Zakule, I'm done with you. I now see the other users frustration with you on this thread. Your arguments make no sense what so ever. None. To even argue the need for a defined benchmark to compare systems, is ludicrous. If you don't understand that simple concept, you are beyond hope. I'm not getting sucked in. Find someone else to play your silly little game.
Very well. But what I'm saying makes perfect sense. Not playing games, just debating.
Consistent minimum framerates in the 3D apps you use is what you're looking for. What is the minimum frame rate that see you in an app when you're playing it? How consistent your system is at staying above a given frame rate is what matters. Getting 175 average FPS doesn't matter if the game drags at 16 FPS during certain points that it shouldn't. I don't see what's so confusing or inflamatory about what I'm saying here. All I've been saying is that how the apps you use perform matters more than how benchmark demo utilities perform.
Consistent minimum framerates in the 3D apps you use is what you're looking for. What is the minimum frame rate that see you in an app when you're playing it? How consistent your system is at staying above a given frame rate is what matters. Getting 175 average FPS doesn't matter if the game drags at 16 FPS during certain points that it shouldn't. I don't see what's so confusing or inflamatory about what I'm saying here. All I've been saying is that how the apps you use perform matters more than how benchmark demo utilities perform.
>>>Getting 175 average FPS doesn't matter if the game drags at 16 FPS during certain points that it shouldn't.<<<
More wild speculation. If my box, fklosters or anyone elses lagged to that extreme they would have noticed it and wouldn't be here bragging about it's performance. And shouldn't have to prove to you that theirs doesn't exhibit that behaviour although if your box does feel free to let us all know about it.
More wild speculation. If my box, fklosters or anyone elses lagged to that extreme they would have noticed it and wouldn't be here bragging about it's performance. And shouldn't have to prove to you that theirs doesn't exhibit that behaviour although if your box does feel free to let us all know about it.
<< More wild speculation. If my box, fklosters or anyone elses lagged to that extreme they would have noticed it and wouldn't be here bragging about it's >>
I know of a few different people that had this sort of frame rate problem with Dungeon Siege on it's initial release. Not at all wild speculation.
i would just like to say, rather than having a flame war, post some numbers or post nothing. This thread is getting rediculously long and isnt even on the subject anymore.
I would like to see some 3Dmark 2k1se scores, maybe some Q3A or Sacrifice benches, anything you guys can bench with the 2 different environments. Otherwise, make another thread if you want to discuss benchmarking techniques and please keep the flames to a minimum, im very suprised this thread hasnt been locked yet.
I would like to see some 3Dmark 2k1se scores, maybe some Q3A or Sacrifice benches, anything you guys can bench with the 2 different environments. Otherwise, make another thread if you want to discuss benchmarking techniques and please keep the flames to a minimum, im very suprised this thread hasnt been locked yet.
Funny, we're seeing an early rise in the RAMBUSt promoters here. Too bad PC1066 is vaporware. I'm not surprised. Maybe they're still trying to figure out how to attach a heatsink to the modules thick enough to half-arsedly cool them without being so thick that 2 RIMMs can't fit side-to-side! Or maybe they're having Thermaltake design an Orb. I can see it now, the RambustOrb. 
DDR400 is real, it is here, and it rocks. It runs cooler, costs less (most certainly less than what PC1066 will, assuming it ever appears), and provides more bandwidth with less latency. What's so hard to understand for some people about that?
DDR400 is real, it is here, and it rocks. It runs cooler, costs less (most certainly less than what PC1066 will, assuming it ever appears), and provides more bandwidth with less latency. What's so hard to understand for some people about that?
Zakule, give one example where a DDR based system has a higher lowest framerate than a RDRAM system, then you we can talk. If you don't even have any examples yourself your whole argument falls to pieces.
BTW, is DDR400 even a standard yet? As far as I know no motherboards even have official support for it and the only modules availible are cas3 which means they're no faster than DDR333 cas2.
hey pabster, eat your words http://shop2.outpost.com/product/3308112/
think before you speak, they will have it in stock in 8 days. the 20 chips they had sold in 25 minutes.
must be because of the crappy performance
think before you speak, they will have it in stock in 8 days. the 20 chips they had sold in 25 minutes.
must be because of the crappy performance
Yeah, well since he's talking about vaporware, you can just as easily overclock PC800 to PC1066 as you can with DDR333 to DDR400.
I disagree Pabster that PC3200 is here to stay, I personally can see it becoming like PC150 did, except it may have a bit more support (ie actual chipsets like SIS 648 and KT400) but still, it will be over priced, and it still can't touch PC1066. As for PC1066's price, we'll see, as for its heat, as I said to Zakule earlier, I do not know why people make such a big deal about the heat as long as it doesn't greately effect operation or price, which it doesn't in the case of RDRAM.
That is just it, early indications from JDEC are that DDR400 will never be ratified as a memory standard. Now SiS 648 and KT400 may make it appear that it is a standard, but it in reality never will, and it will never fully take off.
To everyone saying DR400 will always have high latency etc.
Check this link out. They are testing various DDR 333 modules (of which many are available here in sweden right now) that are capable of 200+ mhz at the most aggressive timings:
link
It's not a new review but It's very nice I think! I studied it more thoroughly a few days ago. And to my surprise the Samsung modules that are the performance leaders of the test are available at a good price even here. Lower than RDRAM anyway. So.. the price is probably even more reasonable in the US. Stuff that aint' mainsteam usually costs alot more here.
We don't know if DDR400 will become a standard, PC3200.. Yet. but since yields seem pretty good I bet it will.
I don't agree that DDR400 won't be able to touch pc1066 performance wise. I can't find the friggin numbers now but I saw in a review somewhere on THG (I believe it was the 845G chipset perform just a little worse than pc1066 on the 850 in Quake 3)..
So I believe that DDR400 should perform somwhere in between pc1066 and pc1200. Ofcourse there is no Intel chipset to support this speed yet. Only crappy SIS chipsets do.
Regarding Pc1200.. Which will be running on a frequency of 150 mhz. How is that connected to a 133 mhz bus?.
What multiplier is that? it seems extremely.. asynchronous. but maybe this doesn't matter.. I dunno.
Check this link out. They are testing various DDR 333 modules (of which many are available here in sweden right now) that are capable of 200+ mhz at the most aggressive timings:
link
It's not a new review but It's very nice I think! I studied it more thoroughly a few days ago. And to my surprise the Samsung modules that are the performance leaders of the test are available at a good price even here. Lower than RDRAM anyway. So.. the price is probably even more reasonable in the US. Stuff that aint' mainsteam usually costs alot more here.
We don't know if DDR400 will become a standard, PC3200.. Yet. but since yields seem pretty good I bet it will.
I don't agree that DDR400 won't be able to touch pc1066 performance wise. I can't find the friggin numbers now but I saw in a review somewhere on THG (I believe it was the 845G chipset perform just a little worse than pc1066 on the 850 in Quake 3)..
So I believe that DDR400 should perform somwhere in between pc1066 and pc1200. Ofcourse there is no Intel chipset to support this speed yet. Only crappy SIS chipsets do.
Regarding Pc1200.. Which will be running on a frequency of 150 mhz. How is that connected to a 133 mhz bus?.
What multiplier is that? it seems extremely.. asynchronous. but maybe this doesn't matter.. I dunno.
Yea I suppose we don't know, although like I said, early indications point to it never becoming a standard. but I have one question, where are you getting tthat companies are getting good yields on PC3200??? jw
I do have to agree on the THG article (845G with DDR333 was 4-6% behind 850E+PC1066, allthough, it is unclear whether the memory controller of 845G will be put in 845E, plus Intel will never adopt DDR400 unless it becomes a JDEC standard.
Well, it would be 4.5x the fsb at 133fsb, however, I expect that the fsb will be increased to 150fsb by then as well. Btw, SSXeon found some very intruiging info on Prescott. Here's the thread, the link he gave needs to be translated, which he did. Here's the important part:
:QRDRAM support next year:Q Prescott maybe coming much sooner than we think:Q 667fsb:Q No word though on PC1200 support, but wow!!!
Athlon4all .. My bad there on the yields.. I meant the yields seem good on PC2700 ram, not PC3200.
The review I sent along there is why I think so:
Three out of five sticks did 200+ mhz at the most aggressive timings while all of them did 215-220 mhz at more modest timings.
So.. if PC2700 is already performing so well. I think it's not too far fetched to believe yeilds are good. And that PC3200 isn't too far away. And there IS PC3200 ram available in stores now. It is expensive. But at least here in sweden a 256 MB stick of PC3200 ram is not more than about 10% more expensive than a good stick of RDRAM.
Thank you for the exciting info on DUAL CHANNEL DDR 333!!
If that info is correct we will have 5.4 GB/sec of bandwith at our hands next summer! By that time PC2700 will be very much mainstream and Intel ofcourse knows this. It wants cheap memory for the P4. Dual channel PC2700 will provide 28% higher bandwidth than PC1066 RDRAM and 12% higher than PC1200 RDRAM.
And don't forget that latency will be lower as well.
You have to admit it now ppl.. Dual channel DDR will seriously outperform RDRAM. Single channel comes veeeeeeery close to dual channel PC1066 with the correct memory controller. And doubling the bandwidth will.. well.. You can figure it out for yourselves.
Oh.. btw.. just imagine 5 GB+ scores in Sisoft sandra
Also let's not forget that by next summer DDR400 could very well have been approved and validated.. Dual channel=6.4 GB/sec.. drooool. Then again.. who knows what Rambus has up their sleves.
Either way.. Rambus has serious competition. Which may force them to move forward a little more quickly, hopefully.
All we want is better performance at a better price, right?
The review I sent along there is why I think so:
Three out of five sticks did 200+ mhz at the most aggressive timings while all of them did 215-220 mhz at more modest timings.
So.. if PC2700 is already performing so well. I think it's not too far fetched to believe yeilds are good. And that PC3200 isn't too far away. And there IS PC3200 ram available in stores now. It is expensive. But at least here in sweden a 256 MB stick of PC3200 ram is not more than about 10% more expensive than a good stick of RDRAM.
Thank you for the exciting info on DUAL CHANNEL DDR 333!!
If that info is correct we will have 5.4 GB/sec of bandwith at our hands next summer! By that time PC2700 will be very much mainstream and Intel ofcourse knows this. It wants cheap memory for the P4. Dual channel PC2700 will provide 28% higher bandwidth than PC1066 RDRAM and 12% higher than PC1200 RDRAM.
And don't forget that latency will be lower as well.
You have to admit it now ppl.. Dual channel DDR will seriously outperform RDRAM. Single channel comes veeeeeeery close to dual channel PC1066 with the correct memory controller. And doubling the bandwidth will.. well.. You can figure it out for yourselves.
Oh.. btw.. just imagine 5 GB+ scores in Sisoft sandra
Also let's not forget that by next summer DDR400 could very well have been approved and validated.. Dual channel=6.4 GB/sec.. drooool. Then again.. who knows what Rambus has up their sleves.
Either way.. Rambus has serious competition. Which may force them to move forward a little more quickly, hopefully.
All we want is better performance at a better price, right?
Ok np, I was just wanting to make sure.
Again, I am not doubting DC DDR's Performance, As a matter of fact, I used to be a huge supporter of Dual Channel DDR, but I would not call Dual Channel DDR "Cheap" by any means. The cost is going to be enormous to produce both the chipset and boards, I'm thinking $200 for a ECS Springdale board at launch, and $250 for an Asus one, where as you'll be able to likely get a Springdale 850E Asus for around $150, $170 at most. Although, DC DDR333 will be faster than even DC 16-bit RDRAM PC1200 on 667fsb. 32-bit RIMM's (real 32-bit RIMM's that provide PC1066 for one RIMM even if its in a Single Channel Chipset, 32-bit RIMM's that on 850e=8.4GB/ps) could change this picture dramtically, and the thing is, they really aren't any more expensive to make from what I understand when compared to 16-bit RIMM's in use today.
Yes, that is all I want, and right now, the performance is there for RDRAM and the price is good, I really don't see that changing until Springdale/Granite Bay hits. And even then, I think RDRAM will remain as the value segment.
Athlon4all
I think you got this wrong. A 32 bit rimm will on the 850E chipset give the same performance as dual 16 bit rimms.
You won't get quad channel RDRAM if you put in two 32 bit rimms.
I imagine quad channel (4x16 bit=64 bit) will be kind of expensive as well. But this is sheer speculation. I know little of this.
I think you got this wrong. A 32 bit rimm will on the 850E chipset give the same performance as dual 16 bit rimms.
You won't get quad channel RDRAM if you put in two 32 bit rimms.
I imagine quad channel (4x16 bit=64 bit) will be kind of expensive as well. But this is sheer speculation. I know little of this.
MadRat
Lifer
- Oct 14, 1999
- 11,967
- 280
- 126
I'm not sure where they were connecting 32-bit RIMMs to quadruple interleaving. The 850s already can use 16-bit channels from a pair of RIMMs in the current configurations for an effective two-way interleave of the memory. It would be nice if that was in fact what would happen, but the 850E would not support a quadruple interleave unless it was totally redesigned. People also confuse interleaving with multiple channels. Interleaving does not offer twice the bandwidth, but rather allows transfers of data from the bus when it would normally be idle while waiting for another transfer. At most its a 30% gain in performance, while often it offers no performance gain at all. The nForce is one chipset that people confuse as being dual-channel when in reality its "T-bar" design is a fancy name for interleaved.
Madrat. You know this interleaving bizniz alot better than I do. But I was under the impression that the N-force can't deliver any better performance due to the fact that the Athlon has a limited fsb. So the chipset isn't at fault here,the Athlon is. And the P4 wouldn't be the limitation, right?
I'm sure you are going to let me know if I am rigth or wrong.
I'm sure you are going to let me know if I am rigth or wrong.
MadRat
Lifer
- Oct 14, 1999
- 11,967
- 280
- 126
The nForce has two advantages that are bottlenecked by the Athlons.
1) Their hardware prefetch is already duplicated in the Palominoes
2) The Athlon is limited to a 64-bit pathway to external memory
So in effect the Athlon cannot take any more data in than an already efficient memory design.
The Pentium 4 probably wouldn't gain much from the nForce-style of chipset simply because SiS and Intel already run efficient DDR chipsets and the nForce doesn't really offer any improvement over them. Nobody will know until nVidia releases one, though, so don't take my word for it.
1) Their hardware prefetch is already duplicated in the Palominoes
2) The Athlon is limited to a 64-bit pathway to external memory
So in effect the Athlon cannot take any more data in than an already efficient memory design.
The Pentium 4 probably wouldn't gain much from the nForce-style of chipset simply because SiS and Intel already run efficient DDR chipsets and the nForce doesn't really offer any improvement over them. Nobody will know until nVidia releases one, though, so don't take my word for it.
MadRat
Lifer
- Oct 14, 1999
- 11,967
- 280
- 126
I didn't say that it couldn't, but nForce has yet to prove it has any more bandwidth than any other DDR system. Interleaving offers minimal gains, something you can already test with VIA chipsets since they support interleaving in all of their post-Apollo Pro chipsets.
MadRat, I still am confused. According to Anand's technology Preview of nForce, Twinbank has 2 64-bit Memory controllers that are interleaved I believe, but because there are 2 memory controllers, each should provide 2.1GB/ps right???? What about 850e with 32-bit RIMM's??? Do you know how that works??? I am more confused than ever. Can someone explain how this works? I am more confused than ever!!!:|
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 Discussion Threads
- Started by Tigerick
- Replies: 21K
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 23K
-
-
Facebook
Twitter