Hey MadRat- How much real-world performance does RDRAM have over DDRAM clocked at the same FSB? I know the Sandra memory scores look impressive, but if its only 1-3% framerate advantage, I can't see going for RDRAM because of the lack of motherboard options.
MadRat
Lifer
- Oct 14, 1999
- 11,999
- 307
- 126
<<Uh no . 160FSB with DDR equates to around 2400/2400 sisoft 2002 scores which is on par with PC800 RDRAM. Provide some proof that RDRAM is blowing the wheels off DDRAM besides Tom's Hardware 3ghz overclock northwood running default PC2100 DDR speeds. Most users that overclock are seegin 300-400mb/s extra boost than what tom's hardware had. Most benchmarks ive seen show default DDRAM about 5% slower at most with PC800 +/- a few points. Anyway why would this irritate you? The combo is still fast as hell anyway and you wont even notice a difference between DDR and RDRAM except in 3dmark2001. 350FPS at 640x480 is quakeIII doesnt mean anything.>>
We aren't talking a "mere 5% difference" in performance, but also a scaling of performance when the Northwood@533fsb is paired with RDRAM. Besides that, its more like 5-15% performance difference in everday applications. The stock Pentium 4 has a theoretical 400fsb that roughly translates to a 3.2GB/sec potential. The overclocked Pentium 4 with a 533fsb translates to around 4.2GB/sec of potential. DDR333 has a limited potential of only 3.0GB/sec, DDR266 is limited to 2.1GB/sec, and DDR200 is limited to 1.6GB/sec.
As you scale the speed of the CPU the bandwidth potential is consumed, meaning that a 2GHz P4 uses more of the potential than a 1.3GHz P4. The Pentium 4's requirement for memory bandwidth takes a huge leap when you begin to scale past 2GHz, which means the Pentium 4's potential will never be realized with the DDR memory no matter how far its FSB is overclocked. Sure, a sub-2GHz P4 can utilize DDR memory just as well as RDRAM, but that is not the point.
Syncronous RDRAM closely matches its potential to the front-side bus, always fulfilling the needs of the CPU regardless if its 400fsb or 533fsb. Even when you run RDRAM at 4:3 (533fsb:400mem) there is more potential than with the DDR systems. RDRAM just makes the most sense in Northwoods systems, overclocked or not.
We aren't talking a "mere 5% difference" in performance, but also a scaling of performance when the Northwood@533fsb is paired with RDRAM. Besides that, its more like 5-15% performance difference in everday applications. The stock Pentium 4 has a theoretical 400fsb that roughly translates to a 3.2GB/sec potential. The overclocked Pentium 4 with a 533fsb translates to around 4.2GB/sec of potential. DDR333 has a limited potential of only 3.0GB/sec, DDR266 is limited to 2.1GB/sec, and DDR200 is limited to 1.6GB/sec.
As you scale the speed of the CPU the bandwidth potential is consumed, meaning that a 2GHz P4 uses more of the potential than a 1.3GHz P4. The Pentium 4's requirement for memory bandwidth takes a huge leap when you begin to scale past 2GHz, which means the Pentium 4's potential will never be realized with the DDR memory no matter how far its FSB is overclocked. Sure, a sub-2GHz P4 can utilize DDR memory just as well as RDRAM, but that is not the point.
Syncronous RDRAM closely matches its potential to the front-side bus, always fulfilling the needs of the CPU regardless if its 400fsb or 533fsb. Even when you run RDRAM at 4:3 (533fsb:400mem) there is more potential than with the DDR systems. RDRAM just makes the most sense in Northwoods systems, overclocked or not.
Besides that, its more like 5-15% performance difference in everday applications.
Link? 5-15% in word,internet explorer? huh? If you want to talk 533FSB dual channel DDR is hitting this summer along with PC1066 RDRAM. Both the RDRAM and dual channel DDR chipsets equate to the same bandwith , which is 4.2 gb/s. Overclock them both and they both are dead even. And still the difference is minimal to nil now. Anyway,everything you just said is all theoretical. DDR doesnt meet the full potential for the P4 now,which is true, but where does it actually show up in applications. The boost in most stuff is not as big as everyone makes it out to be. BTW I'm pretty sure DDR333 is 2.7gb/s not 3.0gb/s.
As you scale the speed of the CPU the bandwidth potential is consumed, meaning that a 2GHz P4 uses more of the potential than a 1.3GHz P4. The Pentium 4's requirement for memory bandwidth takes a huge leap when you begin to scale past 2GHz, which means the Pentium 4's potential will never be realized with the DDR memory no matter how far its FSB is overclocked. Sure, a sub-2GHz P4 can utilize DDR memory just as well as RDRAM, but that is not the point.
I dont know if this is entirely true either. In most CPU intesive tasks DDR is on par or even better than RDRAM. But theres less memory bandwith and yet the thing still runs at full potential at its speed. I can see were the difference is when you have cpu and memory intensive tasks going on. But still sometimes some of that bandwith wont even be used or needed in those applications. Anyway basically even at toms hardware with the 3ghz and PC2100 DDR ,it was winning basically every single cpu intesive task. So even with extra memory bandwith its just not needed in these tests. The cpu is running 3ghz really no matter how much bandwith you've got to a certian extent.
Link? 5-15% in word,internet explorer? huh? If you want to talk 533FSB dual channel DDR is hitting this summer along with PC1066 RDRAM. Both the RDRAM and dual channel DDR chipsets equate to the same bandwith , which is 4.2 gb/s. Overclock them both and they both are dead even. And still the difference is minimal to nil now. Anyway,everything you just said is all theoretical. DDR doesnt meet the full potential for the P4 now,which is true, but where does it actually show up in applications. The boost in most stuff is not as big as everyone makes it out to be. BTW I'm pretty sure DDR333 is 2.7gb/s not 3.0gb/s.
As you scale the speed of the CPU the bandwidth potential is consumed, meaning that a 2GHz P4 uses more of the potential than a 1.3GHz P4. The Pentium 4's requirement for memory bandwidth takes a huge leap when you begin to scale past 2GHz, which means the Pentium 4's potential will never be realized with the DDR memory no matter how far its FSB is overclocked. Sure, a sub-2GHz P4 can utilize DDR memory just as well as RDRAM, but that is not the point.
I dont know if this is entirely true either. In most CPU intesive tasks DDR is on par or even better than RDRAM. But theres less memory bandwith and yet the thing still runs at full potential at its speed. I can see were the difference is when you have cpu and memory intensive tasks going on. But still sometimes some of that bandwith wont even be used or needed in those applications. Anyway basically even at toms hardware with the 3ghz and PC2100 DDR ,it was winning basically every single cpu intesive task. So even with extra memory bandwith its just not needed in these tests. The cpu is running 3ghz really no matter how much bandwith you've got to a certian extent.
MadRat
Lifer
- Oct 14, 1999
- 11,999
- 307
- 126
<<5-15% in word,internet explorer? huh?>>
In Office applications is where memory bandwidth really shows itself, so perhaps yes. Considering how much the CPU does, even IE will benefit from the extra bandwidth.
<<I dont know if this is entirely true either. In most CPU intesive tasks DDR is on par or even better than RDRAM. But theres less memory bandwith and yet the thing still runs at full potential at its speed. I can see were the difference is when you have cpu and memory intensive tasks going on. But still sometimes some of that bandwith wont even be used or needed in those applications.>>
I've yet to see a P4 using DDR memory perform on par, let alone exceed a P4 with RDRAM, in any application. The article on Tom's Hardware points out a huge advantage between the Pentium 4 at 400fsb versus one at 533fsb, both using synchronous memory. All of the DDR articles prior to his newest article have shown a 3-5% difference in applications comparing DDR memory on VIA chipsets to RDRAM on the 850 chipset. We are talking performance at lower speed ratings of 1.5GHz, not even the 2GHz or better that an overclocked Northwood will often reach.
Well WinXP seems to be sucking up about 190MB of memory right now. Very few applications are going to fit into that tiny 512k of L2 cache, meaning my everyday applications are running from memory not cache. Sure the cache holds the immediate information, but if even 1% of that 190MB is needed then that easily exceeds that 512k cache's potential.
This means that the Pentium 4 is going to rely heavily on the memory for performance when doing realworld tasks. Benchmarks that stress memory bandwidth are often better than the 10-15% difference when considering DDR versus RDRAM. You want to know if non-memory intensive benchmarks matter whether on DDR or RDRAM. In every realworld case where memory is heavily consumed by background tasks it will matter.
<<Both the RDRAM and dual channel DDR chipsets equate to the same bandwith , which is 4.2 gb/s.>>
Interesting, but what dual-channel DDR chipset is that? So far there are no dual-channel chipsets. Are you talking VIA chipsets? Funny, but there is a shortage of motherboards out there being touted with this mythical dual-channel VIA chipset. Oh, wait - you probably meant interleaved. So from what I gather you feel interleaved equates to dual-channel. Not hardly, more like 2/3rd's of a true dual-channel. Its an improvement over plain single-channel memory, but its not the miracle worker you may think.
In Office applications is where memory bandwidth really shows itself, so perhaps yes. Considering how much the CPU does, even IE will benefit from the extra bandwidth.
<<I dont know if this is entirely true either. In most CPU intesive tasks DDR is on par or even better than RDRAM. But theres less memory bandwith and yet the thing still runs at full potential at its speed. I can see were the difference is when you have cpu and memory intensive tasks going on. But still sometimes some of that bandwith wont even be used or needed in those applications.>>
I've yet to see a P4 using DDR memory perform on par, let alone exceed a P4 with RDRAM, in any application. The article on Tom's Hardware points out a huge advantage between the Pentium 4 at 400fsb versus one at 533fsb, both using synchronous memory. All of the DDR articles prior to his newest article have shown a 3-5% difference in applications comparing DDR memory on VIA chipsets to RDRAM on the 850 chipset. We are talking performance at lower speed ratings of 1.5GHz, not even the 2GHz or better that an overclocked Northwood will often reach.
Well WinXP seems to be sucking up about 190MB of memory right now. Very few applications are going to fit into that tiny 512k of L2 cache, meaning my everyday applications are running from memory not cache. Sure the cache holds the immediate information, but if even 1% of that 190MB is needed then that easily exceeds that 512k cache's potential.
This means that the Pentium 4 is going to rely heavily on the memory for performance when doing realworld tasks. Benchmarks that stress memory bandwidth are often better than the 10-15% difference when considering DDR versus RDRAM. You want to know if non-memory intensive benchmarks matter whether on DDR or RDRAM. In every realworld case where memory is heavily consumed by background tasks it will matter.
<<Both the RDRAM and dual channel DDR chipsets equate to the same bandwith , which is 4.2 gb/s.>>
Interesting, but what dual-channel DDR chipset is that? So far there are no dual-channel chipsets. Are you talking VIA chipsets? Funny, but there is a shortage of motherboards out there being touted with this mythical dual-channel VIA chipset. Oh, wait - you probably meant interleaved. So from what I gather you feel interleaved equates to dual-channel. Not hardly, more like 2/3rd's of a true dual-channel. Its an improvement over plain single-channel memory, but its not the miracle worker you may think.
<< "what speed is pci bus?"
Asus locks the PCI and AGP @ 33mhz and 66mhz respectively over 133mhz FSB.
>>are u sure
MadRat, the chipset is called the Granite bay. Its a chipset directly from intel hitting xeon/northwood cpus. It's listed here under this roadmap. Even anandtechs main website has a preview of the dual channel ddr xeon version. Even if what your saying is true about the extra cache the northwood has it still has equated to a 10-15% boost over the williamette in almost all applications. GamePC's review of overclocked 1.6A's show how memory bandwith doesnt matter nearly as much in cpu intensive tests,or PC2100 DDR is providing more than it needs for those types of tasks right now. I dont know what reviews you've been reading but in everything I've seen,anything cpu intensive...the faster the clock the faster the application runs. Xibit Labs has a review showing the differences of PC800 RDRAM and PC2100 DDR with a 2ghz P4. DDR is less than 5% slower and on par in all tests. Tom purposely cripped the DDR board just to show how effective extra bandwith is for the P4,and if you notice even with horrible DDR memory scores in sandra 2002, 2050/2050. The 3ghz northwood still won around 50% of the tests,most of which were cpu intensive. RDRAM was winning the 3dmark2001's and quake 3's at 640x480. Not to mention theres many boards out there that allow higher DDR benchmarks. I'm running 2500/2500 memory scores with my setup right now, he could have used the same board with the 3ghz setup and seen much better results with even much higher memory benchmarks than mine. That board he used for the DDR setup has been known around here as not one of the best to use. Anyway , you seem to just be throwing numbers around with no links. Really overall a DDR mobo will lose by 10% in some tests and be on par on some tests,but overall it will be around 5% slower at default speeds. RDRAM does use the full potential on the P4 right now but in no way does it blow the wheels off of a DDR setup. Some of full potential is being seen in some tests but of a lot of it is in 3dmark2001, a synthetic benchmark, and quakeIII at 640x480. Who runs quakeIII at 640x480 with a cpu that fast? Some of its being wasted and you wont even see it put to use. Once more memory reliant applications come out in the future that will all change, thats why intel is going dual channel DDR. And if you think about it,it fits the P4 perfectly. You have PC1600 running which is 100 FSB x4=400FSB and 3.2gb/s and you have PC2100 running 133FSBx4=533FSB and 4.2gb/s. So everything will run in sync just like RDRAM does now and will in the future. Anyway to get back to the main point I still dont see how your so irritated that people are running 1.6ghz@2.4ghz with DDR memory running PC800 RDRAM speeds,its probably the most easiest to setup and best bang for your money right now.
well said christoph83!
And also, you't play benchmark all day.
Real-life performance is all that matters.
I will be getting a 1.6A NW, PC2700 512 and an Asus board..
And also, you't play benchmark all day.
Real-life performance is all that matters.
I will be getting a 1.6A NW, PC2700 512 and an Asus board..
MadRat
Lifer
- Oct 14, 1999
- 11,999
- 307
- 126
christoph83-
<<MadRat, the chipset is called the Granite bay. Its a chipset directly from intel hitting xeon/northwood cpus.>>
We are talking availability today, not 3Q2002.
<<GamePC's review of overclocked 1.6A's show how memory bandwith doesnt matter nearly as much in cpu intensive tests,or PC2100 DDR is providing more than it needs for those types of tasks right now.>>
GamePC's does what?? Look at there tests, none of there tests are sensical. The test rigs are such a hodgepodge they have no correllation from one system to the next. You can compare 533fsb only in the PC1066 system to the DDR400 system:
CPU Arithmetic
PC1066/P4-2.2GHz: 4154
DDR400/P$-2.2GHz: 3924
Memory Bandwidth
PC1066/P4-2.2GHz: 3306
DDR400/P$-2.2GHz: 2631
ScienceMark
PC1066/P4-2.2GHz: 133.7
DDR400/P$-2.2GHz: 131.87
3DMark 2001 1024x768x32
PC1066/P4-2.2GHz: 8623
DDR400/P$-2.2GHz: 8381
3DMark 2002 1600x1200x32
PC1066/P4-2.2GHz: 5314
DDR400/P$-2.2GHz: 5268
Valley of the Jaguars 1024x768x32
PC1066/P4-2.2GHz: 51.5
DDR400/P$-2.2GHz: 37
Quake!!! Fastest 512x384x16
PC1066/P4-2.2GHz: 343
DDR400/P$-2.2GHz: 316.7
Quake!!! High Quality 1024x768x32
PC1066/P4-2.2GHz: 241.5
DDR400/P$-2.2GHz: 232.3
Quake!!! High Quality 1600x1200x32
PC1066/P4-2.2GHz: 123.1
DDR400/P$-2.2GHz: 122.8
Return to Wolfenstein ATDemo6 - High Quality - 1024x768x32
PC1066/P4-2.2GHz: 52.9
DDR400/P$-2.2GHz: 54.4
Return to Wolfenstein ATDemo8 - High Quality - 1024x768x32
PC1066/P4-2.2GHz: 150
DDR400/P$-2.2GHz: 157.5
Photoshop 6.01 Filter Benchmark (v.2) *
PC1066/P4-2.2GHz: 93.1
DDR400/P$-2.2GHz: 93.5
Kinetix 3D Tank Model Rendering Time *
PC1066/P4-2.2GHz: 16:42
DDR400/P$-2.2GHz: 16:57
Windows Media 8.0 50MB MPEG Video to Windows Media Video Encode
PC1066/P4-2.2GHz: 3:26
DDR400/P$-2.2GHz: 3:26
L.A.M.E. 200MB Wav to MP3 File Encoding Time *
PC1066/P4-2.2GHz: 68
DDR400/P$-2.2GHz: 67
* = lower is better
<The Final Word
With that said, it's clear that RDRAM at PC-1066 speeds offers better overall performance compared to DDR400 memory paired with the Pentium 4.>
They don't offer any reason why the DDR boards beat the RDRAM boards in a few tests, namely ones that involved file transfers. Well, its pretty obvious the DDR boards had an advantage in the IDE subsystem. Its pretty obvious when you make one system run a castrated (stock) subsystem, versus a hugely overclocked subsystem, that the castrated one will lose.
In all CPU intensive tests it looks like a m00t point that the PC1066 system is killing the DDR system.
<<MadRat, the chipset is called the Granite bay. Its a chipset directly from intel hitting xeon/northwood cpus.>>
We are talking availability today, not 3Q2002.
<<GamePC's review of overclocked 1.6A's show how memory bandwith doesnt matter nearly as much in cpu intensive tests,or PC2100 DDR is providing more than it needs for those types of tasks right now.>>
GamePC's does what?? Look at there tests, none of there tests are sensical. The test rigs are such a hodgepodge they have no correllation from one system to the next. You can compare 533fsb only in the PC1066 system to the DDR400 system:
CPU Arithmetic
PC1066/P4-2.2GHz: 4154
DDR400/P$-2.2GHz: 3924
Memory Bandwidth
PC1066/P4-2.2GHz: 3306
DDR400/P$-2.2GHz: 2631
ScienceMark
PC1066/P4-2.2GHz: 133.7
DDR400/P$-2.2GHz: 131.87
3DMark 2001 1024x768x32
PC1066/P4-2.2GHz: 8623
DDR400/P$-2.2GHz: 8381
3DMark 2002 1600x1200x32
PC1066/P4-2.2GHz: 5314
DDR400/P$-2.2GHz: 5268
Valley of the Jaguars 1024x768x32
PC1066/P4-2.2GHz: 51.5
DDR400/P$-2.2GHz: 37
Quake!!! Fastest 512x384x16
PC1066/P4-2.2GHz: 343
DDR400/P$-2.2GHz: 316.7
Quake!!! High Quality 1024x768x32
PC1066/P4-2.2GHz: 241.5
DDR400/P$-2.2GHz: 232.3
Quake!!! High Quality 1600x1200x32
PC1066/P4-2.2GHz: 123.1
DDR400/P$-2.2GHz: 122.8
Return to Wolfenstein ATDemo6 - High Quality - 1024x768x32
PC1066/P4-2.2GHz: 52.9
DDR400/P$-2.2GHz: 54.4
Return to Wolfenstein ATDemo8 - High Quality - 1024x768x32
PC1066/P4-2.2GHz: 150
DDR400/P$-2.2GHz: 157.5
Photoshop 6.01 Filter Benchmark (v.2) *
PC1066/P4-2.2GHz: 93.1
DDR400/P$-2.2GHz: 93.5
Kinetix 3D Tank Model Rendering Time *
PC1066/P4-2.2GHz: 16:42
DDR400/P$-2.2GHz: 16:57
Windows Media 8.0 50MB MPEG Video to Windows Media Video Encode
PC1066/P4-2.2GHz: 3:26
DDR400/P$-2.2GHz: 3:26
L.A.M.E. 200MB Wav to MP3 File Encoding Time *
PC1066/P4-2.2GHz: 68
DDR400/P$-2.2GHz: 67
* = lower is better
<The Final Word
With that said, it's clear that RDRAM at PC-1066 speeds offers better overall performance compared to DDR400 memory paired with the Pentium 4.>
They don't offer any reason why the DDR boards beat the RDRAM boards in a few tests, namely ones that involved file transfers. Well, its pretty obvious the DDR boards had an advantage in the IDE subsystem. Its pretty obvious when you make one system run a castrated (stock) subsystem, versus a hugely overclocked subsystem, that the castrated one will lose.
In all CPU intensive tests it looks like a m00t point that the PC1066 system is killing the DDR system.
A comment on the test then on madrat...
1) the memory marks are wrong....I booted my system p4s333 using pc2700 at fsb of 117 for 2.1ghz and 395mhz ddr or very close to 400mhz ddr and I was just at 2700 memory scores...I am running 2.4ghz now with only 333mhz ddr and I get 2525...When true cas 2 pc2700 comes on line it will get better...
comment on madrat...Are you an idiot??? you have been a zealot everywhere I see you...just live with your system and let ppl make their own choices...
All I really care about is I have a system at 2.4ghz stable that is right with rdram pc800 at that speed...I am not saying I will beat a stock 533fsb 2.4ghz w? 1066, but I don't care...I am not saying I am faster then someone with oc'd rdram memory...However I will be even with a stock 2.4ghz 400fsb though, and I did it at a fraction of the price and that chip isn't even available here in the states...A 2.2ghz was right around 480-500 when I bought my chip for 208, and the likelihood I couldn't have oc' but 2.6ghz anyways
The test show exactly what christoph83 was saying...percentage differences in order
rdram vs ddr
+6%
+25%
+1%
+3%
+1%
+39% (never seen test before)
+8%
+4%
+4%
+0.1%
all the rest the ddr wins in....
Anything really less then 5% in these test is neglible...So WTF have you proved..NOthing!!!
1) the memory marks are wrong....I booted my system p4s333 using pc2700 at fsb of 117 for 2.1ghz and 395mhz ddr or very close to 400mhz ddr and I was just at 2700 memory scores...I am running 2.4ghz now with only 333mhz ddr and I get 2525...When true cas 2 pc2700 comes on line it will get better...
comment on madrat...Are you an idiot??? you have been a zealot everywhere I see you...just live with your system and let ppl make their own choices...
All I really care about is I have a system at 2.4ghz stable that is right with rdram pc800 at that speed...I am not saying I will beat a stock 533fsb 2.4ghz w? 1066, but I don't care...I am not saying I am faster then someone with oc'd rdram memory...However I will be even with a stock 2.4ghz 400fsb though, and I did it at a fraction of the price and that chip isn't even available here in the states...A 2.2ghz was right around 480-500 when I bought my chip for 208, and the likelihood I couldn't have oc' but 2.6ghz anyways
The test show exactly what christoph83 was saying...percentage differences in order
rdram vs ddr
+6%
+25%
+1%
+3%
+1%
+39% (never seen test before)
+8%
+4%
+4%
+0.1%
all the rest the ddr wins in....
Anything really less then 5% in these test is neglible...So WTF have you proved..NOthing!!!
Like you could tell the difference if you sat down to two machines side by side. One rdram and one ddr.
A faster hd or video card in one or the other would make a more noticeable increase in apps than rdram would.
I still don't see where rdram is blowing the wheels off ddr.
A faster hd or video card in one or the other would make a more noticeable increase in apps than rdram would.
I still don't see where rdram is blowing the wheels off ddr.
Exactly....too many variables and in even setups in testing like above you can see it may be difficult to tell...I like in quake and in many of the games the speeds where most are playing to work best with their monitors are 1% or less...no raving endorsement there!!!! Who plays at 480x640 with the vid cards we have now and the insane fps...can you say overkill anyways with current games???
MadRat
Lifer
- Oct 14, 1999
- 11,999
- 307
- 126
Okay, now I am a zealot for using his own DDR=RDRAM link to show exactly what I was saying about RDRAM>DDR. Great comeback. If 5% is no big deal then why do people buy 100-200MHz upgrades? If 5% is no big deal then why do people overclock past 533fsb? They do because 5% is significant.
What tests did the DDR example win? The ones that relied on the IDE transfers were the ones. The IDE was 33% overclocked on the DDR system and running stock on the RDRAM example. If you want to compare fairly then benchmark with both IDE systems at either the same speed, or both at their maximum overclocks. If they redid those tests with the RDRAM sytem using insane overclocks on the IDE then surely the RDRAM system would have won those tests, too.
btw - It was only within 5% on generous tests. If they set out to prove RDRAM>DDR then it would have been more satisfactory to see a 160MHz DDR versus a PC1066 RDRAM paired to a Pentium 4, which is more like what people are doing. We're comparing 200MHz DDR versus PC1066 in those benchmarks, which means your example of DDR is gaining 25% more than the typical overclock right now.
What tests did the DDR example win? The ones that relied on the IDE transfers were the ones. The IDE was 33% overclocked on the DDR system and running stock on the RDRAM example. If you want to compare fairly then benchmark with both IDE systems at either the same speed, or both at their maximum overclocks. If they redid those tests with the RDRAM sytem using insane overclocks on the IDE then surely the RDRAM system would have won those tests, too.
btw - It was only within 5% on generous tests. If they set out to prove RDRAM>DDR then it would have been more satisfactory to see a 160MHz DDR versus a PC1066 RDRAM paired to a Pentium 4, which is more like what people are doing. We're comparing 200MHz DDR versus PC1066 in those benchmarks, which means your example of DDR is gaining 25% more than the typical overclock right now.
You are a zealot for the reasons that I have seen in many flame wars....Some guy is commenting on how he got a nice oc and you crap all over the thread...
Whether it is your blind faith for amd, or you do like rdram (I doubt it), basically telling him, so what you have a neutered system is false....
Why don't you try to use some tact and say "hey I think the rdram platform could have given you a bit more performance for not much more cost" Then we could have debated the likelihood of rdram allowing a 160fsb oc anyways....
I agree on the 100-200mhz thing...I have jumped in increments of 400mhz and usually change in critical things like fsb or architecture...
I didn't say 5 percent isn't a big deal...In some programs like divx encoding 5 percent translates into plenty of time...seti is another one....so it really depends on a persons apps they are using...
Whether it is your blind faith for amd, or you do like rdram (I doubt it), basically telling him, so what you have a neutered system is false....
Why don't you try to use some tact and say "hey I think the rdram platform could have given you a bit more performance for not much more cost" Then we could have debated the likelihood of rdram allowing a 160fsb oc anyways....
I agree on the 100-200mhz thing...I have jumped in increments of 400mhz and usually change in critical things like fsb or architecture...
I didn't say 5 percent isn't a big deal...In some programs like divx encoding 5 percent translates into plenty of time...seti is another one....so it really depends on a persons apps they are using...
<< Yeah, I just ordered a Radeon 8500 and I might try to switch out it's tiny HSF for a large 55mm Alpha HS and see how it runs fanless. >>
Or you could just run the fan at 7v if you want to save a buck or two. I had done this on my other system with a Geforce 3 with no problems, and you can't hear it.
Hey MadRat look at that review again. You seem to be overlocking a lot of stuff.
btw - It was only within 5% on generous tests. If they set out to prove RDRAM>DDR then it would have been more satisfactory to see a 160MHz DDR versus a PC1066 RDRAM paired to a Pentium 4, which is more like what people are doing. We're comparing 200MHz DDR versus PC1066 in those benchmarks, which means your example of DDR is gaining 25% more than the typical overclock right now.
This is wrong as actually they memory is only running DDR333 speeds. I have the same Asus board and you cannot see DDR400 speeds with it. Its a i845d board not a sis645 board with PC2700. As Duvie said his memory benchmarks are wrong as my memory is running DDR333 and I have basically the same memory bandwith. The reviewer said RDRAM is overall the better performer,which is true,but did he say anywhere that DDR does not provide enough performance to be a valid option for some people? In a lot of those tests performance was neck in neck. Especially in those last cpu intensive tasks. At default speeds in that tests its the same way. These numbers are right here in front of you but yet you insist RDRAM blows the wheels off RDRAM and is not a option for a northwood setup. Build a DDR setup and then go talk trash about it.
What tests did the DDR example win? The ones that relied on the IDE transfers were the ones. The IDE was 33% overclocked on the DDR system and running stock on the RDRAM example. If you want to compare fairly then benchmark with both IDE systems at either the same speed, or both at their maximum overclocks. If they redid those tests with the RDRAM sytem using insane overclocks on the IDE then surely the RDRAM system would have won those tests, too.
Once again you dont know how some of these boards work. How was the IDE overclocked 33% when the asus board they are using locks the PCI/AGP/IDE busses at default speeds? The IDE was not faster the CPU was just running equally as fast as the RDRAM setup. 5% is something in benchmarks but its not getting its wheels blown off. You have to remember where alot of these DDR setups are coming from. Past AMD systems. So a lot of these people had DDR to begin with,grabbed a 1.6A and hit 2.4ghz and saw faster than PC800 Speeds at that clock. Dual channel DDR is coming out soon and after you always have the option of going clawhammer. I think a lot of people here,including duvie , see that you hate intel with a passion and want to pick out something to bash them about it. When a small group of people try out the 1.6A thats different,but theres almost a 300 post thread going into the cpu and overclocking forums of people overclocking these things like mad and seeing great performance out of them. But yet all you can talk about is how RDRAM is blowing the wheels off DDR by winning tests in 3dmark2001 and quakeIII at 640x480. If you want to see that extra performance play quake III at 640x480 all day.
btw - It was only within 5% on generous tests. If they set out to prove RDRAM>DDR then it would have been more satisfactory to see a 160MHz DDR versus a PC1066 RDRAM paired to a Pentium 4, which is more like what people are doing. We're comparing 200MHz DDR versus PC1066 in those benchmarks, which means your example of DDR is gaining 25% more than the typical overclock right now.
This is wrong as actually they memory is only running DDR333 speeds. I have the same Asus board and you cannot see DDR400 speeds with it. Its a i845d board not a sis645 board with PC2700. As Duvie said his memory benchmarks are wrong as my memory is running DDR333 and I have basically the same memory bandwith. The reviewer said RDRAM is overall the better performer,which is true,but did he say anywhere that DDR does not provide enough performance to be a valid option for some people? In a lot of those tests performance was neck in neck. Especially in those last cpu intensive tasks. At default speeds in that tests its the same way. These numbers are right here in front of you but yet you insist RDRAM blows the wheels off RDRAM and is not a option for a northwood setup. Build a DDR setup and then go talk trash about it.
What tests did the DDR example win? The ones that relied on the IDE transfers were the ones. The IDE was 33% overclocked on the DDR system and running stock on the RDRAM example. If you want to compare fairly then benchmark with both IDE systems at either the same speed, or both at their maximum overclocks. If they redid those tests with the RDRAM sytem using insane overclocks on the IDE then surely the RDRAM system would have won those tests, too.
Once again you dont know how some of these boards work. How was the IDE overclocked 33% when the asus board they are using locks the PCI/AGP/IDE busses at default speeds? The IDE was not faster the CPU was just running equally as fast as the RDRAM setup. 5% is something in benchmarks but its not getting its wheels blown off. You have to remember where alot of these DDR setups are coming from. Past AMD systems. So a lot of these people had DDR to begin with,grabbed a 1.6A and hit 2.4ghz and saw faster than PC800 Speeds at that clock. Dual channel DDR is coming out soon and after you always have the option of going clawhammer. I think a lot of people here,including duvie , see that you hate intel with a passion and want to pick out something to bash them about it. When a small group of people try out the 1.6A thats different,but theres almost a 300 post thread going into the cpu and overclocking forums of people overclocking these things like mad and seeing great performance out of them. But yet all you can talk about is how RDRAM is blowing the wheels off DDR by winning tests in 3dmark2001 and quakeIII at 640x480. If you want to see that extra performance play quake III at 640x480 all day.
MadRat
Lifer
- Oct 14, 1999
- 11,999
- 307
- 126
<<This is wrong as actually the memory is only running DDR333 speeds. I have the same Asus board and you cannot see DDR400 speeds with it.>>
Hey, if the website is this far wrong then why did you point it out?
<<Once again you dont know how some of these boards work. How was the IDE overclocked 33% when the asus board they are using locks the PCI/AGP/IDE busses at default speeds?>>
Asus seems to say they are adjustable, not fixed.
<<I think a lot of people here, including Duvie, see that you hate intel with a passion and want to pick out something to bash them about it.>>
First you say I hate DDR memory. Now I hate Intel? Both are news to me.
<<But yet all you can talk about is how RDRAM is blowing the wheels off DDR by winning tests in 3dmark2001 and quakeIII at 640x480. If you want to see that extra performance play quake III at 640x480 all day.>>
Do I detect a squirm? Excuse me, but the RDRAM system is better at gaming in the higher resolutions, too. The DDR example won when it came to tests reliant on IDE transfers, plain and simple. Everything else was a RDRAM win. RAMBUS may suck, but the RDRAM has grown wings. The Pentium 4 is best used above 2GHz with RDRAM, and the tests have shown up to my claim.
Hey, if the website is this far wrong then why did you point it out?
<<Once again you dont know how some of these boards work. How was the IDE overclocked 33% when the asus board they are using locks the PCI/AGP/IDE busses at default speeds?>>
Asus seems to say they are adjustable, not fixed.
<<I think a lot of people here, including Duvie, see that you hate intel with a passion and want to pick out something to bash them about it.>>
First you say I hate DDR memory. Now I hate Intel? Both are news to me.
<<But yet all you can talk about is how RDRAM is blowing the wheels off DDR by winning tests in 3dmark2001 and quakeIII at 640x480. If you want to see that extra performance play quake III at 640x480 all day.>>
Do I detect a squirm? Excuse me, but the RDRAM system is better at gaming in the higher resolutions, too. The DDR example won when it came to tests reliant on IDE transfers, plain and simple. Everything else was a RDRAM win. RAMBUS may suck, but the RDRAM has grown wings. The Pentium 4 is best used above 2GHz with RDRAM, and the tests have shown up to my claim.
The asus P4B266 locks the PCI and AGP buses with turbo 1 mode in the BIOS. The thread in cpu's and overclocking explains this. The only thing ive seen a noticable difference in RDRAM and DDR at 1024x768 is that valley of the jaguars from that gamepc review, everything else is less than 5% and most of it is on par,ie quake III.Notice the Return to wolfenstien benchmark? DDR was actually faster there. The website made a small miscalculation on what the 3:4 option does on the Asus P4B266 towards the beggining of the review,doesnt mean the benchmarks are less valid. Your miss-interepting a lot of information without a hands on expierience with a DDR setup.
MadRat
Lifer
- Oct 14, 1999
- 11,999
- 307
- 126
<<The asus P4B266 locks the PCI and AGP buses with turbo 1 mode in the BIOS. The thread in cpu's and overclocking explains this. The only thing ive seen a noticable difference in RDRAM and DDR at 1024x768 is that valley of the jaguars from that gamepc review, everything else is less than 5% and most of it is on par,ie quake III.>>
Wait a minute, I just showed you the listing from Asus that states they are variable. The benchmarks seem to lean towards the DDR example as having an overclocked PCI bus, IDE channel, and AGP port. You insist on this not being the case, claiming they are locked at their standard 33/66MHz speeds. Its pretty hard to take what you claim as fact when you've already said the website's testing is full of inaccuracies.
You seem to be suggesting that they are running in the mode where the safeguards kick in. From what I understand on locking the PCI/AGP speed of the Asus boards you have to be at 533fsb or above. Yet the testors intentionally stay a hair below 533fsb just to avoid the safeguards, so its unlikely they have the PCI/AGP running at 33/66MHz.
<<The website made a small miscalculation on what the 3:4 option does on the Asus P4B266 towards the beggining of the review,doesnt mean the benchmarks are less valid.>>
Its either A=B or A<>B. What you are saying is that A<>B but that this disimilarity "doesn't matter". Explain how something can be compared if its disimilar.
<<Your miss-interepting a lot of information without a hands on expierience with a DDR setup.>>
Did I say I do not have a DDR setup, or have never used one? Keep on topic, none of this misdirection crap. What I/you own and/or use has nothing to do with this debate.
Wait a minute, I just showed you the listing from Asus that states they are variable. The benchmarks seem to lean towards the DDR example as having an overclocked PCI bus, IDE channel, and AGP port. You insist on this not being the case, claiming they are locked at their standard 33/66MHz speeds. Its pretty hard to take what you claim as fact when you've already said the website's testing is full of inaccuracies.
You seem to be suggesting that they are running in the mode where the safeguards kick in. From what I understand on locking the PCI/AGP speed of the Asus boards you have to be at 533fsb or above. Yet the testors intentionally stay a hair below 533fsb just to avoid the safeguards, so its unlikely they have the PCI/AGP running at 33/66MHz.
<<The website made a small miscalculation on what the 3:4 option does on the Asus P4B266 towards the beggining of the review,doesnt mean the benchmarks are less valid.>>
Its either A=B or A<>B. What you are saying is that A<>B but that this disimilarity "doesn't matter". Explain how something can be compared if its disimilar.
<<Your miss-interepting a lot of information without a hands on expierience with a DDR setup.>>
Did I say I do not have a DDR setup, or have never used one? Keep on topic, none of this misdirection crap. What I/you own and/or use has nothing to do with this debate.
Well let me tell you why I'm building my friend a DDR based P4 system. I originally was looking to build him a RDRAM based system but I saw a few problems with it. First off, I wanted to build him a system that would be completely stable overclocked and had two choices to go with Asus P4T-E or Abit TH7-II. Well the P4T-E is just a big pain in the ass to get a board with teh right clock generators on top of that, if you do get the right clock gens, you'll have to do dip switch overclocking PLUS do a wire trick to bump the voltage. OK, screw that, too much trouble even if I was building the system for myself, and this is coming from a guy that owns a P4T-E. Well the next viable option is the TH7-II. This is a great board but now you have to hope your PC800 can hit PC1066 speeds. Alot of people are hitting PC1066 speeds but that is with 128MB sticks and even with 128MB sticks, you have to look for a certain types of PC800 that would give you a better chance of doing 533MHz FSB. WAAAAAAAY to much time being wasted here. I also want 256Mb sticks as 512MB is not enough memory for me. If they came out with 256MB PC1066 sticks, it would be a clear choice for me.
Now with a DDR system, all I need is a P4B266 board and some PC2700 memory and I am good to go. I ordered a P4 1.8A GHz chip and some Samsung PC2700 and I'm hoping to hit 2.4GHz and see if I can run my memory at DDR333 speeds. We'll see...
Now with a DDR system, all I need is a P4B266 board and some PC2700 memory and I am good to go. I ordered a P4 1.8A GHz chip and some Samsung PC2700 and I'm hoping to hit 2.4GHz and see if I can run my memory at DDR333 speeds. We'll see...
Did I say I do not have a DDR setup, or have never used one? Keep on topic, none of this misdirection crap. What I/you own and/or use has nothing to do with this debate.
I never said you did. I said your mis-interepting informtation because I own a Asus P4B266 and It does infact lock the PCI and AGP busses up to atleast a 133FSB. Ask any Asus P4B266 user here. Infact why not go read the thread ive been talking about in cpu's and overclocking and see for yourself instead of blocking it out. Your pulling what you want out of reviews. There was no boost to the IDE performance.
Its pretty hard to take what you claim as fact when you've already said the website's testing is full of inaccuracies.
I never said this. If you would have read the beggining of the of the review you would have saw where they made a miscalculation. The memory benchmarks show they are running DDR333 instead of DDR400 . This doesnt mean the rest of the review is screwed up. People make miscalculations all the time just like you did when thinking DDR333 runs 3.0gb/s. This is the closest you can get to comparing defualt DDR vs default RDRAM and overclocked DDR to overclock RDRAM. The fact is now DDR is the easiest way to overclock a 1.6A to 2.4ghz. You'll have memory running RDRAM speeds. DDR hangs very close in most applications but overall RDRAM will be faster at overclocked and default speeds. DDR is a much easier overclock because its already running 133FSB. Either way you go your going to have a very,very fast setup.
I never said you did. I said your mis-interepting informtation because I own a Asus P4B266 and It does infact lock the PCI and AGP busses up to atleast a 133FSB. Ask any Asus P4B266 user here. Infact why not go read the thread ive been talking about in cpu's and overclocking and see for yourself instead of blocking it out. Your pulling what you want out of reviews. There was no boost to the IDE performance.
Its pretty hard to take what you claim as fact when you've already said the website's testing is full of inaccuracies.
I never said this. If you would have read the beggining of the of the review you would have saw where they made a miscalculation. The memory benchmarks show they are running DDR333 instead of DDR400 . This doesnt mean the rest of the review is screwed up. People make miscalculations all the time just like you did when thinking DDR333 runs 3.0gb/s. This is the closest you can get to comparing defualt DDR vs default RDRAM and overclocked DDR to overclock RDRAM. The fact is now DDR is the easiest way to overclock a 1.6A to 2.4ghz. You'll have memory running RDRAM speeds. DDR hangs very close in most applications but overall RDRAM will be faster at overclocked and default speeds. DDR is a much easier overclock because its already running 133FSB. Either way you go your going to have a very,very fast setup.
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