![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
-
Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.
AM2 and 4 mem modules?
- Thread starter O-6
- Start date
(cf)Eclipse
Freelance Overclocker
- Joined
- Feb 18, 2003
- Messages
- 30,027
surprisingly hasn't really been said yet. i'm assuming it will though
Tetrahedron
[H]ard|Gawd
- Joined
- Jun 24, 2004
- Messages
- 1,579
bandwidth > speed
defaultluser
[H]F Junkie
- Joined
- Jan 14, 2006
- Messages
- 14,398
O-6 said:
That was actually only ealier boards. Most recent 939 boards, coupled with a Venice core processor, will run 4 sticks at DDR 400, with a 2T command rate. It does cause a %1-2 performance hit in games and applications, and about a %5-10 performance hit in certain benchmarks.
But we really have no idea what the AM2 brings to the table. How about you be bold and daring, and find out for us
banGerprawN
[H]ard|Gawd
- Joined
- Apr 15, 2005
- Messages
- 1,113
The AM2 MC suffers the same problem that the X2s and Venice-> onwards processors did : It can only support four dimms at 2T, and having all 4 slots populated lowers it's overclocking potential (including lowering latencies to the level that was possible with only 2 dimms.)
(cf)Eclipse
Freelance Overclocker
- Joined
- Feb 18, 2003
- Messages
- 30,027
people still believe that eh? perhaps you should take a look at this. AMD HAS NOT changed the spec to 200mhz at 2T for 4 double rank modules.defaultluser said:That was actually only ealier boards. Most recent 939 boards, coupled with a Venice core processor, will run 4 sticks at DDR 400, with a 2T command rate. It does cause a %1-2 performance hit in games and applications, and about a %5-10 performance hit in certain benchmarks.
But we really have no idea what the AM2 brings to the table. How about you be bold and daring, and find out for us
http://www.amd.com/us-en/assets/content_type/white_papers_and_tech_docs/26094.PDF
go to page 182
linkage to where you have discovered this?banGerprawN said:
burningrave101
[H]F Junkie
- Joined
- Sep 9, 2003
- Messages
- 11,825
Tetrahedron said:
bandwidth = speed
Tetrahedron
[H]ard|Gawd
- Joined
- Jun 24, 2004
- Messages
- 1,579
burningrave101 said:
sorta..
10 lane highway speed of 40mph = 40x10 = 400
20 lane highway speed of 30mph = 20x30 = 600
i would rather have more bandwidth and lesser speed.. allows for more throughput overall.
banGerprawN
[H]ard|Gawd
- Joined
- Apr 15, 2005
- Messages
- 1,113
It took me ages to find, I couldn't remember anything from the article apart from the "memory heatsink" the retarded writer did at the start.(cf)Eclipse said:linkage to where you have discovered this?
http://www.legitreviews.com/article/347/4/
The bottom paragraph...
of all the problems i have seen with AM2 the memory downclocking is not one of them, i have put 4x1GB DDR2-667 PC2-5300 and i had not noticed any slow downs with XP32bitO-6 said:
we have not transfered to 64bit yet as that is not what the mass market uses
other problems with AM2 the FX-62's are picky on the memory brands, even though no one will tell u that
(cf)Eclipse
Freelance Overclocker
- Joined
- Feb 18, 2003
- Messages
- 30,027
nice, thanks. that's good to knowbanGerprawN said:
but atm with DDR2, K8 prefers bandwidth to be sacrificed a bit for "speed", or overall latency. there's so much bandwidth, the core doesn't know what to do with it. lower latency will be the way to go to get the most performance.Tetrahedron said:
(cf)Eclipse
Freelance Overclocker
- Joined
- Feb 18, 2003
- Messages
- 30,027
yes, but the extreme majority of 512mb and all 1gb modules are double rank, so that point is kinda moot, unless you like using 256mb modulesdrizzt81 said:
(and that is also the spec for non-revE memory controllers, as shown by page 179)
look a wee bit further down. you see that 4 double rank is still ddr333
Tetrahedron
[H]ard|Gawd
- Joined
- Jun 24, 2004
- Messages
- 1,579
Spazilton said:
yeah like who really uses all of 2GB of memory that much anyways.. I dont so... so why get 2GB or even 4GB?
even when I have photoshop, 3ds max, media player, crytek sandbox editor, browser (with multiple tabs open), trillian with all of em cranky out renderings 2GB of memory is not being used.
so yeah.. I dont even fill my highway up with cars with 2x1GB modules.
Tetrahedron said:
Wrong again. You're using Windows, which pages a lot to keep memory free. Use another OS that pages less, keeps more in memory, and you'll see why having more memory is important (ask Trovalds). Regardless, a single game, like Battlefield 2 benefits from an extra GB of memory. As for 4 GB, use a 64 bit OS, or a useful 32 bit OS (Win2k3 Server, Linux, etc) and you'll see that you can benefit from that as well, depending on the application (server apps, high load density apps such as running a VXML browser, etc.).
All of this is actually off topic and not at all what the other guy was commenting on. What he said is fill up the high way -- as in THROUGHPUT, not static data filling the memory. He's talking about data being written and read from the memory. This is bandwidth, the amount of information read and written from memory..usually measured at its peak amount.
This doesn't make any sense. Bandwidth is a measure of maximum combined speed, so what the hell do you mean? I could come up with an arbitrary scenario where this isn't true, which is what Spazilton was trying to say.
Great, I have 10 cars to send from pt. A to pt. B. They're limited by the speed of their highway...so I send them on your first highway and they get there in X seconds, then I try on your second highway and they get there in X + C seconds, since they had to go slower. Therefore, the scenario fails. Basically, I've looked at the worst case scenario and flipped your example to show that the reverse can be true.
We can look at the other side as well, if your architecture is memory bandwidth starved and has a long pipeline (e.g. p4), then you need high memory bandwidth and latency isn't as important b/c the processor doesn't spend much time accessing memory to begin with (it can't, since it might only access memory 1/24 of the time, for example, depending on several factors). If you have a shorter pipeline with relatively frequent memory access (e.g. A64), then latency becomes more important than in the prior scenario because the processor will spend more time accessing memory. This is why there isn't some drastic difference in performance w/ the AM2 A64s w/ DDR2.
If your theory were true, there would be a massive performance increase w/ the switch to DDR2 on the A64. They did manage to keep any possible performance hit from the higher latency of DDR2 minimal with the new memory controller, however.
Cliff Notes: More bandwidth is good, but if you can't use it, you might as well work on latency to get data from place to place more quickly.
(cf)Eclipse said:nice, thanks. that's good to know
but atm with DDR2, K8 prefers bandwidth to be sacrificed a bit for "speed", or overall latency. there's so much bandwidth, the core doesn't know what to do with it. lower latency will be the way to go to get the most performance.
Yep... that's like building a 20 lane highway when a 2 lane highway with a higher speed limit would perform better overall.
(cf)Eclipse
Freelance Overclocker
- Joined
- Feb 18, 2003
- Messages
- 30,027
sadly true.drizzt81 said:
GilmourD
[H]F Junkie
- Joined
- Jun 17, 2004
- Messages
- 11,185
That would apply here... If only you could change the bit width of the memory channels. You can't, so you've got to work with 128bits in dual channel.Tetrahedron said:
Therefore, considering the limitation, speed DOES equal bandwidth.