BARTON 3200+ PREVIEW

[paralazarguer]

Nice i didnt know the Nforce2 supported 200Mhz FSB.

So I just rebooted went into the BIOS and lowered the multiplier and raised the FSB to 200.

Processor x86 Family 6 Model 8 Stepping 0 AuthenticAMD ~1904 Mhz

Not to bad for cheap little 1800+ 9.5x200

I dont know if my PC2700 is gonna like running at this speed though. I relaxed the timings to CAS3 just to make sure it wasnt the RAM that would be causing any crashes.

Time to sell my RAM and buy some PC3200

Right on buddy! Woot!

 

[Rand]

Likewise, there may be situations where the 200 MHz FSB may cause some performance degradation regardless of the "interface setting".

The question is WHY did they see a performance degredation it certain tests with a faster FSB, worst case scenario they should perform identically. I really wish they would have attempted to do some sort of testing to see what the issue was...chipset latency increasing to attain stability at 200MHz FSB... what?

I find it mildly interesting that HT caused a performance hit in most of their tests.

 

[paralazarguer]

Maybe because they were really running a 394Mhz bus and therefore not really 2.2ghz but only close while the other chip was actually 2.2ghz.

 

[Rand]

Originally posted by: paralazarguer
Maybe because they were really running a 394Mhz bus and therefore not really 2.2ghz but only close while the other chip was actually 2.2ghz.

It crossed my mind, but they should be running at virtually identical clockspeeds,,, and the incrementaly difference wouldnt be nearly enough to account for the performance difference shown even assuming performance scales perfectly with clockspeed.

197x11= 2,167MHz
13x166.66=2,166.58MHz

Clearly 0.42MHz isnt going to amount to anything even if performance scales 100% linearly with clock frequency, even if it did the 200 (197)MHz FSB system should have performed ever so incrementally faster.... not slower.

 

[paralazarguer]

They say that they had trouble running some benchmarks under optimal settings.

If not further indicated, it means that the benchmark was run at the "Optimal" setting.

For the record, in some of the benchmarks, the resulting scores are included, in those cases, we differentiate between the (opt) and (agg) setting.

 

[UlricT]

This post is just to prove that paralazarguers is wrong... he went into a post which HE reposted, and called it a repost of this thread, mildly insulting the person who posted it! Didnt think that was very nice sooo.......

Thread Title: Athlon XP 3200+.....
Created On 03/03/2003 3:18 AM

paralazarguers post

Thread Title: BARTON 3200+ PREVIEW
Created On 03/03/2003 9:52 PM

now who has done the repost????????

 

[paralazarguer]

In looking at the benches more carefully I'm not even sure that I see the degredation that you're referring to. Would you mind elaborating?
If you're referring to this page it looks like they shot the lowest settings in the chart with the P4 and then list the agressive settings at the bottom under it. This is all I can find with any degredation.

 

[paralazarguer]

Okay sorry, I see the paragraph that you're talking about now. I'm not sure why that is happening. I can try to find out.

Interestingly, we were not able to get anything higher than a 377 fps score in AWadvs04, compared to some 420 fps at the 333 MHz speed. Fixing the AGP speed to 66 MHz (which we belive is the auto setting anyway) did not change any of the scores either. In other words, the "aggressive" setting may have a few hidden surprises, not only with respect to stability but also with respect to performance. Likewise, there may be situations where the 200 MHz FSB may cause some performance degradation regardless of the "interface setting".

 

[paralazarguer]

Suffice it to say that any asynchronism between the processor and the memory bus causes severe penalties because of the additional latencies caused by the interspersing of fifo buffers to bring everything back to one speed.

Something that Mr Fiffy pointed out. What does that mean, anyway? Isn't that B.S? Doesn't make sense to me...

 

[chizow]

Originally posted by: paralazarguer

Suffice it to say that any asynchronism between the processor and the memory bus causes severe penalties because of the additional latencies caused by the interspersing of fifo buffers to bring everything back to one speed.

Something that Mr Fiffy pointed out. What does that mean, anyway? Isn't that B.S? Doesn't make sense to me...

That's actually just a reference to running mem and FSB async, so all that jargon is just a fancy way of saying "bottleneck."

As for the degradation in the SPEC viewperf tests, I'd attribute it to the chipset more than anything else, which is why I didn't necessarily agree that the next Barton would be 400 MHz FSB. A 400MHz Barton is on AMD's roadmap, but again, chipset support has to be there. Its not really surprising that there is a performance degradation at DDR 400, as others have had issues with high FSB frequencies. Could just be a driver/BIOS issue, or it could be an actual hardware limitation (chipset or voltage).

Chiz

 

[paralazarguer]

That's actually just a reference to running mem and FSB async, so all that jargon is just a fancy way of saying "bottleneck."

Of course, If I would have read it more carefully I would have known that. For some reason I was thinking that it was a comment about them running synchronously not async...it's getting too late for this.

 

[zephyrprime]

The performance drop that resulted by turning on aggressive timings is weird. I would definitely like to get to the bottom of that. They should see if the motherboard they were using exhibits the same problem when running with pc2700 mem instead of pc3200 so the problem can be isolated to the chipset or memory or whatever.

 

[SeybaNumberTwo]

Hahaha, that review reminds me of when VIA released their Hyperion 4in1s.




HYPERION ==================================================14000 3dmark2k1
4in1s

old 4.31 ==========.............................................................................................13800 3dmark2k1
4in1s

 

[paralazarguer]

Maybe memory errors at DDR 400 agressive settings (cas2) ?????