Silicon Investor (SI) -- The First Internet Community

STOCKTALK

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor. We ask that you disable ad blocking while on Silicon Investor in the best interests of our community. If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.

Technology Stocks : Advanced Micro Devices - Moderated (AMD) -- Ignore unavailable to you. Want to Upgrade?

To: Joe NYC who wrote (85823)	7/25/2002 9:19:05 PM
From: Petz	Read Replies (1) \| Respond to of 275872

Would a 2GHz TBred be slower or faster than a 2.4/533 NW? I thought it would be worthwhile to do a careful examination of this question. You found TBred 2.0 GHz (estimated) beating NW 2.4/533/1066RDRAM on only 10 out of 27 benchmarks as a "rough guess." I have to disagree on On Sysmark 2002 Office, TB 1.8 and 1.6 have scores of 165 and 154, respectively. I don't see any chance that 2.0 would not get to 173, the score fot the NW 2.4. Whether you extrapolate from 1600, 1666 or 1733 MHz, the result is better than NW 2.4. In any event, the performance scales horribly (10%) for the P4 on this benchmark. A 10% performance improvement here would be at least a GigaP4Hz. I also see TBred 2GHz tieing in SpecviewPerf DRV08 and winning in PROE01. On DRV08, the extrapolated score is either 40.84+3(40.84-39.44)=43.64 40.84+1.5(40.84-39.01)=43.585 40.84+(40.84-39.03)=42.65 The NW 2.4/533 actual score is 43.17, smack dab in the middle, call it a tie. In PROE08, the 2200+ (1.8 GHz) is already at 99.4% of NW 2.4/533 performance. Its pretty clear that a 1.86 or 1.93 TB would be superior. On PC Mark 2002, I also came to the opposite conclusion as yours. The score extrapolating using 1.8 and 1.73 GHz is 5412+3(5412-5223), or 5979. Using 1.8 and 1.66 GHz we have 5412+1.5(5412-5040), or 5971.5. Using 1.8 and 1.60 we have 5412+(5412-4857), or 5967. All three are well above the NW 2.4/533 score of 5928. With these changes, a 2 GHz TBred wins on 13 of 27 benchmarks (P4 scalability index in parentheses): 3DMark2000 Pro (60%) XMPeg4.5 (81%) Sysmark 2002 Office (10%) LameMP3 (88%) Cinema 4D XL R7 (102%) 3D Studio Max (74%) SPECViewperf DX07 (32%) SPECViewperf LIGHT05 (69%) SPECViewperf PROE01 (41%) SPECViewperf UGS01 (25%) Sandra CPU (57%) Sandra Multimedia (100%)* PC Mark 2002 (96%)* Average P4 scalability excluding * = 58% The NW 2.4/533 wins on 13 of 27: Quake (55%, 34%, 63%) 3DMark2001 Pro (60%) Newtek Lightwave (94%) Sysmark 2002 (60%) Sysmark 2002InterCC (102%) MP3 Maker (78%) WinACE (53%) Comanche 4 (74%) Pinnacle Studio 7 (214% ?)* Sandra Memory (5%)* PC Mark Memory (36%)* Average P4 scalability excluding * = 67% A 2 GHz Clawhammer won't have any trouble with the benchmarks in the first list, even with only a 20% IPC improvement over TBred. That 20% IPC would be equivalent to 40% higher clock speed than the 2.4 GHz P4, equivalent to a 3.36 GHz P4. I'm still a disbeliever in HT and believe most people will disable it until software is rewritten to take advantage. But the benchmarks in the second list have a pretty high scalability index for the P4, and the 2 GHz TBred is only at about a 2.2 GHz P4/533 level. Maybe Intel can get an average 55% scalability going from 2.4 GHz to 3.4 GHz using PC1066 RDRAM. Then, for the second list, a 2 GHz Clawhammer would be equivalent to a 2.2 + 2.2*(0.2 to 0.25)/0.55. This gives a range of PR 3000 to PR 3200, which is right where Intel will be. Maybe a 2 GHz Clawhammer with PC2700 would only win 60-75% of the benchmarks against an NW 3.066/533/PC1066 system. (nearly all the benchmarks in the first list + 1/3 to 1/2 of the second list) Petz