Tenchusatsu - probably not for the last time :-)
Sorry if it wasn't clear - OK DDR 133 is only 166% faster than PC100, compared to Rambus 800's being 33% faster than Rambus 600 (or having 133% of the bandwidth)
- 166 to 33 percentage increase - is that OK? (you're right, at least once I mixed the two measures)
The benchmarks are the ones I've seen. I expect rambus to do very well on memtest also (actually, I expect it to do better than DDR 266) but these benchmarks seem to show that you don't need the cost and complexity of rambus to get big increases in raw memory throughput. The benchmarks also show that big increases in raw memory throughput make very little difference in most standard applications - which makes it hard to justify memory substantially more expensive than PC100.
>all you're doing is repeating your own arguments
I'm reposting thoroughly documented tests - where are the rambus tests?
>PC133 has shown to have no benefit in real-world ...
In CAS 2 100 vs CAS 3 133 it doesn't - but that's a rigged test.
I doubt that latency 50 rambus 800 will do as well as latency 40 rambus 600 in real world tests either. Please post the tests for that comparison if you want to reference the samsung test. Hey, just post any test of rambus :-)
>memory controller design hasn't matured yet
Or something, but just wait until... To me, that's the issue with rambus - it was conceived when it looked as though 100MHZ was it for DRAM - it had been a rough transition from 66 to 100. But that was a long time ago, and the SDRAM variations have proven to have had more potential than was expected. A lot of time has gone by since the rambus specs were established. The world has changed in the meantime.
The EV6 bus is designed to eventually run at up to 400MHZ at double data rate, just like rambus - and it's 64 bits wide instead of 16 bits. Sure, someday rambus will get its latency down, but the rest of the world won't have stood still meanwhile.
>Why is AMD saying Athlon will move towards DDR SDRAM
Because for VC DRAM, there's nothing to "move towards". VC DRAM at 100 and 133 MHZ is already supported by the new VIA chipset. This is possibly why Intel was so hysterical about VIA. VIA's support of VC DRAM 133 in the same time frame as rambus's initial release makes it difficult to trumpet rambus as a performance leader - and it certainly doesn't look as though it's going to be a price leader. VC is already here, you can buy it now, and the PC motherboards that use it will ship just before, or just after, the rambus motherboards. Meanwhile DDR is Q1 00, and VC burst technology is applicable to DDR (and rambus - but rambus would have to give up some or all of their royalties)
==============================================
via.com.tw
Apollo KX133 Features
VT8371 North Bridge
200MHz Alpha EV6 FSB for the AMD Athlon processor
66/100/133MHz memory bus settings
Supports PC133, PC100, VC SDRAM, and EDO DRAM types in mixed combinations up to a total of 2.0GB
AGP 2X/4X
ATA-33/66
Concurrent CPU and AGP access
AC-97 link for audio/modem
4 USB ports, UHCI compliant
Integrated IO/APIC
Hardware monitoring
Advanced mobile power management and clock stop features
PC98/99 compliant
516 BGA VT8371 North Bridge
352 BGA VT82C686A South Bridge
===============================================
>Dan, just wait. I'm sure at the Intel Developer's Forum later this month, there will be official benchmarks for DRDRAM.
There had better be, I think a lot of people are tired of waiting.
Look, I'm not trying to say that rambus is "dead". Intel will make sure that it has some future in PC's, at least for awhile. Rambus has captured the video game market, is making good progress in video cards, and as I said in a recent message - how about HDTV? -
But I think that the near term prospects for rambus hinge upon this single issue: to what degree is Intel willing to lock itself into a rambus only strategy. Given Intel's enormous rambus investment, it has much to gain by sticking with rambus, and much to lose if it doesn't. What I've been trying to show is that, for price/performance reasons, this isn't the certain outcome that some people think it is. The big question for discussion IMHO: is it worth it for Intel to exclude everything but rambus in all future products?
I'd like to see more discussion about just how much Intel stands to lose if it accepts alternatives to rambus. If the loss is too great, then intel sticks with rambus and intel's market power will insure that rambus does very well - even in the near term. The discussion about rambus's huge performance benefit that guarantees market acceptance at any price - that's what I don't think makes sense.
Regards,
Dan |
