to spool up, but when the benchmark is run a second or a third time in short succession, the clock rate is at a maximum already.

Here are my results of re-running the benchmark several times in short succession.

I am still waiting for my Xeon to arrive, I think it should work in with my mainboard.

Attachments

  • XP-box-Tom2D.png
    Filename
    XP-box-Tom2D.png
    File size
    20.56 KiB
    Views
    569 views
    File comment
    Tom2D first run
    File license
    Public domain
  • XP-box-Tom2D-rerun1.png
    Filename
    XP-box-Tom2D-rerun1.png
    File size
    20.76 KiB
    Views
    569 views
    File comment
    Tom2D second run
    File license
    Public domain
  • XP-box-Tom2D-rerun2.png
    Filename
    XP-box-Tom2D-rerun2.png
    File size
    20.47 KiB
    Views
    569 views
    File comment
    Tom2D third run
    File license
    Public domain

Reply 10 of 12, by pete8475

User metadata
Rank Oldbie
Rank
Oldbie
gustep wrote on 2021-04-02, 06:24:
Hi Pete, […]
Show full quote

Hi Pete,

Yes, same here: When I re-run the benchmark, it scores higher on subsequent runs. I think it may have to do with SpeedStep and other dynamic CPU / GPU clock adjustments. The first run starts with idle clock rates, and takes a moment to spool up, but when the benchmark is run a second or a third time in short succession, the clock rate is at a maximum already.

Here are my results of re-running the benchmark several times in short succession.

I am still waiting for my Xeon to arrive, I think it should work in with my mainboard.

Ok cool, good luck with the upgrade!

Reply 11 of 12, by gustep

User metadata
Rank Newbie
Rank
Newbie

Hi Pete,

So if I get that right, you could test the GTX980Ti, GTX970 (your current results), and GTX770 in your XP system? That sure would be interesting!

Here's my update: I now upgraded from an i5-2300 to a Xeon E3-1225 CPU. Scores improved to ~9k but occasionally come in at ~7k. It can vary from run to run, quick repeat runs are no guarantee of a high score, it just varies.

The greatest variability in the Tom2D benchmark appears to come from the "Text" sub-score. Some sub-scores are also repeated exactly in some runs, so perhaps the benchmark isn't granular enough on such fast systems.

Anyways, here are the results:

Attachments

Reply 12 of 12, by pete8475

User metadata
Rank Oldbie
Rank
Oldbie

I just put the GTX770 into my XP machine, booting now.

Will post the benchmark results soon.

I'll do the 980ti tomorrow if I can find a few minutes.

EDIT - ran it 5 times on the GTX770. It's a 2GB EVGA branded card.

Attachments

  • 770-1.PNG
    Filename
    770-1.PNG
    File size
    18.94 KiB
    Views
    425 views
    File license
    Public domain
  • 770-2.PNG
    Filename
    770-2.PNG
    File size
    19.01 KiB
    Views
    425 views
    File license
    Public domain
  • 770-3.PNG
    Filename
    770-3.PNG
    File size
    19.01 KiB
    Views
    425 views
    File license
    Public domain
  • 770-4.PNG
    Filename
    770-4.PNG
    File size
    18.94 KiB
    Views
    425 views
    File license
    Public domain
  • 770-5.PNG
    Filename
    770-5.PNG
    File size
    18.95 KiB
    Views
    425 views
    File license
    Public domain