They (celerons) are just that bad. The later models got a little less sucky because Atoms took their place as the "budget" models.

I guess the Celeron 700 is bandwidth starved with its 66MHz FSB while the P2-400 can actually move more data through its 100MHz bus. So it may be the bottleneck.
It's like the Celeron could calculate more frames, it is indeed a faster CPU on the inside, but it just can't get the data to calculate from. Or can't put it in place fast enough where it can be picked up by the VGA.

Look at this review: https://www.anandtech.com/show/568/9

But cpu's are cheap as chips so just buy a faster one if you need it....

May I also recommend giving UTGLR a try? Unreal Engine games are notorious for running like crap on anything but Glide, so this alternate OpenGL renderer is fantastic and might actually fix your speed issues.

meljor wrote:

Look at this review: https://www.anandtech.com/show/568/9

OK... but according to that performance should be just a bit slower than an Athlon 500, which is faster than a P3-500. The Celeron is definitely slow, clock for clock, due to its high multiplier and low cache, but a C700 should hold its own against a P2-400.

I suspect something else is also holding CPU back. OP, are you running these CPUs on the same system? If not, what are the rest of the specs?

I recall moving from Celeron 366A to 700 at work around 2000. It was a nightmare. Not because the CPU itself was slower, but because we went from an i440EX + discrete (4MB) Rage Pro VGA to i810 with integrated VGA. So half the - already tight - memory bandwidth was eaten by the iGPU, resulting in terrible performance. The "upgrade" from Win98SE to WinME didn't help either...

Some time ago I ran a couple of 600Mhz cpu's in PCMark 2002
ASUS P3V4X - 2Gb RAM - Win XP
oooooooooooooooooooo
600 Celeron fsb66 coppermine core (@slotket) :
CPU score: 1470
Memory score: 1067
HDD score: 934
oooooooooooooooooooo
PIII 600B fsb133 Katmai:
CPU score: 1546
Memory score: 1360
HDD score: 1038
oooooooooooooooooooo
PIII 600E fsb100 Coppermine:
CPU score: 1599
Memory score: 1396
HDD score: 1043
oooooooooooooooooooo
PIII 600EB fsb133 Coppermine:
CPU score: 1623
Memory score: 1504
HDD score: 1031
oooooooooooooooooooo

for comparison:
oooooooooooooooooooo
PIII 1000 fsb133 :
CPU score: 2571
Memory score: 1747
HDD score: 1040
oooooooooooooooooooo

But for some reason the Celeron 600 performed very poorly in 3DMark2000 1024x768x16 compared to the others (with XFX ti 4200 - driver 41.09).
2700 3dMark2000 points for the Celeron / 4246 3dMark2000 points for the PIII 600E

IIRC, back in the days, a Celeron 300A @450 was very close to be on par with a PII 450, cribbled off course by the size of the cache, but most was won back by the full-speed on-die cache.

But again - that was @100 MHz FSB, so ..

Gahhhrrrlic wrote:

Is it normal for a P2 400 to do a better job at playing say... Unreal Tournament than a Celeron 700?

no

Gahhhrrrlic wrote:

The weird thing is, the minimum system requirements are 200MHz.

233Mhz, and those are post stamp sized window at 15fps absolute minimum specs

Gahhhrrrlic wrote:

So, under what circumstances is the game playable at 200 MHz

not playable, its runnable

Gahhhrrrlic wrote:

Half-Life actually plays better on my P400, which seems to support the hypothesis that the system is CPU bottlenecked.

P400 in same system? or another computer?

Celeron 700 should be somewhere between Pentium III 450 and Pentium III 500E

jmarsh wrote:

They (celerons) are just that bad.

the only bad celerons were cacheless Covington and later P4 variants. Mendocino, Coppermine (OPs 700 Mhz) and Core based ones were brilliant perf/$ after overclocking. Almost all Coppermine 66MHz fsb ones could be forced to run at or close to 100MHz equaling Duron clock for clock in games.

cribbled off course by the size of the cache

Overclocked Celeron Mendocino is better or equal to Pentium II in practically any test. Only Pentium III starts to get faster due to SSE support and better clock speeds (600+ mhz is possible even on Katmai core).

equaling Duron clock for clock in games.

Duron should be faster.

rasz_pl wrote:

8><CUT […]
Show full quote

8><CUT

jmarsh wrote:

They (celerons) are just that bad.

8><CUT
Mendocino, Coppermine (OPs 700 Mhz) and Core based ones were brilliant perf/$ after overclocking. Almost all Coppermine 66MHz fsb ones could be forced to run at or close to 100MHz equaling Duron clock for clock in games.

It seems hardly fair to judge midrange Celerons only for their 'gaming performance'. The 600Mhz Celeron that I used for the benchmarking earlier in this thread works at standard 1.5 volt and runs at fsb 83 without overvolting. For what it was meant to be and how it was meant be used it was possibly indeed bordering on 'brilliant'.

The 66FSB Celerons were terribad for gaming. There's barely any performance scaling past 533MHz due to that dinky 66Mhz bus that's straight outta 1993. I wouldn't be surprised to see a Pentium around the 450MHz mark beat a 700/66 Celeron in games. Try running it at 100MHz FSB if possible. Or, if your motherboard supports 133MHz FSB, just get a dirt cheap PIII EB and double your gaming performance.

PARKE wrote:

But for some reason the Celeron 600 performed very poorly in 3DMark2000 1024x768x16 compared to the others (with XFX ti 4200 - driver 41.09).
2700 3dMark2000 points for the Celeron / 4246 3dMark2000 points for the PIII 600E

I've noticed that 3DMark 2000 is extremely sensitive to cache size. Far more than 3DMark99 and 2001 even. The 128K Celerons based on Coppermine, Willamette, and Northwood absolutely tank in this benchmark. Only the Tualatin Celly does somewhat OK, thanks to its 256K L2. Still, it's fairly limited by the 100MHz FSB though. What's funny is seeing an overclocked PIII-S with a full 512K L2 and 150+FSB easily take out a Northwood Celeron running a full GHz faster in this particular benchmark. 😜

Time to play catchup on some posts:

So what I'm seeing so far is, yes the C700 sucks but actually it should be able to run the game at 30 FPS at least. Mine is in the neighbourhood of 5, if playing the map "Liandri" out in the open and not in a corridor.

My Mobo is a PC Partner Apollo Pro 133

I have a different system with the P2B running a 400MHz (nominal) slot CPU. That does pretty good on both HL and Unreal (considering its speed).

I tried OC'ing the CPU. However as soon as I bump the FSB, the PCI frequency goes with it so I can't do much better than 788 MHz or therabouts. Multiplier is locked and I can't find a way to move only the FSB without the PCI being affected.

Going to try UTGLR

EDIT:

No effect on speed (although is looks nice)

What version(patch) is UT? Earlier versions were heavily optimized for glide (3dfx cards) and D3D/OpenGL cards were very slow. Even the latest patch will run quite slow, unless you have a good cpu/video. Try running the game with a 3dfx card, if you have one. V3 3000 should be a good choice for that CPU. V2 or V3 2000 should work fine, too. Banshee is ok, but not great, since it lacks texturing. V1/Rush is too slow for UT.

Radeon 9500 might seem a fast card for UT, but it isn't that fast, I'd say that with celeron 700 it will run slower than a v3 3000 (using glide). That's not because v3 3000 is faster, but because UT really likes glide.

But what if the graphics aren't the problem? Granted the renderers have their pros and cons and the 9500 may not be a good match but it's still a 9500. It's like an 8.4L V8 vs a 2.0 twin turbo. The TT might put out 300 HP which is amazing for its displacement but the 8.4 is still putting out 600 something HP, as inefficient as that might sound. The 9500 ought to be able to run UT99 at 20FPS at all times on either DX or OpenGL. If it can't, and if the resolution has no effect on FPS, isn't it rational to conclude that the problem lies elsewhere?

I have already edited the .ini to increase the cache to 12MB from 4. It made load times longer and probably helped smooth out transitions but I can enter the same exact space repeatedly and get the same lag (suggesting that it's not caching from the hdd for the first time that's causing the slow down). I'm using PC133 RAM anyway in this system and with a 66MHz FSB, it is probably actually running at 133 so that can't be too slow. The HDD is balls and I have my doubts as to whether it's actually using DMA (although the primary ide says it is) but even on the very first deathmatch map, which is rather small and should be easy to buffer into ram AND only has 1 opponent, it's still slowing down to the point I can't kill the bastard because I'm waiting for the screen to catch up when I aim my shots.

All signs pointing to the CPU and yet I still can't believe that 700 MHz anything isn't good enough for UT99. It just doesn't compute.

Gahhhrrrlic wrote:

My Mobo is a PC Partner Apollo Pro 133

I never had PC Partner products in my hands in Europe, but according to redhill.net.au they were average run of the mill, definitely not terrible.
Apollo Pro 133 on the other hand .... expect tons of compatibility problems

Gahhhrrrlic wrote:

Mine is in the neighbourhood of 5

most likely you are missing 4in1 AGP driver, what gpu are you using?

Gahhhrrrlic wrote:

I have a different system with the P2B

so average shit chipset mobo versus the best slot1 motherboard ever made? 😀

Gahhhrrrlic wrote:

I tried OC'ing the CPU. However as soon as I bump the FSB, the PCI frequency goes with it so I can't do much better than 788 MHz or therabouts. Multiplier is locked and I can't find a way to move only the FSB without the PCI being affected.

I suspect board doesnt have voltage control? you wont do much without it/modding the board. 700 has a chance of running 1050MHz at 1.8V

Still you should be getting 40 fps even on stock clock. Start with AGP drivers for crap VIA chipset

I should have mentioned I installed the via_hyperionpro_v524a.zip drivers from phil's lab

Gahhhrrrlic wrote:

Time to play catchup on some posts: [....] […]
Show full quote

Time to play catchup on some posts:
[....]

My Mobo is a PC Partner Apollo Pro 133

[...]

I have a different system with the P2B

Yep, that's the difference. You're comparing rotten apples to glorious oranges.

- ApolloPro133 is nice for high FSB, but a CPU at a given speed on ApolloPro133 clocked at 133MHz FSB loses badly to an equally clocked CPU at 100MHz FSB on i440BX. You're running at 66MHz, not 133MHz, so mem performance will be downright abysmal - and with that 66MHz FSB and high multiplier, that's the Celeron's bottleneck for startes.
- Asus' late 1990s boards were the cream of the cream, with proven rock-solid stability and very nicely tuned performance. PCPartner is Vtech's PC brand, and while not abysmal, isn't even in the same league as Asus either in terms of default settings or of optional ones you can tweak yourself.

So yeah, that explains why the C700 is performing a good 25% below what it should be doing. Still, that would result in about equal performance, so check those 4-in-1 drivers. If they're missing that explains the rest.

Next time when you want to compare, do it on the same platform...

All I said is "try a 3dfx video card". Yes, you can draw a conclusion based on facts and logic, but sometime you don't have all the facts. The car analogy is not that good here, but if you like it think this way: it doesn't matter how many "liters" or horse powers an engine does have if you don't know other "facts". You can have a Ferrari or Lamborghini with a huge engine, if you try to go with that car on a very bad forest road you won't get far.

And I only noticed now the motherboard. Slot 1 motherboards with VIA chipsets can be extremly bad. I had a lot of experience with this, I worked a long time ago in a computer store, there were a lot of computers with slot1 motherboards and various VIA chipsets that were returned with problems. Customers complained that they were slow and crashed often (blue screens, freezes, etc). BX440 motherboards were much more stable.

Do yourself a favor, switch to BX440. Also, a PIII cpu would be much better, you should find them quite cheap, it's not 1999 anymore 😀

Untile then, you could try one more test - Quake III. I bet it will run much better than UT. If you want to play UT on a retro machine get a 3dfx card. Or a really fast non 3dfx card, like 9800XT or 5950 Ultra, but then you need AGP4X slot at least, also it won't be quite that retro. If you want time period correct experience and good FPS go with BX440 motherboard, PIII 550, 256 MB RAM and V3 3500 (or V3 3000/V3 2000). It should run very good in 1024x768x16 bit with all details maxed out.

I didn't realize the ram would run at 66. I thought there'd be some 2x multiplier or something at play.

The 4 in 1 drivers appear to be for 98 and below only. I am doing my tests on XP (because it is more stable for my sound drivers) so can I even install the 4 in 1 on XP?

RavenX: I think you might be right about Q3 engine games. I'm going to test your theory next.

I do enjoy working with authentic childhood hardware as it is - it's the real deal and if you can squeeze something out of it, it feels more satisfying. Then again, when you're working with garbage, it diminishes the potential return on investment. I wish I could OC the CPU more without messing with the PCI frequency. It seems stable enough at 37 MHz but if I go any higher something is guaranteed to blow. I've got literally every slot in this computer filled with something. Lots of chances for things to go wrong.