AMD... you better come up with something quick.

Intel would love to get rid of AMD just as long as they didn't get the same anti-trust DoJ hassles M$ had.

Actually they might not care as M$ won eventhough they lost the case.

DaddyB said:

Anyway... the on die memory controller for the A64 is only single channel, not dual channel like it's big brother so there goes half the bandwidth.

Some motherboards may use their own memory controller that will allow you to use an A64 with a dual channel setup but then you lose the benefits of the on die controller (lower latency etc.).

Click to expand...

Are you talking about the fact that the Athlon64's memory controller will be like the Opteron's in that there is only one big channel, not two small ones like Intel's (Canterwood)?

I don't believe there will be any Northbridge chipset-based motherboards for the K8 - the latency gains by using the on-die controller are just too much to ignore.

big_cochino said:

Well.......I'm editing video and burning DVDs. The conversion of
the video files from the .AVI "capture" format to the mpeg2 files
needed for use in TV style DVD players is *extremely* mHz
dependant. More speed = more projects per week = more $$$.

Click to expand...

Most of the speed in encoding programs is from SSE2 optimizations, not clock speed. It is not a simple matter to change the "GenuineIntel" checks the lazy programers use to check for SSE2 compatibility however, and this is why you'll probably see Intels still beating AMD in this arena for a while. When they start changing those checks, you'll see a huge performance jump with the K8.

Mark Larson said:

Are you talking about the fact that the Athlon64's memory controller will be like the Opteron's in that there is only one big channel, not two small ones like Intel's (Canterwood)?

I don't believe there will be any Northbridge chipset-based motherboards for the K8 - the latency gains by using the on-die controller are just too much to ignore.

Click to expand...

Good point I should have clarified, Opteron has one 144bit wide memory bus which is broken down into 2 64bit memory channels (128bit total) and 16bits for the ECC of the ECC RAM it uses or maybe it was 142 total and 14 for ECC... it's one of the two. A64 will have one 64bit wide bus capable of only single channel RAM and no ECC. So to answer your question no it won't be one big bus like with Opie.

I agree that there shouldn't be a motherboard based memory controller, the gains in latency and overall bandwidth were quite impressive in the Opie reviews but I have heard that there might be some. If they do make them you wil likely be able to chose between using the on die memory controller or the motherboard one in the BIOS.

Northbridge and southbridge as we know it are going to change with the Opteron and A64, the typical northbridge controls the RAM and AGP, while the southbridge controls the PCI, IDE and ISA (in older boards). With the 64bit AMD's the southbridge will control AGP and all the regular southbridge stuff, if there are hammer motherboards with a memory controller onboard I wouldn't be surprised if it was all in one chip or the memory controller was on it's own chip (with AGP or PCI express graphics being controlled by the southbridge).

DaddyB said:

Good point I should have clarified, Opteron has one 144bit wide memory bus which is broken down into 2 64bit memory channels (128bit total) and 16bits for the ECC of the ECC RAM it uses or maybe it was 142 total and 14 for ECC... it's one of the two. A64 will have one 64bit wide bus capable of only single channel RAM and no ECC. So to answer your question no it won't be one big bus like with Opie.

Click to expand...

I can't believe AMD will cripple it like that...

From the AMD Website
The AMD Athlon 64 processor will contain one HyperTransport link offering 6.4 GB/s data transfer while the AMD Opteron processor will offer three links. The processors will also contain different amounts of cache.

Click to expand...

Hmm... still has 6.4GB/s of bandwidth.

Also:

The AMD Athlon™ 64 processor and the AMD Opteron™ processor directly addresses this bottleneck by integrating a DDR memory controller into the processor, revolutionizing the way x86-based processors access main memory. By running at the processor’s core frequency, an integrated memory controller greatly increases bandwidth directly available to the processor at significantly reduced latencies. The performance-enhancing effect is even more dramatic within an AMD Opteron multiprocessing environment, because each additional processor has its own memory controller thereby scaling over all memory bandwidth.

Features of Integrated DDR DRAM Memory Controller:

* Available memory bandwidth scales with the number of processors added
* 128-bit wide integrated DDR DRAM memory controller capable of supporting up to eight (8) registered DDR DIMMs per processor
* Available memory bandwidth up to 5.3GB/s (with PC2700 memory) per processor

Click to expand...

Doesn't say whether this is exclusive to the Opteron. Opteron has a 144-bit wide bus, with 16 bits for ECC.

Benchmarks done by tom compairing intel to amd and putting intel ahead are the sort of meat products that give you gout (and not liver or kidneys). That said, 2.4 wont be consistantly beating a 3200+ across the board. if you believe that, you are a fool. 3000 and 3200mhz chips might come a bit closer, but there will still be many instances where amd will outperform the pentium 4's.

Lithan said:

That said, 2.4 wont be consistantly beating a 3200+ across the board. if you believe that, you are a fool.

Click to expand...

if you would actually READ what i typed you would find i said that the 2.4b beat the 3200+ on SOME tests. i did not say at any time that it beat the 3200+ consistently.

and to be precise i believe it was 2 tests out of the whole comparison.

There is a movie on that I want to watch so I'll make this quick.

If you want to read more about the memory bandwidth of A64 then you should read this preview of the A64 (page 6 is about the memory but the whole article is worth a read. here is a quote from the article:

As we see, the single-channel Athlon 64 controller is much faster than the dual-channel controller of the nForce2 core logic, which cannot show its best because of the limited processor bus bandwidth. At the same time, Athlon 64 had hard time trying to compete with the bandwidth of i875P dual-channel controller.

Click to expand...

I doubt that they will change their mind before A64 is released, after all there is very litle that is different between Opie and A64, mainly half the memory bus on A64, Opie uses ECC RAM, there will be an Opie with more L2 cache and an A64 with less cache and A64 supports ddr400 as opposed to Opie's ddr333. I think that's all the differences but I can't think cause I'm watching the sum of all fears.

Maxvla said:

these 800fsb processors are literally untouched through the whole test.

Click to expand...

There's a 2.4 800fsb. So you either said that, or you used very poor wording and in doing so, insinuated as much.

If either AMD or Intel were to die, it would be a blow to everyone here...

Re: Re: AMD... you better come up with something quick.

Lithan said:

There's a 2.4 800fsb. So you either said that, or you used very poor wording and in doing so, insinuated as much.

Click to expand...

when you compare the same processors in the different lines.

like the 2.4c compared to the 2400+ and so on. not that all of the 800fsb chips are beating every AMD chip every time. must you always think in absolutes?

i probably could have worded it a little better but the point remains that the 800fsb chips are consistently beating chips with a higher PR value.

start said:

If either AMD or Intel were to die, it would be a blow to everyone here...

Click to expand...

how so?

Re: Re: Re: AMD... you better come up with something quick.

Maxvla said:

i probably could have worded it a little better but the point remains that the 800fsb chips are consistently beating chips with a higher PR value.

Click to expand...

Yup, this is the point where amd's pr-scheme crumbles. It just does'nt take into consideration the possibility for such a change in intel platforms performance. (and please don't start telling me pr's are comparing to old athlons)

Re: Re: Re: Re: AMD... you better come up with something quick.

dropadrop said:

Yup, this is the point where amd's pr-scheme crumbles. It just does'nt take into consideration the possibility for such a change in intel platforms performance. (and please don't start telling me pr's are comparing to old athlons)

Click to expand...

The PR-rating is comparing to the Athlon 1Ghz.

Skiing Squirrel said:

how so?

Click to expand...

Because then Intel would be the only manufacturer of x86 CPUs and we would be back to CPUs costing more than $ 1000.

Did you know? The 3.0C was the only CPU ever made by Intel that was less than $ 600 at launch! This is because of competition by AMD, not the tooth fairy.

Re: Re: Re: Re: Re: AMD... you better come up with something quick.

Mark Larson said:

The PR-rating is comparing to the Athlon 1Ghz.

Click to expand...

Athlon 1.4 actually

dropadrop said:

Click to expand...

u said it

Re: Re: Re: Re: Re: Re: AMD... you better come up with something quick.

OC Detective said:

Athlon 1.4 actually

Click to expand...

Whoops! My bad!