It is disabled in the cpu. Would be silly for Intel to have a huge launch of HT even though all the current cpus already run it? It is known that HT is 'inherant' in all P4's but just disabled till the 3.06
Actually Randum, 2 128MB sticks of PC2100 Crucial will cost $90 at Crucial ($45 each, I just checked last night). But, I bet that GB/Sis 655 boards will be at least $30 (prolly closer to $50-60) more expensive tyhan 850e boards with similar featureset/manufacturer
Newegg has 256mb sticks for $74 including shipping. Total for two sticks and 512mb total is $150.Link
This is about the price of ONE 256mb stick of Samsung PC1066 RDRAM. On pricing, I would bet that the SIS 655 would be priced at about the same price-point as current 850E mobos and then come down once a steady supply has been established.
Current DDR platforms have absolutely, positively, beyond a shadow of any doubt, NO latency advantage in real world system performance over pre-fetching i850e / PC-1066 systems. Get over the rambus/latency fallacys Ilmater...they are non-existant!
We are not talking about current DDR platforms here. We are talking about the up and coming Granite Bay, which will undoubtedly have all the tricks of the 850E such as buffering and pre-fetching applied to a Dual DDR system.
This is from the AT FAQ:
The main drawback of RDRAM is its high latency. However, today's chipsets such as the Intel i850 employ buffering, prefetch and other techniques to effectively "hide" the latency and reduce its performance penalty. Each additional RDRAM device on the bus increases the bus length and therefore latency. Thus from a performance standpoint, purchasing an 8-device RIMM would be preferential to purchasing a 16-device RIMM.
DDR does lose latency compared to its frequency as higher speeds are achieved such as DDR333 and 400 but I would attribute this to it running asynchronously with the FSB. Dual DDR GB will be running synchronously when using PC2100. Even the nforce2 for the AMD platform runs best when running memory synchronously with the FSB (best balance of latency/bnandwidth).
Newegg has 256mb sticks for $74 including shipping. Total for two sticks and 512mb total is $150.Link
This is about the price of ONE 256mb stick of Samsung PC1066 RDRAM. On pricing, I would bet that the SIS 655 would be priced at about the same price-point as current 850E mobos and then come down once a steady supply has been established.
Current DDR platforms have absolutely, positively, beyond a shadow of any doubt, NO latency advantage in real world system performance over pre-fetching i850e / PC-1066 systems. Get over the rambus/latency fallacys Ilmater...they are non-existant!
We are not talking about current DDR platforms here. We are talking about the up and coming Granite Bay, which will undoubtedly have all the tricks of the 850E such as buffering and pre-fetching applied to a Dual DDR system.
This is from the AT FAQ:
The main drawback of RDRAM is its high latency. However, today's chipsets such as the Intel i850 employ buffering, prefetch and other techniques to effectively "hide" the latency and reduce its performance penalty. Each additional RDRAM device on the bus increases the bus length and therefore latency. Thus from a performance standpoint, purchasing an 8-device RIMM would be preferential to purchasing a 16-device RIMM.
DDR does lose latency compared to its frequency as higher speeds are achieved such as DDR333 and 400 but I would attribute this to it running asynchronously with the FSB. Dual DDR GB will be running synchronously when using PC2100. Even the nforce2 for the AMD platform runs best when running memory synchronously with the FSB (best balance of latency/bnandwidth).
Ummm, in fact, WE WERE commenting on what Ilmater had said ( "Considering that currently DDR has lower latency," ...Ilmater) about "current DDr" platfroms were performing like. You took my statement out of context.
& I could care less what is said about rimm latency in the FAQ, the bottom line is that in real worlds performance, rimm latency is not an issue, and hopefully won't be with future DCDDR implementations.
RanDum, we're talking about Dual Channel Motherboards here. There's no point in comparing prices using only 1 stick. 1 256MB stick of PC2100 nor 1 256MB stick of PC1066 will work in GB/SiS 655/850e mobo's. They must have 2 sticks. So. and 2 128MB sticks of Sammy PC1066 cost $75 a piece ($150 total), so your $150 price for PC1066 is still correct. I will add further, that a single stick of Sammy RIMM4200 costs only $120. The downside to that is the price of the P4T533 is $180, so bottom line is that things pretty much balance themselves out. I do agree that in all likelyhood for example, the price of the Asus P4SDX will cost close to the P4T533-C/P4T533. If SiS 655's memory controller is good enough (which I believe it will be considering how good the 648's is), then it may be the best, but like we keep on saying, only offical benchys, and retail prices will give a final verdict.
Maybe, but see, what Fkloster is wondering (which I am as well0 is what kind of Latency increases/decreases will occur with the move to Dual Channel DDR? I honestly don't know enough about DDR and DDR Memory controllers to be able to say whether it will increase or decrease. Something that may be worth mentoning is that IIRC, there is little, if at all difference between i860 with PC800 and E7500 with PC1600 (running synchorously with the Xeon's 100 QDR fsb and PC1600's 100MHz DDR clock). Anyways, all this is speculation. Soon enough reviews will be here.
RanDum, we're talking about Dual Channel Motherboards here. There's no point in comparing prices using only 1 stick. 1 256MB stick of PC2100 nor 1 256MB stick of PC1066 will work in GB/SiS 655/850e mobo's. They must have 2 sticks. So. and 2 128MB sticks of Sammy PC1066 cost $75 a piece ($150 total), so your $150 price for PC1066 is still correct.
Thats why I priced TWO sticks of 256mb PC2100 Crucial for a total of 512mb since 512mb of memory is the preferred configuration. I am comparing prices of PC2100 in relation to PC1066 RDRAM. So if 512mb of PC2100 is equal in price to ONLY 1 stick of 256mb PC1066 RDRAm, then 512mb (two 256mb sticks) of PC1066 is TWICE the price (get the comparison?
) I don't know whats so hard with that.
You can probably buy a Dual-DDR mobo AND PC2100 memory for the price of the RDRAM alone (again, this is based on 512mb or two sticks of 256mb each) and still get the same or even greater performance.
UPDATE: Newegg now has Samsung PC1066, 256mb for $175.Link
Maybe, but see, what Fkloster is wondering (which I am as well0 is what kind of Latency increases/decreases will occur with the move to Dual Channel DDR?
If the nforce2 is a good indication, there will be NO latency penalty. LINK. Notice how the latency for 64-bit and 128-bit DDR333 is exactly the same at 244 under Cachemem.
Thats why I priced TWO sticks of 256mb PC2100 Crucial for a total of 512mb since 512mb of memory is the preferred configuration. I am comparing prices of PC2100 in relation to PC1066 RDRAM. So if 512mb of PC2100 is equal in price to ONLY 1 stick of 256mb PC1066 RDRAm, then 512mb (two 256mb sticks) of PC1066 is TWICE the price (get the comparison?
) I don't know whats so hard with that.You can probably buy a Dual-DDR mobo AND PC2100 memory for the price of the RDRAM alone (again, this is based on 512mb or two sticks of 256mb each) and still get the same or even greater performance.
UPDATE: Newegg now has Samsung PC1066, 256mb for $175.Link
Maybe, but see, what Fkloster is wondering (which I am as well0 is what kind of Latency increases/decreases will occur with the move to Dual Channel DDR?
If the nforce2 is a good indication, there will be NO latency penalty. LINK. Notice how the latency for 64-bit and 128-bit DDR333 is exactly the same at 244 under Cachemem.
Oh, I see, now suddenly the story changes. As I was a part of the conversation, I'm quite sure that WE WERE talking about whether or not RDRAM has higher latency, not whether or not it shows up in real world performance. As I said before, plain and simple, RDRAM has a higher latency than DDR. Period. You first stated:
Indeed, this is not the case with current DDR. The latency of DDR is not worse than that of RDRAM. That's why I said what I said in the first place. And, if you look back to the original comment:
... you'll see that he was clearly speaking about latency as well. Now, you could have made the claim that because of prefetch and other optimizations, the performance difference isn't there, but you know that our answer would be that they could use the same technologies with DCDDR memory controllers (see RanDum72's comment), effectively giving them both their better latencies as well as prefetch and buffering. There's no reason not to. Before, Intel was so concerned about bandwidth (and rightfully so, as their procs are very bandwidth hungry) that they were willing to stick with RDRAM and find other ways to get around latency troubles. Now that DCDDR will equal RDRAM in bandwidth, Intel will use the lower latency of DDR, along with prefetch and buffering, to further space DCDDR ahead of Rambus. Which is why, I believe, Intel has stopped talking about RDRAM and have made some indications that it will not be running their high-end systems at least by next year at this time, if not sooner.
Fallacy? What fallacy? It looks like a fact to me.
Now, your point about the complicated timing schematics could be very correct, but that remains to be seen. Is working pairs the reason for the low latencies present in RDRAM today? Maybe. I'd be interested to know if that was the case, or if it was, rather, an issue with the RIMM interface or just in the nature of the way RDRAM chips work. When all's said and done, the most important fact is that we simply don't know, and anything we say is just conjecture for now.
I used to dismiss Rambus altogether because of their corporate practices and high cost. After the launch of dual channel PC800 I admitted that the technology was sound. After they launched dual channel PC1066 on a single RIMM and called it RIMM4200 I was amazed. Rambus really has listened to the market of late and I can't wait to see the performance of the single channel 64bit 1333mhz modules.
Toshiba might be using some new Rambus tech in graphics cards memory.
" Mr. Cross noted he doesn't expect shipments of products based on Yellowstone technology until 2004 or 2005. Initially, the technology will be targeted at graphics-intensive electronic applications such as three-dimensional games and graphics for PCs, he said."
Link
Toshiba might be using some new Rambus tech in graphics cards memory.
" Mr. Cross noted he doesn't expect shipments of products based on Yellowstone technology until 2004 or 2005. Initially, the technology will be targeted at graphics-intensive electronic applications such as three-dimensional games and graphics for PCs, he said."
Link
Will there be a cheap version of the GB? Like 2.4ghz or similar, as in not top of the line... how much would this cost around abouts?
I think you're possibly confusing GB (Granite Bay) with HT (HyperThreading) the former is a DC DDR (Dual Channel Double Data Rate) chipset (therefore independent of CPU speed) the latter is a CPU technology that will only be present in P4s clocked at 3.06GHz and higher.
GB is expected to be released (and thus, reviewed) on or about Nov. 14th - which is also the approximate date that NV30's NDA is expected to expire, so it should be an exciting week
As I have shown in past threads about this lowly pc2100 in dual channel ddr mobo will be slightly better then pc1066 right now, and that is with older pc2100 sticks....
Pc 1066 gets 3240's in sandra and I believe pc1200 gets around 3600's and pc1333 should be in the 4000 range, right???
Now go take a look at hardocp.coms recent tidbit about the DCDDR mobo he had in which he had dual 376mhz ddr, and yes ocers are going to thoroughly enjoy these baords much better then most, and he had 4500's!!!!! 13% increase in bandwidth over pc1333 chips and more then current pc1066 rdram has oevr pc3200 ddr. I myself will enjoy running 342mhz ddr hopefully with my 171fsb and get around 4000 as per that thread at hardocp.com....
Then lets dream a bit and hope for some eventual mem ratio support and could you imagine most doing 400mhz or dual channel; pc3200ddr...can you say pc1333 killer???? It will score about 4800-4900 or near 20% plus....
when is the pc1333 coming out???
Pc 1066 gets 3240's in sandra and I believe pc1200 gets around 3600's and pc1333 should be in the 4000 range, right???
Now go take a look at hardocp.coms recent tidbit about the DCDDR mobo he had in which he had dual 376mhz ddr, and yes ocers are going to thoroughly enjoy these baords much better then most, and he had 4500's!!!!! 13% increase in bandwidth over pc1333 chips and more then current pc1066 rdram has oevr pc3200 ddr. I myself will enjoy running 342mhz ddr hopefully with my 171fsb and get around 4000 as per that thread at hardocp.com....
Then lets dream a bit and hope for some eventual mem ratio support and could you imagine most doing 400mhz or dual channel; pc3200ddr...can you say pc1333 killer???? It will score about 4800-4900 or near 20% plus....
when is the pc1333 coming out???
Sorry, my mistake.. just new this as I done all my research on rdram.. was going to buy rdram (literally) until I heard about dual-channel ddr (G.bay).. I just wanna know is g.bay the make of ddr as crucial is a make also?
And another thing, this new intel chipset (HT), do I need to get it to make full use of dual-channel ddr? or can the p4 2.4ghz 533fsb do just as good job?
Thanks...
your post is confusing
lets break it down.
Granite bay is a new chipset from intel for the P4.
it will support Dual Channel DDR PC2700
It should run any DDR memory whether it be from crucial, mushkin, corsair, etc...
hope that helps
I don't think it will, not even unoffically. Springdale will be Intel's first DCDDR333 chipset.
Its 'unofficial support' of PC2700 depends on the motherboard manufacturer, IMO. The 845-series chipsets didn't support PC2700 but that didn't stop manufacturers from adding the feature. The 845PE 'officially' supports PC2700 but its moot when earlier versions run the same memory just as well.
Its 'unofficial support' of PC2700 depends on the motherboard manufacturer, IMO. The 845-series chipsets didn't support PC2700 but that didn't stop manufacturers from adding the feature. The 845PE 'officially' supports PC2700 but its moot when earlier versions run the same memory just as well.
Speculation is that any DC DDR chipset which supports 333 mhz - still won't outperform the same chipset using 266. The reason is that 333 will run asynchronously to the system (533), while the 266 will be in perfect lock step. When Springdale is out - the system bus will up to 667 - and thus 333 will be the preferred and highest performance DDR memory. The basis for this theory is current performance of the NForce with various memory speeds. Can anyone confirm -?
DynaOne
DynaOne
Just wait until dual channel RIMM5200(PC1333):Q
That should satsify almost *anything*, and is way beyond the reach of even DCDDR400. HAH!
That should satsify almost *anything*, and is way beyond the reach of even DCDDR400. HAH!
Try Quad channel rambus, 2 x 32bit RIMM 4200's in dual channel configuration. Thats about 8.4 GB/sec bandwidth right there.
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: 23K
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 23K
-
-
Facebook
Twitter