G4 400 Spanks Dual P3 600

Kurt · Nov 9, 1999

According to <a href=http://www.barefeats.com/pentium.html>this</a> article anyway. Interesting reading, although it's a pile of hooey.Special thanks to whoever posted this bit of humor over at AI, it had me entertained for about 10 minutes.

resteves · Nov 9, 1999

And what, pray tell, leads you to decide this is hooey??He even lists the configurations used.

John · Nov 9, 1999

It's not "hooey," it's just meaningless. It's a few random tests. It doesn't even attempt to give a comprehensive picture of performance. But hey, if you really need to render that Bryce rock island over and over and over again in your daily work, these are the benchmarks for you! View image: /infopop/emoticons\icon_wink.gif

Easy Rhino · Nov 9, 1999

Slagenthor · Nov 9, 1999

I don't know about "pile of hooey"...but questionable, yes.First, off, with the possible exception of Photoshop, the test is really between a G4 400 and a single P3 600. I am not sure about Bryce, but I believe it does not have SMP support, I could be wrong though. Quake2 and Unreal certainly do not have SMP support, so regardless of the fact they are running on NT, only one CPU gets used for them at one time.NT does not run either Q2 or Unreal as well as Win9x, I believe. Primarily due to driver support and lack of Direct X support above 3.0.Then numbers for Quake2 and Unreal on the PIII look low. Especially Quake2 Timedemo1...they look more like Crusher timedemo numbers which are usually about 30% lower.Photoshop is well known to be better optimized for the PPC than for the x86, this is no secret, and no bad thing....just don't tell me that it's a good yardstick for cross-platform benchmarking, it's not. That would be comparable to using Quake1 scores to say the PIII romps on the G4.

total1087 · Nov 9, 1999

<BLOCKQUOTE>quote:<HR>66MHz PCI slot with 3Dfx Voodoo3 3000 running at 166MHz<HR></BLOCKQUOTE>This is quoted from the Dell side of things. It's kinda pathetic to point out that on the PC side, we do not have 66Mhz PCI slots. If they were, they'd be the 64-bit version that the G4 system has. So, with that concluded, this test is pure nonsense, like John stated. Especially when he never stated what he did to secure both test beds to run optimally. I can also tell you for a fact that 3dfx's Voodoo3 NT drivers are crap. My friend upgraded his TNT PCI card to one of these on his NT 4.0 WS system at home, and his system slowed down by half. Heck, he couldn't use Photoshop anymore without it lagging real hard on him. That's why if you get a Voodoo3, make sure you aren't going to be running it under NT, but run it under Win98 and Linux. What pissed me off was he never stated if any of these programs were set to use the SMP functionality on the Dell side of things. He never stated if he re-installed NT or not, if the NT kernel saw both CPU's when booting up, etc (he obviously knew what he was doing on the G4, but definately didn't know what he was doing on the Dell, as proved by the quote). So yeah, this whole thing was meaningless, and it at least shows his ignorance on the Dell. Please, take this guy's supposed benchmarking with a grain of salt the size of China.

Evil_Merlin · Nov 9, 1999

I think the dolt who ran this "test" had better learn the difference between having an SMP system and USING SMP. Is the person using WindowsNT? What video card? How much ram is in the Yikes! System? Dear god... Resteves, Are you that slow to think this was a good test???[This message has been edited by Evil_Merlin (edited November 09, 1999).]

Geon · Nov 9, 1999

I thought the G4's has ATi Rage 128, NOT Voodoo3 3000.Yeah, I think this guy's tests are hooyeh (sp?), mostly.I am very impressed with the G4s, and suspect that in ideal cases they would spank the P6 clock-for-clock. Unfortunately, P6 is available (theoretically) 733mhz, compared to 450mhz for G4.As for objectivity in reporting system setup - specifying processor speed, memory bus, and memory (in ONE of the machines - for all anyone knows, that G4 had 512MB RAM - the G4 memory was ommitted).

resteves · Nov 9, 1999

Do you guys just not bother to read? Most of the questions (not all) are answered right on the site.I think the dolt who ran this "test" had better learn the difference between having an SMP system and USING SMP ----------- He stated it was a friends system, so he probably didn't do any of the installation stuff. . Is the person using WindowsNT? -------- YES What video card? -------------- Voodoo3 3000 How much ram is in the Yikes! System? ------------- 256 meg, same as Dell I thought the G4's has ATi Rage 128, NOT Voodoo3 3000. ----------- The *come* with rage128, you can still put in another cardAs for objectivity in reporting system setup - specifying processor speed, memory bus, and memory (in ONE of the machines - for all anyone knows, that G4 had 512MB RAM - the G4 memory was ommitted). -------------- Nope, it is listed for both Resteves, Are you that slow to think this was a good test??? ------------ Yes I am slow ;-) But I think this test is not very comprehensive. As pointed out, many apps don't take advantage of SMP and this just points at a few apps, if these are what you run, then it is a good benchmark.

Kurt · Nov 9, 1999

It's still a bunch of hooey. <BLOCKQUOTE>quote:<HR>Mixed results but I call the G4 the winner. After all, it's got one CPU running at at two-thirds the speed of the Dual Pentium. That's like picking a fight with a gorilla with one hand tied behind your back. <HR></BLOCKQUOTE> No, it's like tying the Gorilla's arm behind its back, since he didn't bother to test anything that would actually benefit from having multiple processors to use. Of course, he did manage to use Altivec for some things. He really should have used some SSE enabled filters, unless that would be cheating or something.His webpage doesn't seem to be loading the entire table now though, it get's to the Dell's RAM and then nothing else shows, which is why he seems to have left a lot of questions unanswered. [This message has been edited by GoofBall (edited November 09, 1999).]

Roman A'Clef · Nov 9, 1999

Let us not forget that the Mac is also using the "Yikes" Motherboard, and so does not have the high memory bandwidth or AGP.Just thought I'd throw that into the mix....

Laen · Nov 9, 1999

resteves I am not sure wether you know this or not, but you can't just throw an extra proccessor in a NT machine and expect it to work. That is why you need information about things like NT4? What SP? Why run tests on an SMP machine not using SMP applicaions? The test maybe ok, but it is purposely misleading, by implying you would try to use these two machines in the same area.

Dan · Nov 9, 1999

Hooey. It's got a certain ring to it.

Taifun · Nov 9, 1999

Aren't Unreal and QII fillrate limited on a V3 3000 at that res anyway?

Beosiman · Nov 9, 1999

Let's wait until Bryce appears on the BeOS, http://www.be.com/press/pressreleases/99-06-22_meta.html then we can try some REAL SMP vs. G450 comparisons.By then we should all be running r5 of the BeOS with full Open GL implementation.Hey, anyone notice that Be's stock closed above 15 points today?

Slagenthor · Nov 9, 1999

Total:Good point about the bus (66 Mhz PCI )...or lack thereof I should say. Also whether or not NT was re-installed for SMP.I agree that its obvious that the guy knew what he was doing on the Mac side of things, and not on the Dell, evidence his saying "117 megs" reserved for EACH application...uh...okay...on the Mac, he managed his memory. certainly not on the Dell.Even so...it's good to see some benchmarks for the G4. I would have hoped to see some Q3A Test benchmarks for it though.Anybody out there know of any yet?

Kurt · Nov 9, 1999

-- View image here: http://www.angelfire.com/biz3/goofball01/images/BeStock.jpg -- Yes, I noticed. Makes me wish I invested in Be when I could have. Doh![This message has been edited by GoofBall (edited November 09, 1999).]

Geon · Nov 9, 1999

About the 256 megs of RAM for the G4 - I've looked at the page using Opera3.6, IE5, and Navigator4.6, all under Win98, and I do not see that reported in the table.Maybe my system is completely screwed. Or maybe I am hallucinating. But I still do not see a number for the G4 in that table.I did not know you could pop in a PC graphics card into a mac and have it work - I thought there were severe driver availability issues - guess I am wrong.

Kurt · Nov 9, 1999

It finally worked again, here's the table with the specs... <TABLE BORDER=1 WIDTH="89%"> <TR> <TD WIDTH=79> </TD> <TD WIDTH=160> <CENTER>Dell Dual Pentium</CENTER> </TD> <TD WIDTH=156> <CENTER>Apple G4</CENTER> </TD> </TR> <TR> <TD WIDTH=79> CPU </TD> <TD WIDTH=160> <CENTER>Pentium III 600MHz times two</CENTER> </TD> <TD WIDTH=156> <CENTER>G4 400MHz with Velocity Engine</CENTER> </TD> </TR> <TR> <TD WIDTH=79> Bus Speed </TD> <TD WIDTH=160> <CENTER>100Mhz</CENTER> </TD> <TD WIDTH=156> <CENTER>100Mhz</CENTER> </TD> </TR> <TR> <TD WIDTH=79> RAM </TD> <TD WIDTH=160> <CENTER>256 MB</CENTER> </TD> <TD WIDTH=156> <CENTER>256 MB</CENTER> </TD> </TR> <TR> <TD WIDTH=79> Graphics </TD> <TD WIDTH=160> <CENTER>66MHz PCI slot with 3Dfx Voodoo3 3000 running at 166MHz</CENTER> </TD> <TD WIDTH=156> <CENTER>66MHz PCI slot with 3Dfx Voodoo3 3000 running at 166MHz</CENTER> </TD> </TR> <TR> <TD WIDTH=79> Operating System </TD> <TD WIDTH=160> <CENTER>Windows NT</CENTER> </TD> <TD WIDTH=156> <CENTER>Mac OS 8.6</CENTER> </TD> </TR> <TR> <TD WIDTH=79> Price </TD> <TD WIDTH=160> <CENTER>$4109 from Dell On-line Store (not counting Voodoo card)</CENTER> </TD> <TD WIDTH=156> <CENTER>$2098 from Apple On-line Store for the Yikes model (not counting Voodoo card)</CENTER> </TD> </TR> </TABLE>

Geon · Nov 9, 1999

Okay, thanks, so something must have been screwed up. I take back what I said about the 256MB RAM

total1087 · Nov 9, 1999

What I wanna know is where he got the prices from. I guestimated that the Dell is a DELL PRECISION WORKSTATION 410, but when I tried to to match the pricing, I could not. The Dell comes with an LVD 9.1gig HDD default, along with a SCSI burner and EIDE reader (dumb setup). It also has all of this warranty extra stuff that could change the pricing around. It uses ECC memory, so this setup can easilty be turned into a nice e-commerce server (among other things). It also comes default with a Diamd Fire GL1, using IBM's new 256-bit video chipset (that rival's the GVX1 among other high-end graphic cards).When I went to try and get the same price on the G4 system, the lowest I could go is 2699. It's still EDIE and I had to take out the v.90 modem (add $100 if you must). For SCSI, you had to add $49 for only Ultra SCSI (Narrow), and the SCSI drives the Apple store is selling are very expensive, but if you WERE to add SCSI to it, it would be 3548 (including the DVD-RAM, so it would make both comparable). Even then, you would need to get a Ultra2 SCSI card for your system (you don't have the option the get a U2W card from Apple, and U2W cards are NOT $49, but range around $250).Oh, and resteves, - <BLOCKQUOTE>quote:<HR>if these are what you run, then it is a good benchmark<HR></BLOCKQUOTE> What a load of BS. He never stated if the Dell side of things are optimized for SMP or even SSE (which on the G4 side of things, Adobe 5.5 is already optimized for PPC and the Velocity Engine). So in fact, this benchmark only proves that the guy didn't know what he was doing when he benchmarked the Dell. It also shows biased results from his ignorance. Even then, the Dell setup smacked it around a bit in most of the benchmarks. But that isn't the point. The point is that this is a poor "vs." benchmarking to begin with. And that is why some(most) of us are pointing out to you that this whole thing is meaningless and "hooey".

Easy Rhino · Nov 9, 1999

And let us not forget that the Mac was running MacOS, which in my book pretty much sends it straight to the crapper before you even turn the power on.

Kurt · Nov 9, 1999

You're right Easy Rhino, that is pretty unfair.

Scott Tarr · Nov 10, 1999

Just for the record, Bryce *isn't* optimized for SMP. I could throw 300 processors at it and it would only see one of them. For a 3D program, it is dumb as dirt. But one must remember, that MetaCreations used to be a Mac only app company and they have that (one processor) Mac mentality. Until Bryce is released for Be (or should I say *even if*), it won't have SMP capability. BeOS looked to be the first OS that would be optimized for SMP for Bryce. Because of the tests and their lack of SMP optimized apps used in testing, the test then becomes null and void. I hypothesize that the person running the tests *knew* of the SMP-dumbness of the apps and run the tests knowing full well of the outcome. Heck, I can buy a Ferrari for $100,000 and stick it next to a F1 car (that cost over $1,000,000) and on the grass infield the Ferrari is going to smoke the F1 simply because of the F1s tires made for pavement and nothing else. The same thing is done here. Neither test is fair and unbiased.

Darken · Nov 10, 1999

hmmm regardless the G4 400 (on the slower Yikes mother board) pretty much out preformed the 600mhz P3. You have to admit that thats pretty good ;-)And dont even get started on the whole "coppermine and athalons are faster" stuff.

total1087 · Nov 10, 1999

Eh? What are you talking about, Darken?? From what the table says that Goofball posted, the G4 is running on 100Mhz FSB, along with 66Mhz 64-bit PCI slots. How is that slow?

JB · Nov 10, 1999

Dual CPU PIII systems are not that much faster than heir single PIII brethren. The advantage of a dual system ( I know be cause I run NT on two CPU's) is the added stability and the ability to _run_ and _use_ many resource hungry apps at once. And yes one needs the NT Resource Kit to be able to go from one CPU to two. The utility to do this does not exist in NT without it.

TGR · Nov 10, 1999

Darken refers to the fact that the Yikes board isn't as fast as the umb board, which has new controllers, the AGP slot, etc. Before the speed downgrade, the 400mhz G4 (Yikes) used the G3 board, now the 350 mhz G4 uses that board (new Yikes) and the new 400 and 450 use the umb board.Here's one thought on why the dual NT 4.0 might not have been set up as some have speculated it should have been: reintall to assure NT detects both cpus, etc. As a Mac guy, the tester probably assumed that the Dual Pentium rig would have been set up correctly and ready to run both CPUs. Why? In a Mac it would have. A Mac user assumes that if the hardware is there and the OS is set properly, then it will work. Just a thought.see ya,

DrPizza · Nov 10, 1999

Here's what I said at AI: [ QUOTE ] ROTFLMAO.Inaccurate? I'll say so.The PC has: 66MHz PCI slot with 3Dfx Voodoo3 3000 running at 166MHz Er, no, it doesn't. It doesn't have a 66 MHz PCI slot. 66 MHz AGP, yes, 66 MHz PCI, no. Or, if they used PCI, then the PC is hindered severely by having 33 MHz PCI, thus halving the bandwidth.Photoshop 5.5 was given 117MB of app size on each machine Er, no, it wasn't. Maybe on the Mac, but you can't do that in Windows.Quake II test was run by pressing "~" after startup and typing "timedemo 1" (return) and "~" again. When the first person dies, I press "~" again to get the frame rate. QuakeII isn't multiprocessor aware, so it's not a dual PIII machine against a single G4, it's a single PIII against a single G4. Waited 'til the first person died? Oh yeah, that skips out all the bits that might tax the processor by a minute amount. The PIII, at least, needs a faster video card to allow the processor to show what it can do.The G4 was faster in one of the Photoshop tests. Kinda contrary to the impression I got from Apple.Using the latest version of Bryce (apparently) makes the Apple slower. As there are (apparently) only two seconds in it, might that make the PC faster?Do we know what hard disks the machines have? Do we know what drivers they were using? Do we know why the PC had a crippled video card? Do we know which OS and SP level was installed on the PC?The answers to all these questions is 'no'.Further, we *know* that the person running the tests is biased. How? Look at the rest of the site. It's a Mac site, and I'm sure that none of the Mac users (or Mac lovers, as they profess themselves to be) would trust a similar PC site.Furthermore, why not compare like with like? Why not compare the *best* PC available for $2000 plus the cost of a Voodoo card, with the *best* G4 available for $2000 plus the price of a Voodoo card. Plus, Dell don't sell their machines on speed; they sell them as reliable workhorses. Not slow, but not hell-bent on performance.Apple, OTOH, tout performance (and have done for a long time) as an important factor.But apart from that, it's a great comparison.[unquote -- deliberate non-UBB code View image: /infopop/emoticons/icon_biggrin.gif]Yes, there _is_ a memory allocation box in Photoshop. Quite what it does, I don't know. It shouldn't be there, it's a dumb thing to have in a Windows program.I didn't mention that Bryce wasn't multiprocessor aware, because I wasn't 100% certain about that.The video card is really crippling the PC.And also, why is a PC whose number one feature is rock solid reliability (it is, after all, a Dell) being compared to a machine whose makers tout the performance above everything else? Why not compare the G4 to a machine made by a smaller company going for out and out performance?And if the person is allowed to tweak Photoshop on the Mac (by having to command-click it and select its memory allocation) then am I allowed to have a play around with the Dell machine and tweak it? <BLOCKQUOTE>quote:<HR>NT does not run either Q2 or Unreal as well as Win9x, I believe. Primarily due to driver support and lack of Direct X support above 3.0.<HR></BLOCKQUOTE> QII is fastest with OpenGL graphics, which doesn't use DirectX, and benchtests *should* be done with sound turned off (hence doesn't use DirectX). Unreal is heavily reliant on DirectX, most of the time, so is probably being hindered to a greater or lesser extent. [This message has been edited by PeterB (edited November 11, 1999).]

Slagenthor · Nov 10, 1999

PeterB: "QII is fastest with OpenGL graphics, which doesn't use DirectX, and benchtests *should* be done with sound turned off (hence doesn't use DirectX). Unreal is heavily reliant on DirectX, most of the time, so is probably being hindered to a greater or lesser extent."Yes QII needs OGL and not DX, barring a wrapper I thought it was OpenGL only. But my point was the OGL drivers too. I have a Voodoo3 running both NT and 98. Suffice to say, the NT driver OpenGl performance blows. IMHO, you should leave the sound on while benchmarking, because that's how the games are played., my $0.02 though View image: /infopop/emoticons/icon_smile.gif

Scott Tarr · Nov 10, 1999

The reason that the sound should be off is that to isolate it to just the cpu-vid card, you want to eliminate any obstacles that would hinder performance. Seeing as how sound is a bit different on a Mac than a PC, you would want to turn it off, in this case. If only, for arguments sake. View image: /infopop/emoticons\icon_wink.gif

Atilla · Nov 10, 1999

Sounds like MP G4's are the way to go when OS X comes. Who cares no one switches OS's simply for speed alone. Legacy apps weigh heavily in any OS switch. What is relevent is that it appears to be much faster than a G3. I can see no compelling reason to switch to Macintosh OS from Windows. Of course, I can see no reason to have started with Windows in the first place. Games and a few accounting apps are just not enough reason to take the plunge into the Windows world.

elevate · Nov 10, 1999

"Games and a few accounting apps are just not enough reason to take the plunge into the Windows world." uh...what?

Boogeyman · Nov 11, 1999

Oh Atilla you're absolutely right. Who needs all those apps that you can get for Windows that you can't for Macs? Who needs game, either? Hell, who needs a better hardware selection? I guess if you want to be running Office 98, Communicator 4.7, and Quake 3/Unreal all on a computer with a 450 MHz processor, 100 MHz fsb, and a lousy video card, then Macs are the way to go.

resteves · Nov 11, 1999

How/When did a Voodoo 3 3000 become a lousy video card??Its like many people define good/bad based on what isn't/is available for a Mac. I have never had trouble getting software programs for my Mac. (excluding games)And a larger hardware selection is not necessarily a better selection.

DrPizza · Nov 11, 1999

<BLOCKQUOTE>quote:<HR>How/When did a Voodoo 3 3000 become a lousy video card??<HR></BLOCKQUOTE> It's always been lousy. Particularly in its 33 MHz PCI incarnation (as in, the version used on the PC in the test). It doesn't do 32-bit colour, it doesn't look that nice, it doesn't do OpenGL properly. <BLOCKQUOTE>quote:<HR>Its like many people define good/bad based on what isn't/is available for a Mac.<HR></BLOCKQUOTE> On the contrary. The Mac has plenty of good technology. It has UltraATA, it has AGP, it has PC100 RAM, it has PCI.... SCSI is good. Ultra2 Wide SCSI (I get it on Saturday. Wahoo!) is better.OTOH, ATI Rage XXX is not good. 66 MHz 32-bit PCI ain't good for graphics cards (versus AGP, AGP x2, AGP x4). Crapola OS ain't good. Single processors ain't good. PC100 ain't as good as PC133 (or even, god forbid it should ever take off, 800+ MHz RDRAM). A maximum of 1.5 Gb RAM ain't good. Three (or so) expansion slots ain't good. Integrated sound ain't good. Closed architecture ain't good.Big bucks for the machines certainly ain't good.

Laen · Nov 11, 1999

Mac users need to be aware of something here. You are getting OpenGl for macs which is great, except you need to know Voodoo's run OpenGL for crap. Glide, 3dfx's API, they do great and in software that supports both Glide and Opengl Voodoo's can compare with TNT2's. OpenGl on the other Voodoo's suck at, and can come NOWHERE near a TNT2 in speed, performance, and looks. Edit Horrible spelling and punctuation. Now its a little better. [This message has been edited by Laen (edited November 11, 1999).]

resteves · Nov 11, 1999

I keep hearing that macs suck because they don't do SMP, yet now people are mad about this comparison because the apps don't usually see/use multiple procs. I think that was a large purpose to this test, that having a dual processor isn't that big of a deal, unless you are using specific apps. (Sort of like altivec, but you guys say that is a waste, yet SMP is the bomb.) Also, as I understand it (and I could easily be wrong) Dual procs will always help, even if the app isn't SMP aware, the OS will let the app use one proc, while the OS uses the other one.For those of you happily assuming the worst about this comparison... I did a little checking with the site owner.YES NT was seeing/using both procs.YES photoshop was using both procs, and using altivec and SSE.Apparently the Dell doc's refer to both 33mhz and 66mhz slots, he is looking into this. He is planning on doing the same test with two AGP cards soon.The default memory allocation for photoshop on windows is 50%, so he set the Mac for an equivalent amount, 117meg.He is planning on doing similar tests with a single P3 running Win98.

Boogeyman · Nov 11, 1999

Yeah but you're forgetting something, resteves. If he was only running one app on the Mac, and the computer was doing nothing else (which I would assume, since the Mac's multitasking sucks so much ass that if he was running something else at the time the Mac would have lost big time), then there really is no difference between the 2 tests. Since the current app on Macs takes almost complete command of the processor, that would be the same thing as an app under NT getting one processor to itself. The Mac would still be at a slight disadvantage, but that's still something you have to consider.But these tests are still bogus - go over to AnandTech or the FiringSquad and take a look at the detail they go into in describing their systems and how they ran the tests. That review is bogus - no 2 ways about it.

Scott Tarr · Nov 11, 1999

resteves stated: <BLOCKQUOTE>quote:<HR>. I think that was a large purpose to this test, that having a dual processor isn't that big of a deal, unless you are using specific apps.<HR></BLOCKQUOTE> Okay, lets use your excuse of this as the reason for this test. Were there any conclusions drawn after the test? NO! So, this isn't the case. You have some idiot (or evil genius) run these tests and effectively tie the second processor behind its back. Get this (because this is the fifth or sixth time I have had to explain this to Mac users...who tell me they know computers [read over at <a href=http://forum.appleinsider.com/ubb/Forum4/HTML/001425.html target=blank>AI</a> what is being said here] the second processor wasn't in use in most of these tests If that were the case, the reviewer should have explained the reasoning behind it. If the reviewer was wanting to show the difference between SMP and non-SMP apps, then SMP apps should have been in the test. This was a con from the beginning and will continue to be one.

G4 400 Spanks Dual P3 600

Ars Tribunus Angusticlavius

Ars Tribunus Militum

Ars Praefectus

Ars Scholae Palatinae

Ars Centurion

Ars Scholae Palatinae

Ars Legatus Legionis

Ars Praefectus

Ars Tribunus Militum

Ars Tribunus Angusticlavius

Ars Scholae Palatinae

Ars Scholae Palatinae

Ars Tribunus Angusticlavius

Ars Praetorian

Ars Tribunus Militum

Ars Centurion

Ars Tribunus Angusticlavius

Ars Praefectus

Ars Tribunus Angusticlavius

Ars Praefectus

Ars Scholae Palatinae

Ars Scholae Palatinae

Ars Tribunus Angusticlavius

Ars Praefectus

Seniorius Lurkius

Ars Scholae Palatinae

Ars Praefectus

Seniorius Lurkius

Well-known member

Ars Centurion

Ars Praefectus

Ars Scholae Palatinae

Ars Scholae Palatinae

Seniorius Lurkius

Ars Tribunus Militum

Well-known member

Ars Scholae Palatinae

Ars Tribunus Militum

Seniorius Lurkius

Ars Praefectus

nproxy.org