That's not likely to happen. With the Prescott problems, I would assume Intel is concentrating on a whole new core/chipset design.
I wonder how a Quad dual-core Opteron would fair against a Intel with it's limited memory bandwidth?
AnandThenMan
Diamond Member
- Nov 11, 2004
- 3,991
- 627
- 126
My thought was that opteron was always the more rounded chip , from LTC8K6 post he hinted that the chip was now more dominant and therefore a better choice over opteron, I just ask if this was the case.
LTC8K6
Lifer
- Mar 10, 2004
- 28,520
- 1,576
- 126
Clarkey01, all I have been asking for is an explanation of Anandtech's benchmarks where the new Xeon crushes the new Opteron in 2 out of the three benchmark sets.
I'll ask again, what happened? Before, as far as I know, there was no contest at all between Xeons and Opterons in any server benches. Opterons ruled without question.
Now there appears to be a question.
I suspected that it was memory bandwidth, so I said that more processors would make a difference. Meaning that the Opteron would win again if more than 2 cpus were used.
Jeff7181 flat out said it was the memory bandwidth, and I agreed that it probably was.
My original post asking about the Opteron's losses in Anandtech's benches and speculating that more CPU's would reverse the results:
I said nothing at all about which chip is "better". Both chips are excellent, and I am glad we have choices or we'd be paying a lot more for CPU's. I am especially glad that Intel seems to be doing better, since I prefer that there be some competiton.
It's a shame when folks automatically assume there's been an attack on the object of their affection. Yes, that's directed at you, clarkey01.
I'll ask again, what happened? Before, as far as I know, there was no contest at all between Xeons and Opterons in any server benches. Opterons ruled without question.
Now there appears to be a question.
I suspected that it was memory bandwidth, so I said that more processors would make a difference. Meaning that the Opteron would win again if more than 2 cpus were used.
Jeff7181 flat out said it was the memory bandwidth, and I agreed that it probably was.
My original post asking about the Opteron's losses in Anandtech's benches and speculating that more CPU's would reverse the results:
I said nothing at all about which chip is "better". Both chips are excellent, and I am glad we have choices or we'd be paying a lot more for CPU's. I am especially glad that Intel seems to be doing better, since I prefer that there be some competiton.
It's a shame when folks automatically assume there's been an attack on the object of their affection. Yes, that's directed at you, clarkey01.
- Aug 22, 2001
- 32,033
- 32,510
- 146
Couple questions for you tech doc lovers.
1. How do the new Xeon and Opteron compare in heat and power?
2. Is there potential for clock throttling with the Xeon under extended heavy load?
1. How do the new Xeon and Opteron compare in heat and power?
2. Is there potential for clock throttling with the Xeon under extended heavy load?
VirtualLarry
No Lifer
- Aug 25, 2001
- 56,587
- 10,225
- 126
Don't the new Xeons have some sort of improved hardware-prefetch logic in them? Could that explain the benchmark results, perhaps the actions of those benchmarks are a good fit to those sorts of improvements?
VirtualLarry
No Lifer
- Aug 25, 2001
- 56,587
- 10,225
- 126
Yes, it more-or-less does, usually the memory controller has to be designed to support either/both (I think - I would assume that it would have to delay some of the timings by one clock-tick to deal with interfacing with the registers on the DIMM boards.)
It's kind of interesting how there is a dichotomy between "unbuffered" and "registered" memory, but there is no "buffered + registered", or just "buffered" DIMMs. However, there are motherboards that add their own buffers (not registers) to the memory control signals for the DRAM array. The Abit BX6-r2 was one such board; I owned one at one point. Abit added an additional set of "buffer chips" on the board, six chips in all, between the chipset's memory-controller and the DIMM slots, in order to stablize the signals to be able to load them down with four double-sided SDRAM DIMMs.
I think that it would generally not be cost-effective to add those buffers to the individual DIMMs, and since current SDR/DDR is based on syncronous memory control signals, registers are a more appropriate choice. Back in the days of EDO/FPM DRAM, I'm sure that there probably were "buffered" memory sticks intended for servers, although I've not really seen any myself.
Intel's "FB-DIMM" scheme is essentially some sort of cross between the two, I guess.
What on that page makes you think that buffered and registered are different things. All I could see was that they said they supported "unbuffered, ECC, and registered" RAM modules. Its ECC and unbuffered that are not mutually exclusive to each other, not registered and unbuffered.
- Aug 22, 2001
- 32,033
- 32,510
- 146
- Aug 22, 2001
- 32,033
- 32,510
- 146
That makes sense. It seems doubtful the power&heat of the new Nocona is any better than the last, and that combined with the added expense of keeping the ambient room temp the same would make the cost of even the same number higher correct?
By then, whats the point? Who cares if Intel is faster if you can get more done with AMD's with a lower power bill? Less power used, less heat generated, less air conditioning needed in the server room. 
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 25K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes + WCL Discussion Threads
- Started by Tigerick
- Replies: 22K
-
News NVIDIA and Intel to Develop AI Infrastructure and Personal Computing Products- Started by poke01
- Replies: 384
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 23K
Facebook
Twitter