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.

Politics : Formerly About Advanced Micro Devices -- Ignore unavailable to you. Want to Upgrade?

To: Petz who wrote (37013)	9/14/1998 3:58:00 AM
From: Tenchusatsu	Read Replies (1) \| Respond to of 1570420

3)the K6-3 L2 cache of 256K will experience 30% more L2 cache "misses" than the 512K cache of the typical Super Socket 7 motherboard. Actually, according to Microprocessor Report, the general rule of thumb is that if you double the cache size, you reduce the cache miss rate by the square root of 2. So a better estimation would be that the 256K cache would experience 41% more L2 cache misses than a 512K cache. This is just a nitpick. 4)the penalty for an L2 cache "miss" will be 20% lower than with the K6-2, because 50% of these L2 cache misses will be satisfied by the L3 cache on the Socket 7 motherboard. That's pretty unlikely as well. The principle of locality will be covered by the on-die 256K L2 cache of the AMD K6-3. In other words, any L2 cache miss is likely going to be so random that there won't be very many L3 cache hits on the motherboard unless the L3 cache was very large. Or, here's another way of putting it. The AMD K6-3 has 32K+32K of L1 cache and a 256K on-die L2 cache. The factor between the cache levels is four. How in the world is the L3 cache on the motherboard going to make any difference unless its size was increased to at least 1 MB? Of course, this is another nitpick as well. All AMD has to do is recommend that the K6-3 be installed on motherboards with at least 1 MB of L3 cache. An extra 512K of L3 cache should cost only $45. The bottom line, a K6-2-400 will just equal a P2-350, but a K6-3-400 will equal a P2-500 while a K6-3-450 will definitely outclass it. Pretty interesting prediction. You think that the K6-3 will be ahead of the K6-2 by three speed grades, and ahead of the Pentium II by two speed grades. I predict that the K6-3 will be ahead of the Pentium II by only one speed grade; i.e. a K6-3 450 MHz should perform just as well as a Pentium II 500 MHz. That still should worry Intel, if AMD can deliver on-time. Tenchusatsu

To: Petz who wrote (37013)	9/14/1998 7:43:00 AM
From: Maxwell	Read Replies (1) \| Respond to of 1570420

John: Your calculations look good. Let's wait till some benchmark is done to see how close your model fits the actual data. My guess is similar to Tenchusatsu that the K6-3-400 is more like Intel 1 speed grade higher, PII-450MHz. 3DNOW SIMD will have an incredible performance in K6-3 though. When the K6-3 hits the street the slot1 will look bad. Maxwell

To: Petz who wrote (37013)	9/14/1998 11:27:00 AM
From: Scumbria	Respond to of 1570420

The bottom line, a K6-2-400 will just equal a P2-350, but a K6-3-400 will equal a P2-500 John, You are suggesting that addition of an onboard L2 is as effective as a clock speed increase of 40% (350MHz-500MHz.) I find that rather difficult to accept. Scumbria

To: Petz who wrote (37013)	9/16/1998 6:52:00 PM
From: Jim McMannis	Read Replies (5) \| Respond to of 1570420

Petz, How do these benchmarks for the K6-2-450 jive with your projections? processor.org K6-2 350Mhz at 450Mhz: we've done it!!! Oh happy day! I just got a K6-2 350Mhz unit and plugged it in my computer. I wasn't expecting much from this chip, because I already had a K6-2 at 112x3 (which is faster than 100x3.5 for most apps anyway). However, I was wrong - this chip turned out to be a pleasant surprise! First of all, let me tell you that I didn't use any special cooling devices. No Kryotech was needed to overclock this little chip (now that makes me wonder how much I would've got with a Kryotech cooler). All I'm using is my faithful 12-buck huge Vantec fan/heatsink combo, which was the same one used for pushing a K6-2 300Mhz unit to 112x3.5 (392Mhz) a few months ago! At 112x4 (448Mhz), this chip is faster than Celeron A at 100x4.5! Both were using the same configuration, ie. 128Mb of PC100 sdram, a 4.3Gb UDMA HDD, a 12Mb Diamond Monster II and an old 4Mb Diamond 3000 PCI, except for the fact that the Celeron A was running on a shiny new Abit BH6 motherboard, and the K6-2 was running at my faithful Soyo 5EH/M. Here are the scores: CPU Business Winstone 98: K6-2 450Mhz (112x4) 31.0 K6-2 450Mhz (100x4.5) 29.1 Celeron(A) 450Mhz (100x4.5) 29.3 What's interesting about these scores is that the slight advantage of the Celeron at equaly settings is gone when we set the K6-2 at 112x4, thanks to the flexibility of being able to switch the multipliers in the K6-2. Now here are the 3DWinbench scores: CPU 3D Winbench 98: K6-2 450Mhz (112x4) 1367 K6-2 450Mhz (100x4.5) 1329 Celeron(A) 450Mhz (100x4.5) 997 As you can see, with the 3DNow! support, the K6-2 at 112x4 (once again, thanks to the flexibility of the non-locked multipliers) can be up to 37% faster than the Celeron A under DirectX (keep in mind that DirectX 6.0 is required for 3DNow! support). This clearly shows us how 3DNow! can make quite a difference in the K6-2's otherwise weaker 3D performance. At the same settings, the K6-2 450Mhz was still more than 33% faster with 3DNow! support. We will soon be publishing the scores under Quake II and Unreal, as well as more business-oriented benchmarks! Please note that 450Mhz with the K6-2 350Mhz is not an almost-certain overclock like the Celeron 300A at 450Mhz is, however, most of the K6-2 350Mhz we've seen had no problems reaching up to 400Mhz! As you can see, the Celeron A currently remains the best bang for the buck, but the K6-2 is still a strong player! Jim