AMD TBred - not as bad as made out

Matt Simis

Hi All,


For all those considering the TBred (aka Athlon XP 2200)*:

I finally got the Tbred from Komplett.ie (€286). Its the 2200 model and runs at 1800MHz stock speed. First run I used a 156W Peltier and a decent Water Cooling System. CPU temp at stock speed and voltage (measured from back of CPU) is 14C.

Max stable speed system runs at is 2060Mhz (1.85V). CPU temp (after 6hrs running) is 32C. It never rises above that, 32 is the high point.


At 2060MHz (almost XP2600 I think) it gets close to the performance of my P4 at 3GHzc at a few things. I'll do a voltage mod to the board (Abit KX7-333) later and see can I get a stable 2100Mhz out of it.



Matt


*Not all Athlon XP 2200s are TBreds, a small number are Palominos. If you see ".13 micron" beside the name, then its a Tbred. ".18" is a Palomino. Soon the entire AthlonXP range will be Tbred based, but they are phasing them in slowly.

Gerry

Nice one, could you post up some benchmarks of games performance?

Bear in mind matt, that the ddr ram of your p4 machine is holding it back a bit.

Matt Simis

I will post some benches... any suggestions?

I already ran the "Four" bench in Q3, 226FPS with all details on and 1280x1024 Res, up from 160 on a 1911MHz Palomino on an AMD760 based board.

It gets something like 350FPs on the fastest (512x384 res) settings, but the P4 system gets 490FPS. When I say its close to the P4, I meant in things the Athlon is historically good at, ie Sandra CPU benches. My P4 has the ram at 178MHz, which is pretty fast. I tested the RAM in it upto 206MHz, got over 3000MB/s in Sandra. Ill be getting better cooling for the P4 soon and aiming for 3200MHz.




Matt

JustHalf

I think when Gerry tested out demo four with Quake III on my machine, lowest settings, he got around 700FPS. (he'll correct me if I'm wrong).

This was a P4 running at 2.8GHz with a 140MHz FSB, with RDRAM running at 1120MHz. GeForce 3.

Gerry

I think that was with a high fps config though..

flamegrill

Originally posted by JustHalf
I think when Gerry tested out demo four with Quake III on my machine, lowest settings, he got around 700FPS. (he'll correct me if I'm wrong).

This was a P4 running at 2.8GHz with a 140MHz FSB, with RDRAM running at 1120MHz. GeForce 3.

All i can say is Bloody hell!! 700 ffps hehehe, how unplayable would that be. Also not recommended for people who suffer from epilepsy.

phaxx

Originally posted by flamegrill
All i can say is Bloody hell!! 700 ffps hehehe, how unplayable would that be. Also not recommended for people who suffer from epilepsy.

*rolls eyes*

It will be perfectly playable. :P

saik

right. as we all know an amd760 board is not the fastest.

i have one of them
and a 1400 thunderbird athlon.
and 256mb 2.5 cas ddr pc2100
and a gf3ti200, that clocks pretty high.

how much speed to i get if i change to the current king of the athlon motherboard hill combined with mad ddr ram, and would keeping the cl2.5 ddr be a big fps drop?

//all hail matt simis, im well impressed.

Matt Simis

Id say youd get about 10% increase across the board by changing to a KX333 based board. Some reviews show huge increases, other not so much. Still, 10% with the same equipment is pretty good. I wouldnt worry about the DDR too much, as Im sure you could run it a CL2 or perhaps with a bit more voltage CL2.5 166MHz. Problem with the AMD platform tho is the FSB is limited to 266MHz, negating the benefits of 333MHz DDR. Of course you could overclock the FSB too...


Im still highly doubtfull of the usefulness of the "four" benchmark. Turning off of all details and res creates an environment that no one would ever use, scoring high here is not representive of realworld performance. Ideally, the same GFX card in different systems, but running at high res is a better benchmark for tangible performance benefits.


From memory, the AMD XP2600 got 226FPS at highest quality, the P4 3000 got 240+FPS or so. Not the 2x the speed as the low quality/res settings would suggest (JH, its quite unlikey your systems would even score 30% more, let alone 300% as your 700FPS suggests). Ill have to confirm those numbers, but the difference wasnt dramatic.

These systems had the same GFX card, a TI4200. The "four" benchmark seems to be a poster child for the [High Latency] High Memory bandwidth situation: little performance benefit in real world situations (higher quality settings), despite massively higher scores in "controlled" tests. Its certainly not suitable for CPU speed comparisons. I only put it in as I knew it would be the first thing asked for!

I think its one of those benchmarks where everyone can see the result they "want" to see.*



Matt


*"Phfft! Facts. You can use them to prove anything. - Homer Simpson

JustHalf

It's the same config Gerry uses, if I remember correctly.

Not one I'd use though

The problem is, once the resolution goes up to 1024x768, full detail, you're graphics card limited (at least with everything up ot the GeForce4's). So *they* don't really test out the CPU/board all that well either.

Gerry

Erm matt, running four with low detail tests the cpu and motherboard. Surely thats what we want to test here?
Yes, quake3 is a posterchild for high bandwidth, but the high bandwidth of the p4 with rdram means that quake3 can get the absolute max out of the cpu. The cpu is still the limiting factor, if the p4 cpu wasn't capable of driving quake3 at 500 fps, no amount of bandwidth will make it do so.

It is indicative of realworld performance if you actually play quake3, as I do. If you use the same graphics card, and it is the limiting factor, it is masking the difference between the machines.
If you want to try a real world application which will always need full cpu power, try encoding divx, or a 3d studio benchmark.

Matt Simis

Well, thats my point about taking multiple "facts" from the Four benchmark.

I dont think they are truely limited by the graphics cards anymore.. since the P4 did score higher than the AMD. Surely if they were limited then all of the following systems would score roughly the same:

All the same amount of RAM, WinXP and GF4 Ti4200.
AMD760, PC2600 RAM, 1911MHz PC Athlon: 160fps
KX333, PC2700 RAM, 2060Mhz Athlon: 226fps
845E, PC2900 RAM (oc'ed), 3000MHz P4: 240fps


As you can see there is a huge range between the systems, so I dont think you can say that the graphics card is really the "limiting" factor here. Im merely suggesting we be wary of other factors.... imagine if the high latency of RDRAM caused it to choke on high res and high detail. Its all well and good sending super simple (and small) textures around the motherboard, but if it cant actaully perform in a more real world situation, then its not "really" faster at all is it? Of course thats just a made up example, Im not suggesting RDRAM (or anything for that matter) does perform like that, its just that if it did, our logic here of testing minimalist settings would never uncover such a problem.

For the record, I do play Q3 and Q3 based games and was testing CPU performance (only) as I have done nothing at all to set up the GFX or memory in an optimum manner. I will be doing that tonight..

The end result with my primary PC (the P4) is to have Prometeia cooling on it with the FSB and memory at 190MHz (3230MHz CPU speed) and at these speeds it should give PC1066 a run for its money!



Matt

spooky donkey

Would you guys not also concider using sysmark 2001 SE to test out youre different systems. I know this is mainly a test of youre video card but if you use the same card in the 2 (eg. the TI 4200, in the XP2200 and p4 3G. ) machines will it not give some intresting results as well.

Praetorian

Where did you get the water cooling equipment Matt?

Id love to have silent cooling.... and thats why my xp1800 is not overclocked at all...I do not fancy the idea of having a jet engine in my Case!

This Xmas im gonna treat myself to the fastest AMD chip available and a nice water cooling system

(ATM There isnt enough of a jump in speed to a XP2200 IMO)

I must remember to talk to Matt for advice in December

Gerry

Yes matt, my point was that it is about cpu performance. About quake3, it thrives on bandwidth, but also on latency. So I don't think rdram gives the p4 an unfair advantage in this benchmark.
If you look at tomhardware.com, in a test where you have very fast ddr ram (sayt pc400) against pc1066, in the games benchmarks they are very close, but the rdram seems to have a definite advantage in the "real world" professional opengl spec viewperf benchmarks, indeed sometimes pc800 rdram bests the fastest ddr.

Matt Simis

Id have to admit Gerry, if i didnt change (sometimes break!) motherboards and other stuff so much, id have 1066 RDRAM in my system, no doubts. Its just too expensive for the me the way I go through things!

I do think however, that the reviews on the web give a poor impression of DDR on the P4, as some of them seem to insist running P4 boards at Intel's rated speeds (crappy 266 on a 845E, and maybe 333 on an 845G). At the moment, im temped to get the SIS648 for DDR400 support or hold off for cheap RDRAM 1066 (or 1200 by then?) mobos from SIS or Intel. However, the Prometeia (its like a VapoChill) costs €600+, so Im not going to be buying much other stuff anytime soon.


Praetorian, I get my Watercooling stuff from OverclockersHideout.com and got a couple of pumps from PetStop too (seriously). My systems are anything but silent though, Watercooling still needs fans and with peltiers and high speeds, big fans. Still big fans dont whine like smaller ones, not as annoying.

spooky donkey, SySmark or 3DMark? Your comments about video cards suggest 3DMark. IN 3dMark, both my systems seem seriously hampered by the GF4 4200 64MB, only scoring 11K or so. I have friends with slower systems and Ti4400s getting 12K!!



Matt