appiah4 wrote on 2022-09-02, 06:24:

The consensus at the time was that the 440BX and later 815E(P) chipsets had better memory throughput and compatibility, which was to be fair correct at the time. However, the VIA 4-in-1 drivers came a LONG way since their inceptions and with mature drivers the VIA chipsets leave little to be desired in my experience. Moreover, they bring things like Tualatin Support + ISA which is basically impossible in Intel land. Hence, I love me some VIA, but you will see that most cookie cutter builds (including my own period correct build for 199-2001) go with 440BX/815E(P) instead.

Also, a lot of the problems come from user error in driver installs.

As mentioned earlier, you need to install 4-in-1/Hyperion before installing AGP card drivers, as the miniport need initialising first. Do it the other way round and you're in for trouble. This is completely logical if you consider the underlying architecture and applies to any chipset with AGP of course. Fact that people had these issues more with Via than Intel was political: Microsoft included Intel chipset drivers in its OSs, but no Via drivers. Defensible in Win98FE (as at that time Via drivers were an awful mess), but by 2001, XP and pretty solid Via drivers it had little to do with the technology and everything with US corporate politics.

bloodem wrote on 2022-09-03, 07:20:

As suspected, the game does not run well at all (that is NOT how it's supposed to run - it's running in slow motion). If you will play further, you will notice there are sections (with multiple moving entities on the screen) where it will even stutter/flicker.
<snip>
Also, not sure if you have the speakers turned on, but I don't hear any music.

Thanks sir... Indeed! 😀 You are correct. Sorry for doubting you.

Oh, the reason I had no sound hooked up was because my physical setup is only just nearing completion with KVMs, new power strips, bunches of cables and whatnot, and I didn't have time to hook up my 4 way speaker selector yet (or my speakers, for that matter).

Yes, you're right, I could not find a single combination of ODCM or ACM in which Titus both played with the correct speed and with proper FM music. When the FM was ok, the game was too slow, when the game was fast enough, it was missing notes.

Were you able to find a good throttling balance for Titus with your VIA setup, and if yes, would you please articulate further?

Good thing my fourth computer in this setup is in fact a 486 for scenarios just like this one.

EDIT: I also want to report that upon further testing, you were right about stuttering with ODCM... Jazz Jackrabbit won't run unless I set ODCM to 1, but it does in fact stutter at certain parts. On the bright side, YMF724 sound does in fact work properly with it (SB Pro selected in setup) with SB-Link, but that's not important anymore, since I just ordered an Ezra-T and will probably be replacing this with a P3B-F and the Ezra-T and an ALS100 sound card. Thanks!

leonardo wrote on 2022-09-01, 07:41:

pixel_workbench wrote on 2022-09-01, 05:40:

Because VIA gives you options and Intel takes them away. […]
Show full quote

Because VIA gives you options and Intel takes them away.

For mainstream Win 98 builds with a P2 or P3, the 440bx is pretty hard to beat. But then Intel got greedy and made some annoying restrictions. The i815 only makes sense if you want a Tualatin Win 98 rig and to run 133mhz FSB with in-spec AGP. But then you lose ISA, so you might as well skip the P3 and move to P4 or Athlon XP, or even Athlon 64. I would argue the i815 has a very narrow niche of applications.

Want to push Tualatin to the max and see how it handles Win XP games? You're gonna need a VIA board with at least 1GB RAM.

Want to maintain good DOS compatibility with that Tualatin? You'll also need a VIA board with ISA.

Want to have it all and use a good Intel chipset? Get the modded Tualeron on ebay and put it on a 440bx board.

Tualerons and 440bx too expensive these days? Get a cheap Coppermine and a VIA Apollo Pro 133A board with an ISA slot. Sure, it's not as fast clock for clock, but you get 133mhz fsb with the right AGP divider, and ISA slot. It might also be less compatible with some primitive early AGP cards, but a Geforce or Radeon works just fine. You might not get AGP 4x stable, but the 440bx doesn't even have that option.

If you're building a P3 rig today, I'd say there are even less reasons to use the i815 than before.

Great answer!

I agree, the only possible exception to this would be when sticking to Win9x and finding and 815E based board (industrial) with properly implemented bridged ISA . They do exist, I have one, but they are now hard to find and expensive.

Well built Via 694(T) based socket 370 boards can be great, as long as one is careful to work around any AGP compatibility issues and 686B southbridge mood swings (i.e. with the SB Live!). Also, if planning to use a SATA HDD or SSD on a 686B IDE port, only Marvell chip based IDE to SATA converters will allow reliable DMA operation, AFAIK (PIIX4E and ICH2 are easier to work with in my experience).

mockingbird wrote on 2022-09-04, 03:16:

Were you able to find a good throttling balance for Titus with your VIA setup, and if yes, would you please articulate further?

Yes, with VIA (depending on the platform that we're talking about) there is usually one combination that works.

My favorite is the Asus A7V600-X (VIA KT600) + Athlon XP Thoroughbred + GeForce 4 Ti 4200/4400/4600 + Yamaha YMF7x4. Actually, with such a combo, you will rarely need ACPI throttling (if at all!).
That's because:
1. The CPU multiplier will be unlocked, so you can achieve speeds between 500 MHz (FSB100 x 5) and 2+ GHz (depending on what CPU you are using; it doesn't need to be the fastest).
2. The motherboard BIOS allows you to toggle both the L1 and L2 cache (IIRC, I think setmul only works for L1, not for L2 on these chips).

For example, if you disable the L1 cache when the CPU is running at its max frequency, you will get ~ 486DX2-66 equivalent speed. Then, if you gradually decrease the CPU multiplier and FSB, you will end up with any speed between ~486DX2-66 and ~386DX-33.
If you want any other speed interval (lower than 386DX-33 or higher than 486DX2-66 - like a Pentium equivalent speed), then ACPI Throttling will easily fill that void (and it will do so without introducing stutters up to levels 9 or 10).
As mentioned earlier, the only (minor) issue with such a build is the fact that the Yamaha YMF7x4 will need to rely on the DSDMA TSR (which in turn needs an expanded memory manager). This will break compatibility with a few games (like Ultima 7 / Jazz Jackrabbit), but 98% or more will work perfectly fine. So, if you're simply interested in playing games with great OPL3 sound, such a build can be amazing and you can forget about needing a dedicated 486 machine - which, contrary to popular belief, will usually be a LOT less flexible in DOS and a LOT more finicky/unreliable (on the other hand, if you simply like tinkering with old hardware and you have a problem-solving mindset, then a 486 is definitely for you). 😀

And, let's not forget about how powerful this build is for Win98 - allowing you to basically play any game up to 2001/2002. Throw in a Voodoo 2 as well, and you've now got one PC to rule them all. 😀 It won't be absolutely perfect, but I would argue that it's more than what most people actually need.

mockingbird wrote on 2022-09-04, 03:16:

EDIT: I also want to report that upon further testing, you were right about stuttering with ODCM... Jazz Jackrabbit won't run unless I set ODCM to 1, but it does in fact stutter at certain parts. On the bright side, YMF724 sound does in fact work properly with it (SB Pro selected in setup) with SB-Link, but that's not important anymore, since I just ordered an Ezra-T and will probably be replacing this with a P3B-F and the Ezra-T and an ALS100 sound card. Thanks!

If you're going for the Ezra-T, I highly recommend to pair it with a Gigabyte GA-6BXC rev 2.0 (or 6BXC rev 1.9 / AMPTRON PII-3100B / Kaimei 650B-ATX - if you are willing to desolder the VRM controller chip and replace it with a HIP6004BCB). These are the only boards that I know of (with the ICS 9148BF-26 clock generator), allowing you to control the FSB frequency from within DOS/Win98. There are surely other motherboards as well, but they remain unknown for now. 😀

bloodem wrote on 2022-09-04, 05:40:

...<snip>
If you're going for the Ezra-T, I highly recommend to pair it with a Gigabyte GA-6BXC rev 2.0 (or 6BXC rev 1.9 / AMPTRON PII-3100B / Kaimei 650B-ATX - if you are willing to desolder the VRM controller chip and replace it with a HIP6004BCB). These are the only boards that I know of (with the ICS 9148BF-26 clock generator), allowing you to control the FSB frequency from within DOS/Win98. There are surely other motherboards as well, but they remain unknown for now. 😀

Excellent, thanks...

So the two platforms I'll be testing are the KT133A + unlocked Barton and Ezra-T. The P3B-F has an ICS 9250CF-08, I don't see that listed with SetFSB (is this the utility you're referencing, and if so, is there a DOS equivalent?) Ok, I may have to look for one of those moels you mentioned... I've already done the HIP6004BCB mod on my P2B-S, so I know where to get genuine ones -- thanks for the tip.

By the way, any reason why you stop at the KT600? There are later chipsets like the K8T890 for example which I believe are compatible with Win98.

EDIT: Ah, ok, I see, you're using RayeR's SMB.ZIP. Ok, not a big deal... There are a couple of Windows programs that can modify the FSB that work with the P3B-F... I'm just experimenting with it -- I still have my trusty 486 machines for hassle-free operation. Man, have you seen the prices on those GA-6BXC boards? No way... That plan's out, too expensive.

mockingbird wrote on 2022-09-04, 06:37:

So the two platforms I'll be testing are the KT133A + unlocked Barton and Ezra-T. The P3B-F has an ICS 9250CF-08, I don't see that listed with SetFSB (is this the utility you're referencing, and if so, is there a DOS equivalent?) Ok, I may have to look for one of those moels you mentioned... I've already done the HIP6004BCB mod on my P2B-S, so I know where to get genuine ones -- thanks for the tip.

You're welcome. 😀
I don't think SetFSB works from DOS. I was referring to RayeR's SoftFSB utility (which works great from both DOS and the Win98 command prompt). This means that you can also add it to autoexec.bat, to set any FSB speed you want during startup.

mockingbird wrote on 2022-09-04, 06:37:

By the way, any reason why you stop at the KT600? There are later chipsets like the K8T890 for example which I believe are compatible with Win98.

The VIA K8T890 is a PCI Express chipset. While it is possible to make PCI Express video cards work with Win98, at that point might as well go with something a lot more powerful (Core 2 Duo/Core i3/i5/i7 era).

However, yes, there are very good Athlon 64 chipsets like the K8T800 (although, you might run into issues with certain motherboards/BIOS versions), and while they can be pretty flexible in DOS, they are not as flexible as a Thoroughbred + KT600 board (*not all KT600 boards are created equal, though, which is why I specifically recommend the Asus A7V600-X).
The KT880 + Thoroughbred (or unlocked Barton) is also fine, but since this chipset is newer, certain BIOS options might be missing (like disabling the L1 / L2 cache - which will give you a bit less flexibility, although, truthfully, it will not make an actual difference in the grand scheme of things, it's still a very flexible platform).

mockingbird wrote on 2022-09-04, 06:37:

EDIT: Ah, ok, I see, you're using RayeR's SMB.ZIP. Ok, not a big deal... There are a couple of Windows programs that can modify the FSB that work with the P3B-F... I'm just experimenting with it -- I still have my trusty 486 machines for hassle-free operation. Man, have you seen the prices on those GA-6BXC boards? No way... That plan's out, too expensive.

Don't look at the crazy sale prices on eBay, look at the sold items prices (they frequently appear for a lot less).

bloodem wrote on 2022-09-04, 07:04:

I don't think SetFSB works from DOS. I was referring to RayeR's SoftFSB utility (which works great from both DOS and the Win98 command prompt). This means that you can also add it to autoexec.bat, to set any FSB speed you want during startup.

Heh, I beat your post with my edit by about a minute. Here are the PLLs he lists as supported:

"Winbond 83194R-02/04/39/39A, 83195R-08, ICS 9148-26 and CY 28349"

However, yes, there are very good Athlon 64 chipsets like the K8T800 (although, you might run into issues with certain motherboards/BIOS versions)

Fascinating... You have certainly gone down this road before, while I am only beginning my journey 😀 Yes, indeed I'm planning an "ultimate" win98 build with such a board (though not quite that model) -- I will have to test whether it exhibits that issue. Thanks again!

mockingbird wrote on 2022-09-04, 07:18:

Fascinating... You have certainly gone down this road before, while I am only beginning my journey 😀 Yes, indeed I'm planning an "ultimate" win98 build with such a board (though not quite that model) -- I will have to test whether it exhibits that issue. Thanks again!

Yeah, I have been doing this for quite a bit of time... (cue the "it's been 84 years" gif) 😅

mockingbird wrote on 2022-09-04, 06:37:

Excellent, thanks... […]
Show full quote

bloodem wrote on 2022-09-04, 05:40:

...<snip>
If you're going for the Ezra-T, I highly recommend to pair it with a Gigabyte GA-6BXC rev 2.0 (or 6BXC rev 1.9 / AMPTRON PII-3100B / Kaimei 650B-ATX - if you are willing to desolder the VRM controller chip and replace it with a HIP6004BCB). These are the only boards that I know of (with the ICS 9148BF-26 clock generator), allowing you to control the FSB frequency from within DOS/Win98. There are surely other motherboards as well, but they remain unknown for now. 😀

Excellent, thanks...

So the two platforms I'll be testing are the KT133A + unlocked Barton and Ezra-T. The P3B-F has an ICS 9250CF-08, I don't see that listed with SetFSB (is this the utility you're referencing, and if so, is there a DOS equivalent?) Ok, I may have to look for one of those moels you mentioned... I've already done the HIP6004BCB mod on my P2B-S, so I know where to get genuine ones -- thanks for the tip.

By the way, any reason why you stop at the KT600? There are later chipsets like the K8T890 for example which I believe are compatible with Win98.

EDIT: Ah, ok, I see, you're using RayeR's SMB.ZIP. Ok, not a big deal... There are a couple of Windows programs that can modify the FSB that work with the P3B-F... I'm just experimenting with it -- I still have my trusty 486 machines for hassle-free operation. Man, have you seen the prices on those GA-6BXC boards? No way... That plan's out, too expensive.

fwiw my K8T890 did not like 98SE in the slightest. ME works fine, but that'll shave some compatibility off for DOS games.

mockingbird wrote on 2022-09-04, 03:16:

Yes, you're right, I could not find a single combination of ODCM or ACM in which Titus both played with the correct speed and with proper FM music. When the FM was ok, the game was too slow, when the game was fast enough, it was missing notes.

looks like one of those games coded by incompetent programmers believing other incompetent people (Adlib company had nobody with a clue on board) Re: Speed Affected Sound Blasters
this game needs a patch instead of going out of your way to get slow enough hardware

https://bochs.sourceforge.io/techspec/adlib_sb.txt wrote:

According to the AdLib manual, the 'official' method of checking for a | sound card is as follows: | | 1) Reset both […]
Show full quote

According to the AdLib manual, the 'official' method of checking for a
| sound card is as follows:
|
| 1) Reset both timers by writing 60h to register 4.
| 2) Enable the interrupts by writing 80h to register 4. NOTE: this
| must be a separate step from number 1.
| 3) Read the status register (port 388h). Store the result

so right there we have a situation where official specification asks you to do two consecutive writes to same data register. This will even fail on OPL3 with fast enough CPU/ISA, not to mention OPL2 requires "23 us delay after data register write"

rasz_pl wrote on 2022-09-04, 17:33:

looks like one of those games coded by incompetent programmers believing other incompetent people (Adlib company had nobody with a clue on board) Re: Speed Affected Sound Blasters
this game needs a patch instead of going out of your way to get slow enough hardware

Jazz Jackrabbit was problematic as well... It only worked on a single combination of ODCM and ACM, and it did not work 100% properly. Yea I'm sure a patch exists for JJ, but that's a sisyphean task, looking for patches.

Besides, I am also now having a second look at my KT7A+Barton and it's looking good. It's set to 16x133, with VIAFSB I can bring it down to 100, and then setmul takes it down to 300.

Then I tried THROTTLE.EXE without success... All settings just basically took it to a very slow 386 -- too slow. So I tried cpuspd ACM ('t' switch), and contrary to what I said earlier, it did in fact complain about ACPI being disabled (it didn't with the P4). Ok, I can live with ACPI, no big deal, as long as I get to assign my sound card IRQ7 in the BIOS. But cpuspd was much nicer and more versatile, and did in fact scale properly (offered 16 choices)... I've got a few errands to finish, but I plan on testing Titus on it and reporting back here.

I'm wondering about contacting RayeR to add support for more PLLs. VIAFSB author seemed to only require the PLL's datasheet for this. Ezra-T is nice, but the motherboards with the limited dos utility support are rare and expensive, and the limitation is artificial in the sense that I'm almost certain someone's going to take the initiative and release a more comprehensive utility for it in the future that supports more boards (I've since found more board models with some of the PLLs he supports... But it's a bit of an investment -- I already have plenty of Socket A boards).

mockingbird wrote on 2022-09-04, 18:28:

Then I tried THROTTLE.EXE without success... All settings just basically took it to a very slow 386 -- too slow.

Did you use the -C switch with throttle? Without that switch it disables L1 cache which would explain why it brought the system that low.

Sombrero wrote on 2022-09-04, 18:41:

Did you use the -C switch with throttle? Without that switch it disables L1 cache which would explain why it brought the system that low.

Ah, indeed. What a dummy. I keep forgetting that. I'll test again and report back.

throttle.exe seems to do the same thing as cpuspd, but throttle.exe does not need ACPI to be enabled in the BIOS (while it doesn't complain like cpuspd, it doesn't actually do anything with it disabled.... Again this still holds true for the P4 though -- it did not need to be enabled to work)

@bloodem:

I tested Titus on my Barton/KT133A and multiplier+acpi throttling, and I was not satisfied with the result.

With a multiplier of 3 and throttle set pretty low, I can get to a 486 DX2/50 (according to Speedsys), but the game runs too slow. At a higher throttle, the game runs ok but it's got the aforementioned sound oddities. Another thing: The title screen fade sequence is not smooth. It pauses througout very briefly, though that is an exaggerated way of explianing it, for lack of better words.

You say you managed to get Titus running perfectly with your Ezra-T?

Thanks

EDIT: I just tested Jazz and it had the same problem as Titus with the menu transition animations being jittery.

Hmmm.... Quite odd.

Maybe K6-2 or Ezra-T are in fact the way to go.

mockingbird wrote on 2022-09-04, 20:16:

With a multiplier of 3 and throttle set pretty low, I can get to a 486 DX2/50 (according to Speedsys), but the game runs too slow.

With an AthlonXP, you don't need to get the multiplier all the way down to effectively use Throttle. Try it at 500 MHz (or higher) and then use different Throttle values until you get the desired speed.

You can use the NSSI CPU benchmark to get a rough estimate of which CPU your current speed matches.

mockingbird wrote on 2022-09-04, 18:28:

sisyphean task ... followed by describing hours of fiddling with hardware

I rather whip out debugger and add few noops here and there to fix the game

rasz_pl wrote on 2022-09-04, 21:10:

sisyphean task ... followed by describing hours of fiddling with hardware/

I rather whip out debugger and add few noops here and there to fix the game

Heh 😉

Joseph_Joestar wrote on 2022-09-04, 20:52:

With an AthlonXP, you don't need to get the multiplier all the way down to effectively use Throttle. Try it at 500 MHz (or higher) and then use different Throttle values until you get the desired speed.

Yes, I thought just so, but good point. I'll perform some further tests.

Joseph_Joestar wrote on 2022-09-04, 20:52:

With an AthlonXP, you don't need to get the multiplier all the way down to effectively use Throttle. Try it at 500 MHz (or higher) and then use different Throttle values until you get the desired speed.

Nada, did not help.

The problem with Jazz Jackrabbit is that other than when employing one specific setting that works with the game, it will otherwise throw this error:

Runtime error 200 at 0009:37F2

And when it does work, there is a jitteriness... The best way to see it manifesting is when exiting the game and there's a closing film lens-type animation at the shareware outro screen when you're sent to the DOS prompt.

The P4 didn't have this issue... It wasn't perfect either though.

Next I'm going to hook up my Cyrix 5x86 system and see what's going on... See, people? This is why you sometimes need actual period-correct hardware.

EDIT: Ok, solved the issue... Swapped the Cirrus Logic PCI videocard with a GeForce 2 GTS and the problem disappeared. All that's needed for Jazz is SETMUL L1D, no need to play with multipliers or throttling. I read @bloodem's earlier post and saw he mentioned that so I gave it a try and turns out he was right.

EDIT2: Titus also surprisingly runs perfectly fine (at least at the level start) with just SETMUL L1D. CPU spec is Barton running at 16x100. Sound is perfect (ALS100 ISA), speed is correct. -update: For Jazz, I set the FSB in the BIOS back to 133 (133x16 = 2.13Ghz), ran SETMUL 3 which brought it down to 400Mhz, and then ran cpuspd t6. Works great.

EDIT3: Titus works perfect on the aformentioned P4 system (at least at the level start) with L1 and L2 disabled (in the BIOS because I did not have the utility handy on this machine). Jazz as well. Jazz won't run at all without L1 and L2 disabled (without throttling), Titus does not run well with only L1 disabled. -update: YMF724 doesn't like such a slow computer, it seems... The right channel keeps cutting in and out with Jazz and Titus. No issues whatsoever with the Athlon XP system though, except maybe slightly jerky scrolling with Jazz, but maybe that's related to the GeForce2 and VESA -update: No! That's related to ACPI throttling... With only cache disabled and CPU running full speed, the scrolling is smooth. An easy way to tell is to look at the sun right when you start the first level and run to the right after killing the turtle. I wonder how many other games don't like ACPI throttling.

mockingbird wrote on 2022-09-04, 21:37:

I wonder how many other games don't like ACPI throttling.

The same question popped to my mind earlier last week on another topic, I had been under the impression ACPI throttling is nearly always problem free. I've started testing my rather small selection of games and I've already found that Doom and Doom 2 doesn't like ACPI throttling, both games seemed to work just fine at first glance but then I noticed they both have these odd little hitches when throttled that get more pronouced the more you throttle.

I'm starting to believe bloodem is correct and it's more common than I thought.

rasz_pl wrote on 2022-09-04, 21:10:

I rather whip out debugger and add few noops here and there to fix the game

Much easier to get a flexible system, than to try and fix all DOS games that have such quirks (and there are many!). 😁

mockingbird wrote on 2022-09-04, 20:16:

With a multiplier of 3 and throttle set pretty low, I can get to a 486 DX2/50 (according to Speedsys), but the game runs too slow.

In my experience, Speedsys is not accurate for fast machines with throttled speed/disabled caches (many times it will show a much higher speed than it should - it's unclear what underlying tests it does to evaluate the performance).
Something like 3DBench 1.0C is much more accurate, because it will show you what the actual VGA performance is. For Titus the Fox, you generally want a score between ~ 19 and 28 FPS in 3DBench 1.0c (equivalent to the 486SX-20 <--> 486DX-33 speed interval). Norton System Info is also decently accurate.
So, depending on the platform, the CPU you're using (and its native frequency), you might achieve this speed by simply disabling the L1 cache, without needing Throttle at all (if you use Throttle, speed might drop too much and you will again introduce slowdown/stutters).

For Jazz (on VIA KTxxx chipsets!), you want to drop your CPU & FSB frequency to the lowest that your CPU/motherboard supports, and then use Throttle up to levels 5 or 6 (maybe a bit more), to get to a Pentium MMX equivalent speed. This will get rid of the Turbo Pascal divide by 0 error (or you can patch the game with the TPpatch and then you can play it at full CPU speed - however, with certain sound cards, sound might be affected/run at a lower sample rate 🤣 ). Do NOT disable the L1 cache, Jazz runs much better on very late 486 or Pentium systems (~ 60 - 150 FPS in 3DBench 1.0C). You can disable the L2 cache, though (but the impact will be limited). Again, this is not an exact science (if it were, things would be much easier). It's up to you to evaluate what performance a game needs and play with the settings to achieve said speed.

Now, it's been a long time since I've tested a KT133 board (the last one I've tested was the ASUS A7V a few years ago), but according to my notes, ACPI Throttle worked fine, just like on newer KT400/KT600/KT880 boards.
The idea is to only use ACPI Throttling if you actually need it. 😀

mockingbird wrote on 2022-09-04, 20:16:

You say you managed to get Titus running perfectly with your Ezra-T?

I doubt that there's a single DOS game that won't run perfectly on the Ezra-T (with the right combination/settings). With the Ezra-T, you don't need ACP Throttle... at all! Man, I should really make a video about this CPU's DOS versatility, but I'm just way too lazy... 😁 (and busy 🙁 )
Of course, the Ezra-T is only part of the equation. The video card that you're using with it will also be important (and, let's face it, this CPU needs a good videocard - either a GeForce 2 GTS/PRO/Ti or at least a Voodoo 3 3000). These video cards are generally very compatible, but there will undoubtably still be some games that have issues with them (still, overall, we're talking about very few games).

As with VIA KTxxx chipsets and Throttle, it's basically a matter of knowing what speed a certain game needs (and knowing what tools actually show the correct speed - again, 3DBench 1.0c has always worked perfectly for me, so it's only a matter of learning what FPS is equivalent to what speed).