A Look At AMDs Socket AM2 Platform

Page 6 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Toggle sidebar Toggle sidebar
I had the same thought. To me it makes more sense to do that. But I think AMD wants to wait and see how the new Core Duo's perform in the desktop arena. My thinking is they'll only give Athlon 64's, X2's, etc 400mhz FSB's if they have to.
Click to expand...

:?: I dont get it :?:
FSB and AMD CPUs. Huh ? No FSB there any more since the K8... There is a Hypertransport reference @200 Mhz, most people are referring falsely to as "FSB". But if you mean that, why raise it to 400 ?? The Hypertransport speed is already at 1000 MHz ... I dont see any performance increases, if you raise the reference freq, and lower the multiplier to get HT1000.

Furthermore to that HT333 topic ... I personally never believed that. Go to hypertransport.org and check the whitepapers, specifications etc. The reference freq. is 200 Mhz. Every hypertransport device has to support that frequency. Anything else in *incompatible*.

Thus, a HT333 would be incompatible, too. Of course AMD can do that, most overclockers do, but why should AMD overclock itself ? To get that fancy "1:1" divider for DDR2-667 speeds ? Nope, first there is no 1:1 divider at all, as the memory freq. is derived with a divider from the CPU clock. Second, if you would do it, what would happen if you wanna use DDR2-800 modules ??
Easiest way for AMD to do it, is like DDR333 is achieved nowadays.

In conclusion, it is obvious, that everything else than a 200
Mhz reference freq. is not needed, and totally nonsense.

When I first heard that HT333 rumours, I though about that these might necessary for testing DDR2-667, because early engineering CPUs might not have the new, needed mem dividers, yet. Now, after I read that "nice" test @DDR2-400 speeds here, I am even more convinced.

At Xtremesystem.org somebody with access to AM2 chips also states, that due to the reported performance bug, memory is running only at DDR2-400 speeds. So all in all it is still not 100% sure, as I do not know how reliable is that Xtremesystem source. But I am convinced 90%, that the given numbers at THG are DDR2-400 ones.

byebye

FalseInformation
 
Yes, there is no 3-Bus FSB in the A64 arthitecture, but the HT (HTT is HyperThreading Technology, to some at least) is the replacement for the FSB in A64's. The point in raising it to 400 is to increase the speed of the internal clock and external clock, thus improving overall performance. If you raise the FSB of an A64, sorry, HT, to say, 250MHz, and reduce the HT from 5x to 4x, even if you can get the internal clock speed to the same (2.2GHz to 2.25GHz, for example), it will still outperform a 200MHz, simply because it is a faster crystal clock.

I do not remember where I read the 333MHz HT, but I was certain AMD said it themselves, but I could be wrong. I wonder what a 1GHz CPU w/o Clock Stepping would be like, you think it'd be faster then a 4GHz P4 or an FX? That's an interesting thought.

~~Mad Mod Mike, pimpin' the world 1 rig at a time
 
...If you raise the FSB of an A64, sorry, HT, to say, 250MHz, and reduce the HT from 5x to 4x, even if you can get the internal clock speed to the same (2.2GHz to 2.25GHz, for example), it will still outperform a 200MHz, simply because it is a faster crystal clock.
Click to expand...

The only performance gain I can see here is that the RAM is also running at 250 MHz, as long as you do not change the default multiplier. Are you referring to that ? Otherwise there is no performance gain, where should it come from ? No device directly run at the reference speed, there is always a multiplier.

byebye

FalseInformation
 
I am not saying it should, I am saying it does, because I have tested it personally on multiple CPU's setting the RAM to make sure it stays at around 200MHz (give or take a Megahert), the performance gain may not be huge, but it's there. Not to be insulting, but I think you're a bit ignorant on the CPU/FSB/RAM thing, and I mean all the respect I can in saying that.

~~Mad Mod Mike, pimpin' the world 1 rig at a time
 
Someone should try to overclock the HT frequency to 333 Mhz, to achieve some DDR2-667 timing ...
But I guess there is no possibiliy on a engineering board, to do that.
FalseInformation
Click to expand...

I would actually assume that the Multiplier on a prototype processor would be
unlocked. AMD has to have tested the processor with DDR2 800 also. The processor probably doesn't work at all with DDR2 800, so they will not give out boards for review that would allow such a thing.

The truth is that this is just a platform change. These things never have huge performance increases right away. Look at the last Intel platform migration. However, by waiting for official DDR2 800 support, when AMD starts selling these things we may actually see a larger increase than we did when Intel changed to DDR2.

I find it rather interesting to see how AMD is conducting themselves now that they have a significant performance lead, with no real threat from Intel in the forseeable futere. Being in this postition has allowed them the luxury of waiting to release this new platform. What I am really waiting for is to see how a QUAD core processor on Windows Vista performs. I am actually not all that familiar with Vista, does anyone know if it is a native 64-bit os or if there will be 2 versions like there are now? I really think that Microsoft should make Vista a 64bit only os. That would of course force anyone with an older computer4 to upgrade. Video game console left 32 programming a decade ago, and it's about time pc make that change. The best way to do it, would be if Microsoft forced everyone to use 64 bit.
 
According to M$, the first Vista (This Christmas) will be 32-bit and 64-bit, Vista 2 (unknown Release) will be 64-bit only, probably Blackcomb (new Desktop OS in 2008ish) will be the 64-bit only. I used the Beta for Vista 1 on 32-bit and 64-bit, I wasn't really testing anything, but the 64-bit didn't seem a whole lot faster than XP, than again, I run XP Pro x64.

~~Mad Mod Mike, pimpin' the world 1 rig at a time
 
I am actually not all that familiar with Vista, does anyone know if it is a native 64-bit os or if there will be 2 versions like there are now?
Click to expand...

There will be two versions. 32 bit and 64 bit. Actually, there will be Itanic support as well. So 3 platform versions total.

I really think that Microsoft should make Vista a 64bit only os.
Click to expand...

Doing so would result in a hefty loss of revenues. Especially in the upgrade versions. You know M$ ain't gonna do that. ;-)

-mpjesse
 
...than again, I run XP Pro x64.

~~Mad Mod Mike, pimpin' the world 1 rig at a time
Click to expand...

How much luck have you had with that? I run it as a second OS, and still find that I run into occaisional stability probrlems from lack of driver support. Then again, I have an nForce3 board, and there are no newer drivers for that.
 
So far, after about a year, I find it to be just what I need. I don't have a printer (paper? pff) and all the drivers I need for my hardware (mobo, vid card) either came on the CD or I got online. I find personally that the OS seems more responsive, and my memory bandwidth (according to Sandra) is 10% greater then the same settings on 32-bit XP Pro (5.2GB/s to 5.8GB/s on x64). I would recommend it as a main OS on nForce 4, due to lack of driver support for previous versions. I did find that on my 3400+ laptop, x64 has drivers for my integrated sound but 32-bit doesn't (gateway's drivers won't work on XP Pro 32-bit for some reason).

The only thing I noticed, was 64-bit ATI drivers, they suck. You have to manually install them because it gives you a thunk.exe error everytime, it's quite annoying, but thankfully my nVidia drivers work fine and my SLI and overall system is very stable, at least in my experience. There is 1 downside to x64 XP Pro, you can't run 16-bit appz, and that totally is a drag 8).

~~Mad Mod Mike, pimpin' the world 1 rig at a time
 
Interesting.

I have found it to be a bit faster in my testing, though I need to have printer drivers and haven't found any yet for either of my printers. Printing to PDF then switching OS's and print in XP32 doesn't cut it.

Hopefully I can upgrade to an NForce4 board soon (If the wife will let me buy a new Graphics card and mobo), then I can really test it out.
 
By him stating "funny how AMD is sitting around praying for DDR2-800" he is implying that there is impending financial doom for AMD. It's quite the opposite actually.

So explain to me how my comments had nothing to do with his statement?

1. I addressed his insipid statement that the switch to AM2 is somehow stupid because it's a change in memory technologies.

2. I addressed his comment about DDR2-800 "saving" AMD. So what does AMD need saving from? Hmmm?

3. Quite the contrary, I'd argue that most financial analysts (esp. in semiconductor) do know the products. Most financial analysts are really research experts. They spend hours studying products, supply, demand, etc. What evidence to you have to support your claim that analysts "rarely have any idea of the products at all?"

You too need to go away. Ur nothing but a troll around here anyways.
Click to expand...

I was gonna say the same thing if you didn't. It is annoying when one person says something that shows there complete ignorance. It is downright ludicrous when someone backs them.
 
I am not saying it should, I am saying it does, because I have tested it personally on multiple CPU's setting the RAM to make sure it stays at around 200MHz (give or take a Megahert), the performance gain may not be huge, but it's there. Not to be insulting, but I think you're a bit ignorant on the CPU/FSB/RAM thing, and I mean all the respect I can in saying that.

~~Mad Mod Mike, pimpin' the world 1 rig at a time
Click to expand...

OK as long as it is respectfully, it is fine with me :)

But ... please show me the numbers then.

On the one hand, you say, it is more or less the same ("give or take a Megahert"), on the other hand you say, that your measured performance increases are not huge.
The conclusion, I draw from that is, that your performance increases were due to a few Mhz more ...

Lets discuss the following configuration:

CPU freq: 2400 Mhz
Memory freq: 200 MHz

1st setup:
HT frequency 240; HT multiplier x4 -> HT960; RAM divider "166", CPU multiplier X10

2nd setup:
HT frequency 200; HT multiplier x5 -> HT1000; RAM divider "200"; CPU multiplier x12

2400 and 240 Mhz are perfect, because it will result in a nice 200,0 Mhz clock for the RAM, by using the "166" mem clock divider. The 40 MHz difference for the HT speed is negligible.

If you cant test that, what setups did you have, and how big were the performance gains ?

If you can proof your claim I will believe you, but as I already said nothing is run by the HT reference clock directly. Thus I do not see where the gains should come from, besides slightly faster CPU or Mem clocks.

byebye

FalseInformation
 
Yea, I can test that, but unfortunately I do not have a 4000+ readily available (a friend does) so I will try to test it with his 4000+ and post my results, if anybody can do it before me and post the results, that'd be helpful. I am not exactly sure what you're questioning, if you could elaborate further as to what you're inquiring about, I could help.

~~Mad Mod Mike, pimpin' the world 1 rig at a time
 
What I found to be ridiculous is their SiSoftware Sandra Memory bandwidth results, they put DDR2 667 at 5.8GB/s, but their past Benchmarks put DDR2 667 in an Intel platform at 6.6GB/s, hmm... 800MB/s loss, that aint due to different RAM, unless this RAM was special RAM...

~~Mad Mod Mike, pimpin' the world 1 rig at a time
Click to expand...

the timings, dude, the timings... I've seen tighter timings on Intel rigs... and I've tried relaxing the timings on my RAM, makes a difference in the bandwidth, and should even more so with DDR2 and dual channel


Timings are overrated man. Yes tighter timings make a difference. Nowhere near that much of a difference. AMD has been outperforming Intel for quite some time now. They have done it being behing a generation in propcess manufacturing. AMD did the smart thing letting INTEL jump to DDR2 and force the development of it. AMD can now jump on the DDR2 bandwagon at the most opportune time, when it is mature and fast and will actually increase performance. When AMD also switches to a 65nm process the performance increase will be noticeable. I believe somewhere in the neighborhood of %10-15.
 
All I been trying to say since the beginning is that those are not credible benchmarks, but everybody is saying "oh they're so true, just read the article" I've read that article once, all it took for my to determine it was BS. Windshear: higher latencies DO NOT affect bandwidth that much, either on an A64 or P4, the most you lose from 2-2-2-5 tight timings to 5-5-5-15 (the worst) is about 200-350MB/s, but you gain so much back going from standard 200MHz to 333MHz which is DDR2 667, there is no way in hell that those benchmarks are true, as stated before, so let this thread die now, it's proven that article was useless.

~~Mad Mod Mike, pimpin' the world 1 rig at a time
Click to expand...


You are %100 Correct
 
Ram Clock Cycles I believe, but the speed of the RAM speed is dependent upon the CPU frequency divided by the divider. As for being dumb, being dumb and not knowing the answer are 2 very diffrent things. Just ask if someone can explain it, don't short change yourself :)
Click to expand...

It is ram clock cycles, hwoever, the CPU speed is dependent upon the RAM SPEED and the clock multiplier.

Also to settle the earlier confusion about DDR memory. A clock cyle is simply one cyle of power. Frequency only exists in AC power. This Cycle is called a sine wave. In your home, in the US your electriciy has a frequency of 60hz, in Europe it is 50. That is to say that the electriciy changes polarity, goes form positive to negative, 60 or 50 times respectively per second. If you were to graph this it looks like a wave, hence the name sine wave. In a computer, Single Data (rate) Random Acess Memory(SDRAM) sends one signal per FULL sine wave a.k.a. clock cycle. Whereas Double Data Rate (DDR) sends one signal per half cyle. It uses the positive and negative sides of the wave seperately instead of together as one. Therefore, DDR running at the exact same speed as SDRAM with transmit 2x the data. DDR and DDR2 are reffered two by there equivalant(sp.) SDRAM speed. So DDR(2) 400 is actually 200mhz but has the same bandwith as SDRAM operating at 400mhz, which of course does not exist.

I hope that cleared up some confusing, if there was any if not, then I feel kinda stupid for telling you all what you already knew.
 
indeed i do not think people realize when they are ranting about how pc3200 performs the same as ddr2 that all that extra bandwidth is not being used it's getting close but its still not there the only way i think a desktop pc could every use ddr2 bandwidth needs would be with a gaming system a bit down the road with a dual graphics card solution a top notch sound card like x-fi
and a physics card but for the average user and even the casual gamer
Click to expand...

WRONG! Everything in the computer is a bottleneck compared to the proccessor, with maybe the exception being the graphics card. The memory is always running at full speed. Any increase in speed will result in an increase in overall systim performance. All the RAM bandwith in the world is not enough to supply the processor. OF COURSE it is all being used. However, with the hardrive being exponentially slower than than both Ram and Proc, the performance increase may not seem like that much. It is there though. It would be impossible for it not to be.
 
hello guys,

Ok i'm not an expert when it comes to memory setting and shizzle like that but i have build over 10 pc's and helped alot of people with their pc's. To my experience i haven't noticed any noticable different when changing to a faster memory althought it should make a few % different.
One thing i'm REALLY impressed over is the performance AMD x2 cpu's have.

I recently build a 4200+ x2 pc and another D 830 Intel pc's.

Got damn what a huge different there was in performance. the AMD box just blow the intel in all category. i had the best motherborad for both syste (Asus for AMD and MSI for intel) and also the fastest memory for each system with a 7800 GTX as graphic card and 2G ram on each system.

The AMD run at 49degree when playing while intel was over 70 degree in the same box with 3 fans at idle the AMD was around 30 while Intel was around 55-60. And the performance was much better with AMD so there is no doubt in my mind that AMD is the leading company.

When it comes to the new socket it make a little different if you get better performance since i think the performance you get today is way way better then any Intel cpu around. Maybe the test results show a different story but if you sit at the pc you WILL notice the AMD power trust me :)

i think the new socket just is the first step to go to the next step which is 65nm cpu's with 2+ cores so dont get hanged over the performance. That is how i see it atleast.

All and all i don't think i need any upgrade with my new pc for atleast another 2 year or maybe more. It's so blazing fast 😀 never been happier with any pc i ever build before. /me proud hehe

cheers
/R
 
You must log in or register to reply here.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom