• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Kyro beating geforce2 pro and Ultra in new review

Page 2 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
funny how it's not getting a welter of enthusiasm, and that it's teasy and powervr2 who are managing to animate this festering horse in Anandtech forums.

if the KyroII is so great, how come so few want to talk about great experiences with it? why is there so much debate about it?
if there is debate and differences, then it is not necessarily bad, but on the other hand it cannot be simply good.
 
BFG10K,

?Wrong, the UT engine removes overdraw in software, ie the CPU. Therefore I'm interested to know why the Kyro2 is pulling so far ahead.?

It can?t remove all overdraw. Heheh ? if it did, everyone would want it and there would be no point for tilers (although, they would be the first to have accelerated that part of the game engine in hardware!).

No. I?m not overly familiar with the system UT uss but I?d imagine it would be reasonably similar to the BSP tree system Quake 3 uses. It cuts the much overdraw by eliminating the rendering on rooms / areas outside of the current area -- If you look at many of the interconnecting parts of quake3 maps they are often obscured by a wall, or a corner etc, which signifies the cut-off point for the BSP; UT is similar.

However, this means that polygon overdraw will still be present for the contents in that area ? i.e. pillars, barrels, crates, models, weapons/ powerups etc will still be present in that area and will have overdraw that can?t have been eliminated by the engine.

However, its highly likely that the reason we see KYRO performing better at UT is probably purely due to the way it handles palletised textures; we all know that this is an area Geforce cards handle exceptionally poorly.
 


<< to LSD:

it isn't

those number mine and yours where from a review (anandtech review) that used prebeta kyro drivers

you better believe that kyro can actually do better than geforce 2 pro or even ultra sometimes because it is the truth...

the more recent is the review the better kyro is... (drivers are getting better every time) >>


Yawn.
 
kyro 2 cannot be overclocked at all. The geforce pro can be overclocked to &quot;ultra&quot; levels quite easily. The geforce pro overclocked beats the kyro 2 across the board.
 
Yawn. ???
please explain that to me what that means?

my primary language is not english (it's portuguese) so I don't understand everything in english ...
what Yawn means ????

the only reason this debates exists is because there are a bunch of geforce users that like to promote what they have. why??? because they have it !
😉
imagine if kyro was a gift (like being sold for only 10 us$) and better than gefoce 3 in 99% of all the benchmarks there would always be debate because there are lots of geforce users...

I had a tnt-1 and I survived to the nvidia tribe
😉
cool
 
<<<kyro 2 cannot be overclocked at all. The geforce pro can be overclocked to &quot;ultra&quot; levels quite easily. The geforce pro overclocked beats the kyro 2 across the board.>>>

Clutching at straws a little aren't you? Any card that relize on overclocking to beat another isn't worth the extra money is it. AFAIK the Herc 4500 can clock to 190, 195 and sometimes 200mhz+, not every Geforce 2 Pro clocks by exactly the same amount either. Some might hardely clock at all and some might clock very high, its better to rely on a card that is best at defualt then something you might be able to overclock. Now I'm not saying the Kyro II is a better card then the Geforce 2 pro, it has its advantages and the kyro II has some too, kyro II wins in some games and Geforce 2 pro wins in others. Kyro II has features that the Geforce 2 Pro doesn't have and the Geforce 2 pro a feature that the Kyro II doesn't have, if the 64mb Geforce 2 Pro was the same price as the Kyro II 64mb then on a system upto 900mhz-1ghz I'd go with the Pro, but if my CPU was more in the 1.1-1.3ghz range I'd go for the Kyro II. Of course the price is not the same even in the U.S and where I live the Geforce 2 pro is over twice the cost of the Kyro II.
 
<<<funny how it's not getting a welter of enthusiasm, and that it's teasy and powervr2 who are managing to animate this festering horse in Anandtech forums.>>>

Well I've never been accused of animating a festering horse before, actually I tell a lie there was that one time but it got dismissed through lack of DNA evidence, all joking aside I don't know what your talking about man, maybe you should present a clearer argument on exactly how I'm &quot;animating a festering horse&quot;, is it just me or is that a pretty unusual thing to say.
 
&quot;Where does it say anything about beating the Ultra ?
I see GeForce II GTS, and a GeForce II Pro ..&quot;
In case you didn't read Here you go...

&quot;Well, what have we here? We have a $120 Kyro II defeating, and very soundly defeating, a GeForce 2 Pro across the board, particularly in 32-bit color, where the 3D Prophet 4500 performs as much as 20% faster than the GeForce 2 Pro. In fact, it's reasonable to assume the Kyro II would very probably defeat a GeForce 2 Ultra as well.

Again, we see the same pattern of almsot identical performance in 16-bit versus 32-bit color. Only at high resolutions does the performance begin to separate, which we suspect is due to the card running out of texture memory when in 32-bit color, forcing it to swap over the slow AGP bus.&quot;

Just remember Kyro is about 2 months old so new drivers may still bring out even more power...😉

P.S. Thnaks fo pointing out what I was talking about Teasy!

gygheyzeus... That is what mainly caught my attention! Talk about value, if Imagination Technologies brings great performance down to great prices we are looking at some exciting times.😀
 
There is no way in hell that the kyro 2 beats the geforce pro across the board. The geforce pro, using the latest drivers, beats the kyro 2 pretty much across the board. If you overclock it, you increase that gap even more. Obviously the idiot doing the review was using outdated drivers or didn't have the geforce setup properly. All other benchmarks show the pro beating the kyro in most benchmarks.
 
That was sort of my point, does anyone else have a simular setup (or know of a review) that could be used to compare the stats?

I have seen manny review, and while Kyro 2 was beating a geforce 2 MX, it wasn't by a huge amount.
 
My brother's rig which is here with me right now has a pretty similar setup. The same stuff actualy except he uses the built in sound and Windows ME. I don't know what this site's problem was, but our rig gets well into 140fps at 640-480 in Unreal Tournament and it scales as follows; 8-6=130fps, 10-7=100, 12-10=Unknown. My monitor can't go that high, but it can go to 11-8 and it only got around two fps slower than 10-7 so you do the math...I would bench Quake 3, but I kinda lost it, and my friend never lemme borrow his to um, &quot;borrow&quot;, yeah, that's it. 😉As far as FSAA goes in UT, I still don't know what they're doing wrong cuz I get well over 60 at 10-7 with the Kyro II. That's at 32bit all they way mind you. And to think my brother thinks it's too slow..::Rolls eyes::
 
Ok, after a brief look arround

The Geforce 2 GTS was performing at the same speed as a Geforce 2 MX, see my benchmark at http://gamershq.madonion.com/compare2k1.shtml?881853.

From what I can see the 3dmark 2001 scores should be a bit like so
Geforce 2 MX ~ 2867 (my setup)
Geforce 2 GTS ~ 4021 (someone else)

From these scores you can see that the I have setup a geforce 2 MX system which is scoring higher than his GTS system (~2700). Given that I wasn't even trying to setup a fast machine (no OC or anything) I think that sort of indicates that the reviwer stuffed up.

What we have here is a reviewer who can't 😀


Edit: These were all at 800x600x16, I didn't see what he was comparing at.
 
bevancoleman &quot;What we have here is a reviewer who can't😀&quot;

Just to let you know there is a Discuss this Article link at the end of the review in which you said the Geforces where possibly reviewed incorrectly. link

A reader asked him his 3dMark settings and they are as follows...

3dmark 2000
1024x768x16

3dMark 2001
1024x768x32

Rerun your setup and I feel you will fall in line with his review or at least similar...😀
 
PowerVR2:

maybe it's because there are lots of textures there

I thought of that but I wasn't sure whether it would make much of a difference. After all, ATi's Hyper-Z helps in texture-heavy situations and that barely makes a difference in the (average) scores.

That could be a possible reason though. I would expect the Kyro2 to have higher minimum framerates because of this.

PotNoodle:

It can?t remove all overdraw.

Oh yeah, I realise that. 🙂

I?m not overly familiar with the system UT uss but I?d imagine it would be reasonably similar to the BSP tree system Quake 3 uses.

Except UT does it in real-time while Quake 3 uses pre-calculated BSP trees. It's one of the reasons why UT is slow as hell while Quake 3 is lightning fast.

However, its highly likely that the reason we see KYRO performing better at UT is probably purely due to the way it handles palletised textures; we all know that this is an area Geforce cards handle exceptionally poorly.

Interesting, I didn't think of that. Perhaps the combo of palletised textures and texture-heavy environments is what puts the Kyro2 out ahead. Either way if the scores are correct, the Kyro2 is a winner in a game which has traditionally only taxed the CPU and shown very little difference between the video cards.
 
&quot;if the game does not take advantage of the multitexturing capabilitys of kyro that does not mean bad results with kyro there will be always overdraw, and the brute force aproach also have to render that fog...&quot;

From what I've heard from owners of a KryoII, their biggest complaint was that once smoke/flash grenade was used in CounterStrike, the FPS went to absolutely nothing. Easily below 10FPS on a >1Ghz Athlon. I'm pretty sure CS wouldn't use the multitexturing capabilities of the KryoII, as it was made an eternity ago in the gaming world. I've heard the Radeon also struggles with this same situation, but not as badly as the KryoII because it does have more raw fillrate.

I have yet to see a GF2 or faster NVIDIA card have a problem with the above mentioned situation.

That's what I was basing my post on.

Too bad the KyroII doesn't really OC much. It would be a monster if you could get the same OC's you can get on almost any NVIDIA card.
 
Why does the kyro do so badly on the low end system in the hardware pub review? It gets sub MX numbers on the Celeron II 700 MHz system.

[edit]
Is a lot of the set up for tiling/hsr done in the drivers or something?

As the reviewer over there notes, having a &quot;budget&quot; card do so poorly on a &quot;budget&quot; system makes it a lot less interesting.
 
a friend of mine just replaced a geforce2MX on a 1200 mHz athlon system with a 64MB inno3d kyro II. the maximum framerates are a lot higher on the kyro than they were on the MX, but the minimum has a lot of fluctuation. in quake 3, with the setting that he uses (i'm not totalyl sure what they are, but they are fairly mediumish) the kyro has as much as a 50 fps difference between the hight and the low (from about 80-30 or even lower at 1024x768). the MX, while not getting quite as high (about 65) it never drops below 50 FPS no matter how many players are on a server, or how much action there is on the screen. minumum frame rate is ususally more important than maximum. i'm sure that a geforce2 GTS would be more stable than having 50+ dips several times in a game. these results were with the drivers that came in the box. does anyone know if the newest drivers help this wild fluctuation?

--jacob
 
BFG10K

?Except UT does it in real-time while Quake 3 uses pre-calculated BSP trees. It's one of the reasons why UT is slow as hell while Quake 3 is lightning fast.?

I?d like to read what UT actually does ? do you have any links?

However the point being is that quite a fair chunk of Polygon overdraw will remain in the scene. This can be displayed by looking at Scores for KYROII against a V5 ? V5 has almost double the fillrate and bandwidth of KYROII (and handles Palletized textures correctly), and I think you?ll find the scores between them both are quite similar.

?Perhaps the combo of palletised textures and texture-heavy environments is what puts the Kyro2 out ahead. Either way if the scores are correct, the Kyro2 is a winner in a game which has traditionally only taxed the CPU and shown very little difference between the video cards.?

I think its gradually becoming accepted that KYRO is a damned good choice for UT based engines ? I?ve seen numerous favorable reports of UT, Deus Ex and Rune etc (including one of the recent reviews that?s been flying about on these forums). In light of 3dfx?s demise it looks like KYROII is gradually picking up the UT engine price/performance crown.


Def

? &quot;if the game does not take advantage of the multitexturing capabilitys of kyro that does not mean bad results with kyro there will be always overdraw, and the brute force aproach also have to render that fog...&quot;

From what I've heard from owners of a KryoII, their biggest complaint was that once smoke/flash grenade was used in CounterStrike, the FPS went to absolutely nothing. Easily below 10FPS on a >1Ghz Athlon.?

I didn?t follow who made the initial statement, however in terms of raw performance actually utilizing multitexturing or using multipass rendering makes very little difference. I?ve seen Serious Sam (which uses as much as 5 texture layers) tests on KYROII that show going from no multitexturing (1 texture unit per pass) to Quad texturing (4 textures per pass) only make about a 10% difference in performance ? and that?s at the fillrate/bandwidth limited case of 1600x1200x32.

The actual true benefits of utilizing multitexturing on KYRO are a little more fundamental to the nature of the architecture.


SlimHarpo,

?Is a lot of the set up for tiling/hsr done in the drivers or something??

No, all of this is done 100% in hardware.
 
Funny how powervr2 counters Kyro2 UT scores with a UT score using 640x480x32.
Well if 640x480x32 is what he wants, then here you go, (from that same review:
Kyro2 loses everyone.
Even GF2 MX is twice as fast as Kyro 2, yes &quot;MX&quot; BEATS Kyro 2
Nope, in Quake3, GTS beats Kyro2
Serious Sam, same results, GTS wins.
I see alot about how great multitexture of Kyro2 is, but:
kyro 2 scores 3rd from bottom, only beating V5 and MX.
MBTR, 2nd from the bottom.

Of course not many run at 640x480, but since powervr2 wants numbers 😀
 
&quot;Funny how powervr2 counters Kyro2 UT scores with a UT score using 640x480x32.
Well if 640x480x32 is what he wants, then here you go, (from that same review:
Kyro2 loses everyone.&quot;

Old review, old drivers. I wonder what the situation would be like now!

(although in many cases I would doubt its drastically different in the T&amp;L cases).
 
It is the same review that powervr2 (and many others) are using for reference, so it is a valid review for comparison. You can't pick 1 score out, and dismiss the rest as being old or using outdated drivers.

I think nam ng sums up these reviews nicely in the other thread:

<< Anantech old driver
slowpoked GTS 64MB ---> 50.4FPS
Lightning KyroII -----------> 41.4FPS

Planethardware improved driver
Hella fast GF2_ULTRA -> 40.8FPS
Lightning KyroII -----------> 44.7FPS >>

Note the Ultra is slower than GTS with new drivers!

I do agree that newer drivers can help. But, don't forget it goes both ways. Bumping up to nVidia 1290 from 1260's gave me a nice 10% boost in some benchmarks!🙂
 


<< SlimHarpo,

?Is a lot of the set up for tiling/hsr done in the drivers or something??

No, all of this is done 100% in hardware. >>



Why is the Kyro so slow on the 700 MHz system in the hardware pub review, then?

Non-tilers (GeForces) don't show anywhere near that kind of cpu dependency (except, of course, at low resolutions, which is not an issue with the video card).

The way it looks to me right now, the Kyro might be a good buy for someone with a 1.4 GHz Athlon or P4, but is not a good buy at all for someone with an 800 MHz P3 or Duron (they should just buy an MX for $65 and get similar frame rates, or spend $140 on a GF GTS/PRO, and get much faster frame rates).
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top