AM2 Tested with 800 Mhz DDR2

[cubicleslave]

The chips of 3-4 years ago are more than powerful enough for your average consumer who uses Word, Internet Explorer, Outlook Express, and Solitare. And even those computers can play some of the games out there today, as not everyone pays the 3D intense games.

What about video encoding? I think that would be a big enough job to bring today's systems to their knees. There's still plenty of use for more CPU power, as people take up digital video work. I would love to be able to encode a 2 hour movie in HD format, in less than 2 hours.

Blah. This is my first time posting. Sorry I was trying to save space by clipping Bront's quote down, but ended up messing up the quote. Bront wrote the quote, not Apache. Thanks.

 

[jkflipflop98]

The stepping that was shown at IDF isnt' even close to the top of the line. We've bin'd out parts capable of 3.5GHz. Mind you, at this stage there isnt exactly a ton of conroes in the pipeline, so they do tend to get a little more attention paid to them than the normal P4's right now, but we're changing that rapidly, getting ready for an intense demand. Good thing for us we have enough manufacturing muscle to make enough processors for Earth in 6 months =D

These babies arent fake, there was no "trickery" going on behind the scenes. We've really made something great this time around, and that's good news for everyone.

We chose a midrange part to compete with AMD's future best, just to kind of show off and have some fun. 2.66's should retail at the $450 - $500 mark, whereas the FX-60 would be almost $1200. However it has come to my attention that AMD will be severly slashing prices like mad in september due to the launch of Core - good news for all if you ask me.

 

[TabrisDarkPeace]

I've been lead to understand the same thing draco, AMD => timings, Intel => bandwidth. Intel is only just now fully taking advantage of DDR2, whereas AMD has been using DDR well the whole time. I can't really help but think AMD should have skipped over DDR2 and went straight to DDR3 or GDDR3. The Xbox360 uses GDDR3 as system memory, so the technology is there at least.

Yeah, I want to save US$135 on an AMD processor (on average around October 2006) only to pay an additional US$350 (if not US$3,500) on DDR3 memory and see maybe a +7% increase in performance, if that.

There is just so much DDR3 being manufactured for use on DIMMs right now isn't there ? 😛 (sarcasm)

Assuming they did, then Intel would be in an even better position total machine cost wise (CPU + RAM + MOBO alone they would be aswell).

With a processor cache hit rate (on each core) of at least 80%, 6.4 GB/sec to/from memory is fine, going higher won't help performance much in most CPUs, with a 95% hit rate 1.6 GB/sec would suffice fine, you'd see maybe a 3% drop in performance (vs 6.4 GB/sec at 95%). If the hit rate was under 70% then yes, DDR2 would help significantly..... but if it was that low in normal software then people wouldn't have purchased the processors in the first place because of their low performance. 😛

Going to DDR2 is more a move to let higher clocked processors based on K8 architecture scale better (per clock cycle) and provide enough memory throughput for quad-core on a chip.

PS: We won't be using (GDDR3) Graphics DDR 3 memory as system memory because of the costs.

Heck, the overheads for going from DDR2-667 on Intel 'Core' (Conroe) to DDR2-800 on AMD Athlon 64 / Opterons will be enough to affect total cost of ownership and total price / total* performance ratios to the point where Intel actually look even more attractive.

Do people really think DDR2-800 is going to cost the same as DDR2-667 in October 2006 ?

It is quite likely anyone looking to save the US$135 on an AMD system (predicted average for Oct '06) is not going to want to spend more than the difference just to get DDR2-800, and still wind up with lower overall performance (vs what they could be getting for the same or less dollars).

Think about it.

My personal estimates put Intel 'Core' (Conroe - 2.66 GHz) at US$530, and future AMD Athlon 64 at US$400 (K8L / 8KL [?] - 3.6 GHz apx), which will still let AMD compete price / performance wise but Intel are taking that market share AMD just gained back (as they Intel to do every few years, just to prove they were not really trying to compete before at all).

If AMD make a quad-core AMD64 on a chip (using 65nm) they'll cost a fair bit for at least the first 9 months, likely be Socket F only (as least initially aswell), but will perform damn well for the price in video encoding and server oriented tasks.

Intel want the enthusiest gamer market, as they are generally the nerds / geeks people go to for computer advise, and they'll start recommending Intel systems (in more cases). Dell also want the enthusiest gamer market but even with the newer Intel 'Conroe' CPUs I doubt Dell will be able to sell to them. 😛

Quadcore on a chip isn't going to help gaming significantly anytime in the next 3-5 years, and by then much newer hardware will be out.

 

[Stimpy]

I don't think AMD ever claimed that the switch to AM2 would give them a massive improvement in performance (although it will help scalability for opteron with socket F).
It is mearly a technology enabler for future optimizations and die shrink.
A good parallel would be the car industry. I.e. VW develop a new chassis, but stick in the old engine range. Initially disapointing as the car goes just as slow with the same economy as the last car range, but may handle a little better.
However with this chassis they have the strength to stick in their latest W12 engine, once the engine design is finalised.
There are a number of things AMD can still do withing the current processor range to speed it up, e.g. better branch prediction, L3 cache, more efficient SSE instructions (inclusion of SSE4), etc, etc.
With the introduction of 64bit Vista they may actually finally be able to show their trump card, i.e. a very effcient 64bit implementation (we know nothing of conroes 64bit capabilities yet).

 

[jkflipflop98]

Since you're one of the guys from Intel.
Can you tell me if Intel is going to fire or dis at the engineers who made the Prescott and Willamette?

Cuz I really want the blood of the engineers who made those two procesoors and made the technology/speed grew so slow. In another word, I want to rip their neck off and hope they never see this planet again. They are the scum bags of human race.

You are an angry, angry little kid with no direction.

 

[BaronMatrix]

I can say I expected process improvments and tweaks to account for 10%, while mobos should give it another 5-10% which does hit the 20% increase. WHo knows until June but I am confident in AMD.

 

[qurious69ss]

Since you're one of the guys from Intel.
Can you tell me if Intel is going to fire or dis at the engineers who made the Prescott and Willamette?

Cuz I really want the blood of the engineers who made those two procesoors and made the technology/speed grew so slow. In another word, I want to rip their neck off and hope they never see this planet again. They are the scum bags of human race.

You are an angry, angry little kid with no direction.

What does WWID mean?
Can you tell we what the top headline is for the coprorate news on circuit????

I smell a poser.....

 

[pip_seeker]

What about video encoding? I think that would be a big enough job to bring today's systems to their knees. There's still plenty of use for more CPU power, as people take up digital video work. I would love to be able to encode a 2 hour movie in HD format, in less than 2 hours.

Your average Joe isn't going to be interested in this, the market for encoding Hi Def anything is an upper end market. There are products that can do this now in various ways, but without a standard HD delivery system it's really to soon for any of this. HD-DVD and Blu-Ray are both fighting for this market and you should remember what happens when you have two or more competing formats. VHS + Beta anyone?

So it's really still too early to worry about HD conversion, it's very hi-end and a specialized market. If money is no concern, there are several products that can do the job now, but why do it if you have no medium to put it on?

To do this in RT or better than RT ha... you might need to take out a loan to do that. A cpu won't crack that nut on it's own, you'll need some dedicated hardware as well as some serious disk space to knock that out, but with that said there are all different kinds of products that do this in various ways... you can even do HD in software only, but no where close to RT... yet and I seriously doubt you could do a 2hr movie in under 2 hours even with hardware / software acceleration.

Plus I should mention that you didn't even say what flavor of HD? HDV? 1080i?, 720p? Most upper end products can do all, but I'm sure as time goes on HDV will be the consumer format of choice more than likely.

 

[apache_lives]

I can say I expected process improvments and tweaks to account for 10%, while mobos should give it another 5-10% which does hit the 20% increase. WHo knows until June but I am confident in AMD.

AMD can gain no where near enough to close the gap of conroe, and keep in mind CONROE WAS CLOCKED AT A LOWER SPEED ASWELL.

There was upto 40% more performacne at a lower clock speed and a minimum of like 10% - Intel for once did there homework!

 

[gOJDO]

Going to DDR2 is more a move to let higher clocked processors based on K8 architecture scale better (per clock cycle) and provide enough memory throughput for quad-core on a chip.

Heck, the overheads for going from DDR2-667 on Intel 'Core' (Conroe) to DDR2-800 on AMD Athlon 64 / Opterons will be enough to affect total cost of ownership and total price / total* performance ratios to the point where Intel actually look even more attractive.

Do people really think DDR2-800 is going to cost the same as DDR2-667 in October 2006 ?

It is quite likely anyone looking to save the US$135 on an AMD system (predicted average for Oct '06) is not going to want to spend more than the difference just to get DDR2-800, and still wind up with lower overall performance (vs what they could be getting for the same or less dollars).

Think about it.

I totaly agree with everything that TabrisDarkPeace wrote in his post.
Read this article where guys from AMD are thinking to change thir focus on the market considering that they lost the perofrmance battle of their Athlon64 FX/X2 and Conroe/EE.
http://www.xtremesystems.org/forums/showthread.php?t=91629

...back to topic,
I was not expecting any performance boost from greater RAM bandwidth on K8. We already saw that when they switched from s754 on s939. With the double of memory bandwidth the s939 gained about 5-7% of performance boost. OK, lets say there was a memory bandwidth bottle-neck and they fixed that.
But, why going further with increasing that bandwidth when the greater bottle-neck is already solved with the transition to s939?
How K8 AM2 is improving the memory bandwidth?
With replacing the integrated DDR memory controler with the DDR2 one. But it is has unoptimized architecture to use the benefit of the high bandwidth of this type of high latency RAM. K8 architecture still suffers from the high memory latencies unlike the Core which has an implemented memory reorder buffer and technology for smart memory access.

 

[endyen]

Read this article where guys from AMD are thinking to change thir focus on the market considering that they lost the perofrmance battle of their Athlon64 FX/X2 and Conroe/EE.

Ha ha. Take a quote as out of context as you like, what the man said was he is going after more of Intel's market share. You know what Intel fanboy? He will get it, and he will do it in the big$ server market.
As to bandwidth reqiurement, you do understand that a dual core has more memory requirements than a single core chip, dont you?
You obviously believe Intel's PR department lies. I hope September doesn't bring too large a disapointment.

 

[gOJDO]

Read this article where guys from AMD are thinking to change thir focus on the market considering that they lost the perofrmance battle of their Athlon64 FX/X2 and Conroe/EE.

Ha ha. Take a quote as out of context as you like, what the man said was he is going after more of Intel's market share. You know what Intel fanboy? He will get it, and he will do it in the big$ server market.
As to bandwidth reqiurement, you do understand that a dual core has more memory requirements than a single core chip, dont you?
You obviously believe Intel's PR department lies. I hope September doesn't bring too large a disapointment.
(http://www.tgdaily.com/2006/03/06/amd_ceo_capacity_could_be_traded/)
He(AMD chairman and CEO Hector Ruiz) said that they are going to stay away from the performance desktop market for a while(well that could be years) becouse unknown to us reasons.
AMD are fighting to conquer that desktop market since they produce chips, so tell me why should they leave the battlefield so cool and so defeated?
So what do you think is the reason of Ruizs statement?

"We don't feel we need to make huge strides in the consumer [space] in the near future. In terms of retail around the world, for example, we have 40-50% of the market. Our desire, frankly, is to just hang on to that while we make much more significant impact on the commercial side."

For me, this means there will be almost no improvements(counting the transition on the high-latency DDR2 and its minor gain od 3-5% in performance as improvement) on the current Athlon64/X2/FX in the folowing next years. On the other side its rival Intel is lowering the prices and they offer dualcore processors for less than 150 euros(PentiumD 805) and will sell Conroe 2.66 for the half of the todays price of FX-60. As today articles and benchmarks are saying(i guess what we read is true), Conroe drasticly overperforms FX-60 in all segments as we are informed today. So even the hardcore AMD (as you named them) fanboys will use their brain and will buy what is better and for less money.
BTW how did you found out I am a kind of fanboy of somthing?!?
I have 6 computer systems, 4 of them are K8, 1 is P4 and 1 is P3...Damn FFS, Intel fanboys should buy only Intel!

 

[endyen]

He said that they are going to stay away from the performance desktop market for a while(well that could be years) becouse unknown to us reasons. So what is the reason?

What he said was that he was going to have Amd focus on the server segment. Dont read into it what isn't there.
What K8s do you have? What mobos?, ram setups?, gfx cards? what bios ver? and drivers?

 

[gOJDO]

What K8s do you have? What mobos?, ram setups?, gfx cards? what bios ver? and drivers?

Why do you need this info?

I am writing this post on this machine:
1. Asus A8N5X s939 nForce4 PCIe HTT1000(5x200 @4x250)
Athlon64 3200+ (Venice Rev E6) 2.0GHz @2.6GHz(260x10)
Inno3D GeForce 6600 256 MB DDR2 350/700 @450/850
2x512MB DDR400 Mosel Vitelic CL2.5 4-4-8 CR2 2.6v @433 CL2.5 3-3-6 CR1 2.85v
AWARD BIOS ver. 502
nVidia nForce4 61.70, GeForce 81.98

2. Soltek K8T800 Pro s939 VIA KT800 Pro AGPx8
Athlon64 3000+ (Winchester) 1.8GHz
Inno3D GeForce 5700 256MB DDR 325/550
2x512 Kingston CL3 3-3-7 CR2
don't know BIOS and driver version, i don't have that PC at hand now.

3. HP Compaq nx6125(PY417EA#ABB, s/n: CND5512LTM)
HP 308B KBC Version 45.23, ATI Radeon X200M X-BUS PCIe
Turion64 ML-34 1.8GHz 1MB L2
2x512MB SO-DIM DDR333 Samsung CL2.5-3-3-7 CR1
on board on chipset ATI Radeon X300 with shared up to 128MB of onboard DDR
HP BIOS 68DTTVer. F.07
ATI Catalyst (i can't see version, probably 6.1: 6-1-igp_xp-2k-dd_CC_wdm_sb_gart_enu_29602.exe)

4. ASUS K8N nForce3 250 s754 AGP
Athlon64 2800+ (Clawhammer) 1.8GHz 512kB L2
512MB PQI DDR400 CL2.5 4-4-8 CR2
MSI GeForce4 MX440 8xAGP 64MB DDR
I don't have it at home this PC also, I think it have nVidia nForce3 5.24 for the mobo and GeForce 81.98 for the VGA.
BIOS is AWARD, I think 1009-003 was the last ver. when I updated it.

 

[cubicleslave]

What about video encoding? I think that would be a big enough job to bring today's systems to their knees. There's still plenty of use for more CPU power, as people take up digital video work. I would love to be able to encode a 2 hour movie in HD format, in less than 2 hours.

Your average Joe isn't going to be interested in this, the market for encoding Hi Def anything is an upper end market. There are products that can do this now in various ways, but without a standard HD delivery system it's really to soon for any of this. HD-DVD and Blu-Ray are both fighting for this market and you should remember what happens when you have two or more competing formats. VHS + Beta anyone?

So it's really still too early to worry about HD conversion, it's very hi-end and a specialized market. If money is no concern, there are several products that can do the job now, but why do it if you have no medium to put it on?

To do this in RT or better than RT ha... you might need to take out a loan to do that. A cpu won't crack that nut on it's own, you'll need some dedicated hardware as well as some serious disk space to knock that out, but with that said there are all different kinds of products that do this in various ways... you can even do HD in software only, but no where close to RT... yet and I seriously doubt you could do a 2hr movie in under 2 hours even with hardware / software acceleration.

Plus I should mention that you didn't even say what flavor of HD? HDV? 1080i?, 720p? Most upper end products can do all, but I'm sure as time goes on HDV will be the consumer format of choice more than likely.

OK, maybe it's a little early for making HD format DVD's, but I'm sure progress will eventually be made so that we have a medium to put it on. But even today, that's not necessarily a roadblock. You don't need any next-generation DVD disc to make an HD format Tivo device, since it stores the video on the hard drive. I am sure there would be demand for such a device today, since every time I walk into a video store, all I see are HD televisions. I think before such a device (HD Tivo) comes into existence, all the ingredients have to be available, first to the computer enthusiast who would be an early adopter. One of those ingredients would have to be a cpu with much more processing speed than available today. I am not up on all the different HD formats, but I got a feeling that even at the lowest resolution format, today's hardware would not be up to the job of RT encoding. I think your mention of "dedicated hardware" encompasses both capture and encoding. I would say that the capture is the easiest part to solve. I am able to squirt video from my camcorder onto my hard drive in real time through the firewire port. But then taking that video and encoding it into mpeg2 for DVD takes bloody ages, at least 3-4 times longer than RT. So the encoding is currently the bottleneck.

 

[jkflipflop98]

Since you're one of the guys from Intel.
Can you tell me if Intel is going to fire or dis at the engineers who made the Prescott and Willamette?

Cuz I really want the blood of the engineers who made those two procesoors and made the technology/speed grew so slow. In another word, I want to rip their neck off and hope they never see this planet again. They are the scum bags of human race.

You are an angry, angry little kid with no direction.

What does WWID mean?
Can you tell we what the top headline is for the coprorate news on circuit????

I smell a poser.....

WWID is world wide id number(cuz we're all robots), and the current circuit headline at Ronler is something about writing blogs. . . i dont really pay much attention besides "My paystub" and "my SPP account" (btw, i have over 2500 shares. . . how many do YOU have?)

I am not a poser.

 

[joset]

Can you tell me if Intel is going to fire or dis at the engineers who made the Prescott and Willamette?

"NetBurst" was a magnificent achievement, in terms of raw processing power. And, whatever will come out, in whichever microarchitecture, GHz & THz will be there, at the end of the day.

The thermal barriers are the same for every brand, chip, microarchitecture; in a sentence, thermal issues are pervasive throughout electronic devices. Nevertheless, IBM's coming out with a >4.6GHz POWER6 and, together with Cell, GHz will - most probably - surpass the 5 mark. NetBurst-like processing approaches will pop-up in every HPC platform, in one way or the other (given thermal issues are overcome, as is the case with POWER6).

Deep, narrow pipelines are only "bad" when thermal, current leakage and power consumption issues are still unresolved matters. Current CPU generations (AMD & Intel, for what matters), are shallower & wider than "NetBurst" was; however, if you'll notice, pipeline stages are still increasing, compared with previous, non-NetBurst approaches.

Intel's "NetBurst" might be dead, but Netburst-like processing is still to come.

(Pardon my assertive tone, without presenting any backup references; but you'll surely notice the conspicuous coincidences, as time passes...)

Cheers!

 

[joset]

AMD certaintly has something planed, but it may not match what Conroe has to bring as a completely new chip. However, we as consumers benifit from all this, as it will push prices down, and ultimately performance even higher as AMD and Intel stuggle to stay on top.

AMD's scaling up to DDR2 it's an inevitability. Certainly AMD has its own R&D at full speed (very well backed up by IBM...) and lately, has made a few [public!] statements in order to show it's not standing still: aside the unavoidable improvement of the IMC and the adoption of DDR2 (and DDR3 is not very far...), the leverage of the now-ubiquitous high-performance interconnect, HyperTranspot, the (soon to be...) transition to the 65nm node, Z-RAM, better & bigger caches and so on (most of it, it's not yet disclosed, of course), one important step might already have been taken towards a [remotely] Cell-like approach: the [probable] partnership with U.K.'s ClearSpeed in order to [eventually] integrate a 96-core chip (CSX600) into its Opteron lines, first (curious, is the fact that Intel has had previous contacts with ClearSpeed as well as with RIKEN - a Japan-based IP company - which also developed the MDGrape-3...).


http://www.reghardware.co.uk/2006/03/15/amd_clearspeed_opteron_maths_co-pro/


Although some of these moves are expected in the near-to-mid-term timespan and, even if some imply a severe revision of the current microarchitecture, AMD might surprise everyone with what it has yet to reveal, in the short-term.


Cheers!

 

[gOJDO]

Although some of these moves are expected in the near-to-mid-term timespan and, even if some imply a severe revision of the current microarchitecture, AMD might surprise everyone with what it has yet to reveal, in the short-term.

I wish that to become true. It is good when there are competitors, it is nice when the prices are going down. But until than we can only guess, wait, read and discuss. I want a good chip in my case now.

 

[phial]

I think I've heard that AMD benefits more from better timings than bandwith(raw mhz speed). Current DDR2 memory still has pretty bad latencies relative to ddr, which is why we don't see much of a difference. On the other hand, Intel would gain more from bandwith upgrade which I think is why the Conroes showed did so well with current ram. The competition will become quite fierce when DDR2 with good latencies becomes available.

I'm not sure about what I've stated but it's what my memory(brain, not ram) tells me. Feel free to correct me.

Yea thats pretty much what I understood as well. Which is why I didnt understand why AMD would start supporting DDR2 in the first place...

 

[MrsD]

Why do you guys continually blabber about benchmarks from something that wont even be out for 3-4 months?????
Secondly I heard that AMD isnt even using DDR2 they are jumping right to DDR3.