hard drives & raid - benchmark and compare!

[red930]

The only way an SATA drive can be faster on an SATA II controller, is if the old controller was not working at 100%. The speed of the new rig could be giving you a false impression of performance.

 

I believe you are right, but also SATA I drives could be faster when running with the new SATA controller on the nF4 chipset. I am not sure, but it feels as if it was improved... I have no way of proving this theory as I only tried a SATA I HDD once when I was testing the motherboard, not sure if the faster benchmarks were result of newer Windows XP or the nF4 controller was more eficient.

[Angry_Games]

did a little more benching with 3x SATA II drives on the new AMD Athlon64 X2 4400+ dual core (2.2Ghz with 2 independent 1MB L2 caches)

 

 

 

32k stripe @ default XP cluster size, 2530Mhz

[Sorrento]

The only way an SATA drive can be faster on an SATA II controller, is if the old controller was not working at 100%. The speed of the new rig could be giving you a false impression of performance.

 

Yes, thats the main theory right now... unless the nforce3 Ultra sata controller was, like you said, not working at full eficiency back then.

 

did a little more benching with 3x SATA II drives on the new AMD Athlon64 X2 4400+ dual core (2.2Ghz with 2 independent 1MB L2 caches)

 

Nice, the AMD64 Dual cores are really fast. Does the HD Tach burst speed increase with more HDD's, or just with the spread spectrum enabled?

[EL_]

Angry can you update the first post to an article style summary with links and procedures? I think the thread has tons of info but most won't go thru the 15+ pages to find all the great work you and others have done.

[Angry_Games]

Yes, thats the main theory right now... unless the nforce3 Ultra sata controller was, like you said, not working at full eficiency back then.

 

it was working at full efficiency. it did not support the SATA II functions, so would never be as efficient or as fast as a true SATA II controller coupled with a true SATA II drive (and we really are not seeing TRUE sata II drives yet, most are still hybrid drives)

 

Nice, the AMD64 Dual cores are really fast. Does the HD Tach burst speed increase with more HDD's, or just with the spread spectrum enabled?

 

the cpu itself has nothing to do with the RAID speeds. Burst increases the more drives you use...you getting a combined cache of 24MB with my 3 drives, 32MB with 4 drives

[Sorrento]

the cpu itself has nothing to do with the RAID speeds. Burst increases the more drives you use...you getting a combined cache of 24MB with my 3 drives, 32MB with 4 drives

I was complementing you on the dual core, and then changing the subject to know what increased the burst speed... this because I remember an earlier post about this and Spread Spectrum enabled on the Hitachi SATA II HDD, and how this increased the burst speed in the HD Tach result.

 

I am asking because I have four Hitachi @ RAID-0 and I get only a little faster burst speed than you with those three Hitachi. Hehe, maybe I got something wrong and the burst speed should be near 400

[Angry_Games]

oh heh, its been a crazy day so i probably didnt read your post as a compliment to the cpu then a subject change lol sorry

 

4 drives should burst better, but not by a lot...i think i only burst about 327-330MB/s with 2 drives, and my 343-347MB/s bursts with 3 drives would put your bursts probably right in line with 4 drives.

 

then again, this dual core cpu is not like any cpu before it (it is in a way, but its two complete cpu cores and that might make a bit of difference...not sure).

 

I will end up retesting 3 drives on the new DAGF (unmodified) and another A02 LP NF4 Ultra without a dual core cpu and see what happens...if i can ever clear out this logjam of filled up test benches =/

 

[Sorrento]

oh heh, its been a crazy day so i probably didnt read your post as a compliment to the cpu then a subject change lol sorry

What about "No hablo Inglés" didn't you understand (below name)??? :shake:

 

Arghh, a friend keeps telling me she wants a dual core... I keep telling her single core CPU's are still, and will be, better. Was I right by saying that???

 

There has to be something on which a faster CPU is involved in access time and how the chipset works... if the memory controller was on the motherboard, like the old nForce2 Ultra400 or the Intel ones a faster CPU had faster memory bandwiths; maybe this applies to the RAID performance too. Diference will be hardly noticeable of course, today's systems are very efficient in every aspect.

[Angry_Games]

Arghh, a friend keeps telling me she wants a dual core... I keep telling her single core CPU's are still, and will be, better. Was I right by saying that???

 

lol hell no

 

P4's can beat the A64 in any multithreaded test (so depends on what type of software she gonna use mostly)

 

the X2 A64 can annihilate the P4 in multithreaded tests as the cpu is actually SMP enabled, not just Hyperthreaded (google is your friend!)

 

THe A64 has a superior fpu unit and has shorter pipelines which is why it dominates in games (less instruction time) and any single-threaded application, and can do it nearly 1000Mhz slower than the P4.

 

Intel has 800Mhz and 1066Mhz FSB's...which are faster than AthlonXP's 400FSB max...

 

but

 

the A64 doesn't really have such a thing as FSB...it's memory controller can run as fast as your memory can, and faster. You simply cannoy possibly feed the memory controller on the A64 enough memory to clock it like you can a standard cpu that has to have a separate NB (memory controller) between the cpu and memory. With the A64, the 'NB' is integrated into the cpu's core....!

 

there is no way any cpu with an integrated memory controller should get outperformed by a cpu that requires a totally separate memory hub (NB).

 

now, that all being said...a 2800Mhz single core A64 would be better than a 2200Mhz dual core in SOME things...but with both Intel (already using Hyperthreading, now going dual core themselves) and AMD using multiple cores/threads, software will now evolve to be at minimum 'multithreaded' and more than likely we will begin seeing 64-bit SMP-enabled software...

 

 

*whew!*

[Sorrento]

now, that all being said...a 2800Mhz single core A64 would be better than a 2200Mhz dual core in SOME things...but with both Intel (already using Hyperthreading, now going dual core themselves) and AMD using multiple cores/threads, software will now evolve to be at minimum 'multithreaded' and more than likely we will begin seeing 64-bit SMP-enabled software...

 

 

*whew!*

Nice explanation

 

Yes I know the A64 Dual Core is better than the Intel's Dual Core, by a lot according to some reviews. Perhaps the Intel cpu is still better at certain encoding aplications, yet I really don't care about how fast at encoding can a CPU be... because the "slower" one will be as close as few seconds behind.

 

Problem for the AMD Dual Core is the price... they actually have to be faster than their Intel counterparts, as one cost $500+ while the other cost $200+... I am not sure about the prices, but I am about the A64 costing more.

 

So, if we consider the price of these Dual Core units, and then compare it to a single core FX55 (or 57)... would you still choose the dual core for the present times? I am sure in a couple of years things will be diferent, perhaps... still I see more than half of Mac's aplications using a single CPU, and these guys have been using dual cpu's for quite some time now. Lets see what Unreal Tournament, for Dual Cores, can do to the market... and in the meanwhile, throw that Dual Core away... just let me know where "away" is :nod:

[Angry_Games]

Problem for the AMD Dual Core is the price... they actually have to be faster than their Intel counterparts, as one cost $500+ while the other cost $200+... I am not sure about the prices, but I am about the A64 costing more.

 

what you need to go back and read is the major differences in the cpu's...Intel's are using shared cache, and are not per-core hyperthreaded for the lower costs ones that you cite...the ones that actually have decent cache AND have the ability to hyperthread per-core...are much more in the range of the AMD X2 cpu's...and they still cannot compare really...AMD did it right by placing independent core/caches on a single die, and Intel did some hybrid crap that is really a weaker design by what I have read. The AMD is still a better choice hands-down. (dont get me wrong, those who know me know I'm a huge Intel fan but I am unbiased between the two and always give my honest assessment)

 

So, if we consider the price of these Dual Core units, and then compare it to a single core FX55 (or 57)... would you still choose the dual core for the present times?

 

sure i would....assuming they were readily available (which they are not and wont be for a bit). Think of the dual-core cpu's as a race car with 2 engines running in synch...around town and such (ie: present day computing) you really only need 1 engine running, but since they both run in synch for the most part, you use only a little bit of both.

 

then you get out to the race track (upcoming software titles that will take advantage of multiple cores) and holy crap most are using their single engine hotrods (FX-55 for example) and can crank up the speed on them...but while their single engine's top speed/acceleration may be larger than either one of your engines by itself, you got to remember you have TWO engines (think SLI also). When you get a on a racetrack that is made to really handle your super fast two-engine racing machine...there's no single engine car that can even match yours.

 

 

think of it as the same as SLI like I said above...a single X850XT is going to freakin rock (trust me, I have one!).

 

then along comes this dual 6600GT...and either 6600GT by itself gets crushed by the mighty X850XT...but when you run both of those 6600GT's together in SLI...its a whole different story...the X850XT might have more pure raw rendering power (ie it will dominate SLI 6600GT's in simple graphics like 3dmark01 doing Dx8), it is when those real tough benches like 03, 05, and aquamark somewhat really kick into gear with not only raw geometry, but lighting, particles, transparency, etc that the 2 6600GT's working together in SLI mode start to shine...6600GT SLI's are about half to a third less than a single X850XT...and can do just as well if not better (my 6600GT's can beat my X850XT PE in 03/05 and in aquamark especially with 76.41 drivers that have profiles for major games/benches that sets the correct rendering modes etc).

 

the same will be said of software or games...games are almost all single-threaded applications...but now with SLI and whatever ATI's dual-gpu solution is, we are going to definitely see entrenched single-threaded code, but more and more either patches or the beginning front-end design of a game's graphics engine will be built to take advantage of dual rendering gpu's working together in SLI states...

 

cpu's will be no different...most encoding applications run better on the P4 since it is hyperthreading enabled, and encoding different tracks/channels can take advantage of this since the code does a 'hyperthreading enabled? if yes - use this branch of code (leading to multithreads) , if no - use this branch of code' (leading to single threads). Thats oversimplifying it a lot, but the basics are there and easy to grasp.

 

very soon we will see in games and apps what we are going to see very soon in the new next-gen consoles...multiple cores/multiple threads on the hardware, and the software devs are going to damn sure take advantage of it....

 

how about 1 core running all the rendering and base engine code for a game, and the second core running the physics and AI? How about on the P4's real expensive cpu's that have HT running on each logical core....1 core splits base engine code to one thread on core 1, and runs the rendering geometry calculations to send to the gpu(s) on the second thread of core 1, while on core 2 you have the first thread of core2 running pure physics calculations and the second thread of core 2 running all of the AI routines...

 

booyah!

 

sure...it will take a while...and by the time it REALLY comes into mainstream, we might have moved on to the socket1990 with quad SMP enabled cores with HT on each logical core (ie a total of 8 simultaneous threads calculating at once) that runs at 4Ghz per core (independent)...

 

but not everyone has the money to just purchase a new computer every time something new comes out...and I am one of them (even if I didnt have this job). I buy my computers like I do cars...the instant i purchase it, its already outdated and worth much less than what I paid for it...but if its a good car then it will last me 2-3 years (for a car, I need them to last 5-7 years, but cars dont evolve as fast as personal computer technology does) until I can afford to get another 'awesome' replacement.

 

so spending $800 or so on an FX single core cpu to have the fastest single core that money can buy is ok I suppose...but I would rather purchase a dual core that is very much future ready (and remember, software that runs on x86 Windows platforms, unless something major happens, will continue to be backwards compatible) and when everyone's games start chugging along even though they have a nice FX cpu, yours might not since that second core can kick in and pick up the slack very nicely.

 

Macs = not even a good comparison lol...Macs are in less than 5% of american households and maybe about 8%-10% of business...they are not good to compare with.

 

Entertainment and Productivity software on the Windows platform (and getting to be some Linux which really takes advantage of SMP-enabled software and 64-bit software even now) is what drives the market, and gamers always demand better faster bigger prettier games, and regular users who use it for home/work always want faster, easier, etc (remember, that extra core can always kick in in future software, multitasking is always better).

 

Games are starting to drive the business of software, but right now productivity software is the king...Photoshop, MS Office, things like that...MS and Adobe and many others are already working on multithreaded applications (anything that is built specifically with the P4 in mind is already multithreaded) and SMP enabled applications...

 

like most things...once MS commits to something, and starts favoring it themselves, everyone else follows along (even Intel and AMD) because....well because they have to really...if you only make Mac or Linux software, you missing about 80% of the total market audience in which to sell your product to. So everyone else sees MS touting 64-bit operating systems and windows itself built or patched around multithreaded/SMP enabled functions and decides that they have to follow to take advantage of the hardware and underlying software to hvae the best product.

 

 

 

so in short answer lol...yes i would still take the dual core over even an FX-57.

 

the dual core would cost less, and would outlast the more expensive single-core in the long run (plus it runs cooler and has more gear left in the box when the going gets tough)

[Sorrento]

Do you sell cars by any chance???

 

You convinced me, specially since you explained the issue with shared cache on the Intel and how the X2 A64 has cache for each core. That example with SLI and the comparison between dual 6600GT's and a single x850 XT PE is scary, I already knew how fast SLI is as I have dual 6800 GT's on my other system... but I still saw the x850 XT PE as a very tough cookie to crack, since there was a review that had it beating dual 6800GT's in HL2... yet I always felt it was with 71.84 or previous drivers. The shared cache on the inexpensive Intel Dual Core explains a lot... I thought AMD's design was the same, shared cache, but it isn't and that explains the price and the extra performance.

 

So the A64 Dual Core is socket 939 and totally compatible with the DFI nF4? There is a new socket coming in 2007, M I think (around 1024 pins)... which means its time to start taking everything the 939 can give. I hope the dual cores are available soon... cause after all that blah blah you just wrote, I want mine :shake:

 

Also, are those 76.41 drivers good? I haven't tried them, I think I used everyone except those and had bad results, so I returned to 71.89 to avoid problems. Should I try these? What problems did you encountered with them?

 

And it is FX55 San Diego core, the FX57 doesn't exist nor will be... unless its dual core too I guess.