Tom's Guide | Tom's Hardware | Tom's Games
 |
 |
 |
Whats the big diffrence between the Athlon t-bird and the athlon xp 1900? Is there a big diffrence in gaming performance?
this page calculates the difference between a tbird ghz and an xp ghz
http://www.overclockers.com.au/techstuff/a_qs_calc/
This page overviews the xp design benefits. The xp is a significant step up from the tbird.
http://www.1steweb.net/web/merchant.mv?Screen=CTGY&Store_Code=eWeb&Category_Code=PROC-AMD-XP
this page shows a quake comparison between the athlon 1400 and athlon 1900+ as being about 40%.
http://www.tomshardware.com/cpu/20030217/cpu_charts-22.html
0 
Johnoh, dont forget that the result for Athlon XP 1900+ was using DDR RAM vs the Athlon 1400C was using SDRAM. If both of them were running at same clock speed. Assume using Athlon 1533Mhz vs Athlon XP 1800+ @ 1533Mhz. They will run roughly the same using both SDRAM or DDR RAM. The XP will out run Athlon 3-10% due to SSE optimization.
0
The fastest athlon was the 1400 so an apples to apples ghz comparison would be the athlon 1400 to the xp 1600+ (which is 1400mhz)
In the above link the 1600+ is faster by:
quake3: 30%
3dmark: 21%
Unreal Tournament: 25%
pcmark cpu: 18%But as you point out, some of these differences have to be due to ddr vs sdram. So instead of 18%-30% what is it? I don't know, but I think its more than 3%-10%.
Amd claimed the palomino was 5%-25% faster than the tbird for several reasons. One was the SSE instructions. The others were:
hardware data prefetching - puts data in l2 cache before its needed (this is why a non-dualled athlon 1200 MP sometimes outperformed an athlon 1400 - the athlon MP had data prefetching which the athlon didn't)
bigger TLB - puts translated main memory addresses in l1 & l2 cache before they are needed
pipelining improvements - results in more instructions per clock cycle than the tbird
add to those 3 things the pipelined/superscalar floating-point unit (which was actually already on the athlon) and you have the 4 components of the xp quantispeed architecture.
Which is why amd named the 1400mhz amd xp a "1600+". Because on average it was as fast as an athlon would be if it was clocked at 1600mhz.
0
What? This I've never heard:
"Which is why amd named the 1400mhz amd xp a "1600+". Because on average it was as fast as an athlon would be if it was clocked at 1600mhz."
I thought the 1600+ was named to show equality with the P4 running at 1.6 Ghz, not in comparison to an Athlon Thunderbird.
The main reason I believe they switched to the XP and value+ designation was due to marketing. Otherwise, they would have stuck with the standard Athlon designation, along with the mhz rating, not the value+ crap. The Athlon and Athlon XP are nearly the same chip. I lost a lot of respect for AMD when they went to this new labeling system.
0
The xp was the first cpu with quantispeed architecture, which was why the xp is faster than the tbird at equal clock speed.
However since the fpu was already in the tbird and the tbird was already at 9 operations per clock cycle (compared to the p4 which does 6), quantispeed was partly in the tbird already. SSE and data prefetch and tlb and some unknown amount of pipelining improvements differentiate the athlon from the tbird.
In amd's "True Performance Initiative" where they outline the rationale for the "1800+" and "1600+", etc, they say
"The 1800+ operates with the performance of a 1.8mhz cpu."
which leaves you asking "which 1.8mhz cpu?". Amd up until then only made a 1.4ghz cpu, the athlon. I cannot find reference to amd ever saying which cpu they were talking about. However,
in that same document they have charts that showing the performance of an athlon 1400, xp 1500+, xp 1600+, xp 1700+, etc. In these charts there is perfect scaling from the 1400 to the 1500+ to the 1600+ to the 1700+. So I conclude that these charts mean that the new model numbers are based on the athlon 1400. They also answer the question "why is a 1.4ghz amd xp (1600+) faster than the athlon 1400?"
That same document shows comparisons of the 1800+ versus various p4 cpus. The 1800+ is faster than a p4 2.0g in these charts, about equal to a p4 2.2g. From this I conclude that the new model numbers are not based on the p4.
If anyone can find a more definitive reference to the model ratings please post it. There are many internet references to how a 2400+ is of course meant to go against the 2.4g p4, but the orginial amd statement appears to base them on the athlon
0
okay I dug around a little more on this.
Looking at benchmarks from the first amd xp versus the then-current p4s, its clear the model numbers could not have been based on the p4. The palomino 1600+ or 1800+ creams a willamette p4 1.6g or 1.8g consistently. But amd was perhaps smart to never answer "which 1.8ghz cpu?" regarding what their model ratings meant, because who knew what intel had up their sleeve.
Note that the first palomino, the 1800+, had 384kb of cpu cache (*) and ran at 266fsb. The p4 2.0g willamette at that time had 256kb cache and a 400fsb.
Since that time intel has doubled their fsb (and tripled it on an overclocked 875 chipset) and doubled their cache, while amd has only had an fsb increase of 50% (65% on an oc'd nforce2) and a cache increase of 66% (640kb barton vs 384kb tbred/palomino)
So intel's drastic fsb and cache improvements got them to parity with the amd model numbers, which is perhaps why everyone assumes that the amd numbers are meant as p4 equivalents.
* - amd l1 and l2 cache is exclusive, meaning additive. 128kb of l1 + 256kb of l2 = 384kb total. Intel uses inclusive cache, where everything in l1 is mirrored in l2, to the total cache is just the l2 total.
0  |
 |
 |
This post is quite old and has been locked from receiving new replies. Please create a new posting instead.
| Ads by Google |
