Conroe at 4GHz

[POGE]

It shouldnt be compared to K8... :-\ Thats like comparing a presscot to an athlon xp.

 

[wazzledoozle]

Still $210 for the cheapest single core Conroe is a bit steep, especially because by its release an Athlon 64 should be <$100 as a 3000+ is $109 as of now.

 

[XooM]

but when that lowest end conroe can beat the highest and AMD single-core (same with dualcores)... its perfectly reasonable.

however, conroe probably won't scale too well with higher clocks, since its already devouring bandwidth at normal speeds. superpi 1M isn't effected, since it fits in cache.

 

[wazzledoozle]

but when that lowest end conroe can beat the highest and AMD single-core (same with dualcores)... its perfectly reasonable.

however, conroe probably won't scale too well with higher clocks, since its already devouring bandwidth at normal speeds. superpi 1M isn't effected, since it fits in cache.

Ill wait for the retail Conroe's to come out before deciding how powerful they are, anyone remember how fast people said Prescott was before it came out?

 

[POGE]

Ill wait for the retail Conroe's to come out before deciding how powerful they are, anyone remember how fast people said Prescott was before it came out?

Yeah, and for the same reason... its was good at superpi.

 

[BigD6997]

its was good at superpi.

yeah id like to see different tests than just superpi, more real world tests, like to see if it does anything to improve frames in games. that and some other cpu benchmarks as well as pcmark and 3dmark

 

[Jimmy 2004]

I don't want intel to become the best again because it means higher prices... come on amd!!

 

[v-zero]

It shouldnt be compared to K8... :-\ Thats like comparing a presscot to an athlon xp.

That's not exactly accurate. K8 is actually just as advanced as Conroe in most areas, and like was said it carries an onboard memory controller - something Intel will do soon too...
The basis for the design of Conroe comes from the Pentium 3 in the most part. Obviously they have made some serious modifications, but if you dig deep enough you see the architechture is still there .
I think the 65nm AMD chips should see 400mhz+ over the Conroe when overclocking, because of Conroe's pipeline length it will struggle to reach the uber-high clocks without exhaustive cooling and voltage measures...

 

[NamesDontMatter]

Still $210 for the cheapest single core Conroe is a bit steep, especially because by its release an Athlon 64 should be <$100 as a 3000+ is $109 as of now.

????? Athlon 64s are all single core, your comparing apples to oranges. All Conroe processors are dual core. It does not even compare to a 3000+, and if your in the processor market and you want to call $210 steep, then you are crazy. Especially for the performance you will be getting. Not to mention AMD dual cores will have to drop in price if AMD expects them to sell against Conroe since the cheapest AMD dual core is the 3800+ and that still has half the cache of the $210 Conroe, and is probably much slower. Conroe is not meant to compete with Athlon 64, because it just destroys it with speed.

Since if seen alot of comparisons using cars on here lately thats like comparing a stock civic to a 05 mustang, just because the civic is cheaper dosen't mean people won't buy the mustang because it's much faster and a little more expensive .

Also this price drop is going to be tough for AMD to cope with because they have not been able to move to full 65 nm production as intel seems to be doing quite quickly. More Chips per Wafer = cheaper for them = more money made.

 

[Jimmy 2004]

$210 is quite steep for a starting processor and I thought that Conroe did have single core processors or is that just me? If they are only dual core then you're right because the cheapest AMD dual core on NewEgg is $297.00.

 

[Alec§taar]

I wonder what a rig like that would do vs. my latest (don't we all?)

Give me a Conroe, FX60, and a week off from school and Ill run more test than all these people have combined

Ha, good reply!

(That actually would be FUN to do imo as well! Imagine: There ARE folks that get income from doing tests such as the one you mention - what a world! BUT, they do keep us informed, so it's cool by me...)



Anyhow - I guess this demo goes to show you, don't count Intel out of any race, I guess... I would like to see what such a system would do vs. my latest one on a benchmark setup!

Anyways - When you guys go for these kinds of rigs in the future, & doubtless a few of you will?

I would like to see what my current rig does vs. them (these CONROE cpu based Intel rigs, & also future AM2 (?) AMD offerings as well).

If I'm around these forums @ that point? It'd be some fun to see how my latest design stacks up against the next-gen stuff coming!

* I'm going to make it a point to test here when those days arrive & some of you all start getting systems based on CONROE cpu's and even future AMD ones (AM2 chipset iirc is what's coming from AMD, using faster memories & such, correct?)...

BigD6997 also has a STRONG point - one single specialized benchmark result "doth not a total system make" & all that... would be nice to see more tests w/ diff. benchmarks.

APK

P.S.=> See, I do wonder what that machine based on an Intel "CONROE" cpu would do in a benchmark "drag-race/shootout" vs. the one I have now:

1.) AMD Athlon 64 x2 4800+ (o/c 5% by BIOS and ASUS AI Booster)

2.) NVidia GeForce 7900 GTX OC by BFG (o/c manually to 686 core/865 memory)

3.) Dual WD "Raptor X" 150gb 16mb cache each 10k rpm in RAID 0

4.) RAID 0 via Promise EX8350 128mb ECC caching controller + Intel SuperI/O cpu onboard to offload System CPU

5.) A CENATEK "RocketDrive" solid-state 2gb ramdisk in hardware (for paging, logging, webpage caching, & temp/tmp ops duty + print spooler location & command interpreter environment variable location).

6.) Windows Server 2003 FULLY patched & tweaked/tuned 'to-the-max' (plus, running SuperCache-II on my primary C drive, a better diskcache than oem MS one).

This "CONROE" release looks to be the "ace up their sleeve" @ Intel!

My guess is that Intel's probably looking to see when/if AMD "hits-the-wall" on the top performance they can achieve, & to just stay 25% or so faster on most tests, & reclaim the leader-spot in market share once more from AMD... apk

 

[Sandman]

I think Conroe is going to be pretty sweet, but I don't think it's going to blow AMD out of the water like people are saying. Remember when intel compared their conroe to a "projected FX-62"? Conroe beat it by a fair amount, but independant tests showed the overclocked FX-60 performing mysteriously worse than a stock model.

Not only was it questionably configured, it was doing 14x200, not 7x400(as the 62 will). There is a huge difference, and many people didn't seem to be paying any attention. Conroe is going to be good, but it will be competing with AMD systems running on a 400mhz fsb, not 200. If the engineers at AMD could make a DDR1 memory controller that gave DDR2 systems a run for their money, they can probably make a kickass DDR2 controller too. I don't see this being much different than the launch of previous lines of cpu's.

 

[XooM]

to everyone who is saying that all there is are superpi tests, i highly reccomend opening your eyes and looking around. so many WRs have been set in the past few days in every version of 3dmark and aquamark as it is.
3d01: http://service.futuremark.com/compare?2k1=8976572
3d06: http://service.futuremark.com/compare?3dm06=302162
3d05: http://service.futuremark.com/compare?3dm05=2022914
from this thread
and i don't think these are even WR runs; they're just the first ones i found. I also recall seeing an AM3 shot in the 190K range.
EDIT: the 3d01 link doesn't work, and i dunno why...

 

[v-zero]

Conroe is going to be good, but it will be competing with AMD systems running on a 400mhz fsb, not 200. If the engineers at AMD could make a DDR1 memory controller that gave DDR2 systems a run for their money, they can probably make a kickass DDR2 controller too. I don't see this being much different than the launch of previous lines of cpu's.

Ok, serious problems with this post:

1. There is no such thing as an FSB when we talk about the K8. It is an HT bus speed, and the two are not equal...

2. A higher HT bus speed currently offers no benefits over a lower bus speed.

3. DDR1 is currently a far more suitable platform (performance wise) for the K8 than DDR2 is. The K8 relies on very low latency memory to give it a performance increase from its onboard memory controller, and any added latencies (like we see with DDR2) slow the system down horribly.

4. The memory controller on the AM2 K8s is not a new controller at all - hence the new core is a revison, like Winchester or Toledo - specifically this is revision F. Specific areas of the design have been altered to allow the use of DDR2. This is a tactical move, and a move looking toward the future (1+ years) when the DDR2 memory and increased HT will finally become useful. I will explain this below.

Ok. Whilst higher HT bus speeds and DDR2 currently offer a performance disadvantage for dual and single core processors, they will offer a performance advantage to the Quad-Core processors that AMD will release. The K8 has never suffered issues from memory or HT bus bandwidth with DDR1, and this is why the slower DDR2 will offer a disadvantage... However DDR1 would limit the Quad-Cores, and so this move to DDR2 is in preperation for that move to four cores...

It is also worth noting that AMD can cut power consumption on motherboards, and on the processor itself through DDR2 - so whilst performance will not be gained (clock for clock) at all with AM2, AMD still have economic reasons to move to the platform...

If you want an AMD processor that competes with Conroe you will have to wait for K8L... I'll just hold onto my Opteron setup and wait for the winter...

 

[Alec§taar]

Ok, serious problems with this post:

1. There is no such thing as an FSB when we talk about the K8. It is an HT bus speed, and the two are not equal...

2. A higher HT bus speed currently offers no benefits over a lower bus speed.

3. DDR1 is currently a far more suitable platform (performance wise) for the K8 than DDR2 is. The K8 relies on very low latency memory to give it a performance increase from its onboard memory controller, and any added latencies (like we see with DDR2) slow the system down horribly.

4. The memory controller on the AM2 K8s is not a new controller at all - hence the new core is a revison, like Winchester or Toledo - specifically this is revision F. Specific areas of the design have been altered to allow the use of DDR2. This is a tactical move, and a move looking toward the future (1+ years) when the DDR2 memory and increased HT will finally become useful. I will explain this below.

Ok. Whilst higher HT bus speeds and DDR2 currently offer a performance disadvantage for dual and single core processors, they will offer a performance advantage to the Quad-Core processors that AMD will release. The K8 has never suffered issues from memory or HT bus bandwidth with DDR1, and this is why the slower DDR2 will offer a disadvantage... However DDR1 would limit the Quad-Cores, and so this move to DDR2 is in preperation for that move to four cores...

It is also worth noting that AMD can cut power consumption on motherboards, and on the processor itself through DDR2 - so whilst performance will not be gained (clock for clock) at all with AM2, AMD still have economic reasons to move to the platform...

If you want an AMD processor that competes with Conroe you will have to wait for K8L... I'll just hold onto my Opteron setup and wait for the winter...

Excellent feedback for me @ least!

It's been many years on Intel is why (not that this is bad, it's not for various things vs. AMD)... & only recently having returned to AMD for my latest system & the last AMD system I had before it was an AMD K6-III @450mhz circa 1999 iirc.



* Thanks much, good read!

Especially the bolded part above, which I will requote here (I was not aware of it):

"Whilst higher HT bus speeds and DDR2 currently offer a performance disadvantage for dual and single core processors, they will offer a performance advantage to the Quad-Core processors that AMD will release. The K8 has never suffered issues from memory or HT bus bandwidth with DDR1, and this is why the slower DDR2 will offer a disadvantage... However DDR1 would limit the Quad-Cores, and so this move to DDR2 is in preperation for that move to four cores..."

APK

P.S.=> Your feedback here in the URL thread below, v-zero, might be useful as well imo:

http://forums.techpowerup.com/showthread.php?p=96833#post96833

TIA (it's about the AM2 platform & what folks say about it, vs. current AMD Athlon 64 x2 setups & upcoming Intel CONROE cpu's etc.)... apk

 

[Sandman]

Ok, serious problems with this post:

1. There is no such thing as an FSB when we talk about the K8. It is an HT bus speed, and the two are not equal...

Well, first of all, FSB is nearly synonymous with HTT anymore. Second of all, “HT bus” refers to Hyper Transport, which has nothing to do with our conversation. I was talking about the HTT.

2. A higher HT bus speed currently offers no benefits over a lower bus speed.

Well, assuming you are talking about hyper transport, you're right, but since AMD isn't changing that at all with AM2, it's hardly relevant. If you are actually referring to the HTT, which you just "corrected" me on, then you are still wrong, because anyone who has ever overclocked a K8 processor knows a higher HTT, even at the same speed, means better performance. 8x300 is much better than 12x200. This is because, assuming you are not running your memory on a divider, you achieve greater memory bandwidth. Since AM2 will support DDR2-800, this will provide approximately twice the bandwidth. How well it will be utilized is another question, but the potential is there.

3. DDR1 is currently a far more suitable platform (performance wise) for the K8 than DDR2 is. The K8 relies on very low latency memory to give it a performance increase from its onboard memory controller, and any added latencies (like we see with DDR2) slow the system down horribly.

DDR1 was a more suitable solution for K8? Of course it was, but we aren't really talking about good ol K8 anymore, are we? Perhaps you've forgotten, but K8L does not use the same memory controller. At any rate, one of the advantages of DDR2 is that the bandwidth is great enough to compensate for the latencies. Alot of people are complaining about how high the latencies will be, but quite honestly, I don't think they're too bad. When you compare the latencies of some DDR2-800 to some DDR-400, they look pretty bad, but imagine what kind of timings that DDR-400 would have if it could even approach 400MHz. Suddenly 4-4-4-12 doesn't look so bad. those sticks of DDR-400 may do 2.5-3-3-8 at 200MHz, but proportionally, the DDR2-800 isn't doing too bad either.

4. The memory controller on the AM2 K8s is not a new controller at all - hence the new core is a revison, like Winchester or Toledo - specifically this is revision F. Specific areas of the design have been altered to allow the use of DDR2. This is a tactical move, and a move looking toward the future (1+ years) when the DDR2 memory and increased HT will finally become useful. I will explain this below.

What's your point? The improved memory controller is meant to handle DDR2, and I think it will do it quite nicely. The previous verion may have worked well withg low latencies, but this one is adressing a new type of memory, and I doubt it shares the same characteristics as the one meant for DDR1.

Ok. Whilst higher HT bus speeds and DDR2 currently offer a performance disadvantage for dual and single core processors, they will offer a performance advantage to the Quad-Core processors that AMD will release. The K8 has never suffered issues from memory or HT bus bandwidth with DDR1, and this is why the slower DDR2 will offer a disadvantage... However DDR1 would limit the Quad-Cores, and so this move to DDR2 is in preperation for that move to four cores...

It is also worth noting that AMD can cut power consumption on motherboards, and on the processor itself through DDR2 - so whilst performance will not b1e gained (clock for clock) at all with AM2, AMD still have economic reasons to move to the platform...

If you want an AMD processor that competes with Conroe you will have to wait for K8L... I'll just hold onto my Opteron setup and wait for the winter...

If DDR2 will only offer an advantage for quad core chips, why would AMD move their entire desktop line to AM2 when the only quad cores will be socket F opterons? This part in particular makes no sense. You are telling me that a memory solution with superior bandwidth will offer inferior performance in all but one scenario. Call me crazy, but if that were the case, I think they would have realized it long ago and moved only their quad core chips to DDR2. Decreased power consumption is not reason enough for such a drastic move.

 

[Alec§taar]

Ok. Whilst higher HT bus speeds and DDR2 currently offer a performance disadvantage for dual and single core processors, they will offer a performance advantage to the Quad-Core processors that AMD will release. The K8 has never suffered issues from memory or HT bus bandwidth with DDR1, and this is why the slower DDR2 will offer a disadvantage... However DDR1 would limit the Quad-Cores, and so this move to DDR2 is in preperation for that move to four cores...

It is also worth noting that AMD can cut power consumption on motherboards, and on the processor itself through DDR2 - so whilst performance will not b1e gained (clock for clock) at all with AM2, AMD still have economic reasons to move to the platform...

If you want an AMD processor that competes with Conroe you will have to wait for K8L... I'll just hold onto my Opteron setup and wait for the winter...

vs.

If DDR2 will only offer an advantage for quad core chips, why would AMD move their entire desktop line to AM2 when the only quad cores will be socket F opterons? This part in particular makes no sense. You are telling me that a memory solution with superior bandwidth will offer inferior performance in all but one scenario. Call me crazy, but if that were the case, I think they would have realized it long ago and moved only their quad core chips to DDR2. Decreased power consumption is not reason enough for such a drastic move.

This is 1 point I would like to see you guys discuss here...

No, I am NOT just "pouring grease on the fire" either so these 2 "get into it" for no reason... but, I would like to see this point hashed-out, fully!



* I can stand to gain here by watching you "motor-heads" (analog in auto world to hardware nuts in PC world imo) debate this point & the others you made!

Mainly, it seems to come down to what I figured it might from other readings I have done here & elsewhere - memory speeds & types!

One of the MOST/MORE complex ones nowadays imo, & I come outta the 1990's time when things were MUCH simpler on all levels, not just memory!

APK

P.S.=> The reason for doing ALL of their systems that way @ AMD? Probably assembly line retooling, purely economic reasons @ the chip fab level...

"Do them ALL, in prep for Quad-Core futures, OR, just to have systems ready for "high-memory-bandwidth" intensive apps"

I dunno, you guys tell me!

This should be a discussion I can gain from @ least: Why? Because, imo, I am just "weak" in hardware CURRENT/State-of-the-Art knowledge!

Heck, it's not my "forte" in this field anymore, nor do I regularly concentrate in it (but, it is VERY interesting nevertheless)

?? apk

 

[Alec§taar]

I don't think they would do something like that across teh board unless itoffered a serious performance boost.

I don't doubt it, but I don't have enough "grounding/expertise" in & about AMD stuff to warrant debate with you on it...

I can only read what you & others here state & gain by it, making my own decisions once I read what you & the other fellow v-zero (possibly others also) have to say!

AM2 and socket F are completely different, so it's not something that they did just to be uniform or for the ease of it.

Right, it's done afaik? To conform to the NEW DDR2 ram from what I've read here & elsewhere...

Question is - Does that DDR2 memory design really give you a gain across the boards, or only say, in gaming (or, other forms of high memory intensive bandwidth apps, of which I have NO idea if gaming qualifies actually, but I would think they do)?

I think they did it because DDR2 offers bandwidth advantages for all of their cpu's, and going to DDR3 from DDR2 will be easier.

Probably, @ least in theory & for certain "high-memory bandwidth intensive apps" (whatever those are... games, scientific applications?)... but, I have to also lean to chipfab design & retooling too, keeping it "future-ready" & consistent across lines in the future as well!

(I think, actually, you guys are BOTH right to a HIGH degree on both your points, on that last account (economics vs. performance only) - so far @ least!)

APK

P.S.=> Gotta run & do Sunday dinner with my Mom & niece (she always brightens up my day, lil' kiddos always usually do)... so, I will read your posts later fellas!

Stay cool & keep it factual so "lamos" like me (lol, in hardware admittedly nowadays) gain... apk

 

[v-zero]

The move to DDR2 is purely economic on the desktop side. DRR2 is quickly becoming cheaper than DDR, uses less power, is less likely to suffer redundancy problems (and hence is more suitable for server use), and above all it "looks" better than DDR1 from a PR stand point...

I shall point you here: http://anandtech.com/cpuchipsets/showdoc.aspx?i=2741 To help you understand the error in what you are saying. It doesn't go into the detail about the core that I'd like it to, but then, not many people look at these things and like to understand them on a level beyond the "numbers game."

I do not mean to be rude, or seem cockey, I just happen to know what I'm talking about in this area...

By the way. HTT stands for Hyper Transport Technology... lol - I have no idea what you thought it stood for, but... That HTT is used along with a divider of sorts to decide the CPU's speed, since this second multiplier is needed to emulate the existance of an FSB.... the HTT speed multiplied by the HTT multiplier (sort of) determines the speed of the HT bus, and hence the onboard memory controller.
These memory controllers are subject to little latency onboard, and hence high latency DDR2 hurts them... This cannot be made up for by higher bandwidth, that is not at all how it works - they are not equal in what they do. The AM2 is still only the K8 by the way, K8L is not coming until 2007...

Oh, and this memory controller is essentially the same. It has been tweaked to allow DDR2 use, and higher bandwidth use. You cannot "optomise" these controllers to work well with poor latency, because they were never originally designed to be used with mega-high bandwidth and high latency solutions.

 

[intel igent]

http://www.pccanada.com/viewitem.asp?id=4535

how close is it?

 

[XooM]

why are we debating the term "HTT"? It's completely irrelevant to the discussion, we're all just trying to sound smart. we know exactly what everyone means when they say "HTT" or "FSB" in regards to athlon 64s: they mean the speed at which the intergrated memory controller is communicating with the RAM. FSB and HTT have simply been picked as misnomers for this connection. stop debating that, because who cares.

secondly, until you can start coming up with some solid facts and decent sources for the argument that the athlon 64 will gain nothing by going to DDR2, you're simply spewing your own opinion and your point is empty. same goes for claiming it'll offer a performance boost. no solid facts, no solid argument. every argument in this thread is based on hearsay, until you have an AM2 ES in your hands or until AM2 is actually launched, your argument is meaningless.

 

[v-zero]

I posted a link to that Anandtech article, how is that not backing my point? It gaisn no more than 8% in real-world performance with some seriously expensive DDR2...

HTT and FSB are not the same. I don't care if you talk about them being the same in an overclocking sense - that's fine, but they are not designed for, and do not do, the same thing...

I don't base my "opinions" (as you put them) on hear say, I am not one to misinform - I say these things because they are what I understand to be the case from a technical side. Having read a lot of things on K8, and the finer details on the comparison between K8 and Conroe (on Anand amongst other places), I am in a position to speak about these matters...

 

[XooM]

ok, the anandtech article. thats good. how about some more like that?

and just shut up about HTT and FSB. it doesn't matter for the present argument so just drop it already. they're not the same and we know it, move on.

 

[v-zero]

Hey, all I'm saying is that I doubt very much that AMD would introduce two new sockets and a line of processors equipped with a new memory controller because it saved a few watts here and there. K8L is will be competing with conroe, and they know it. AM2 is not a socket based entirely around power consumption.

I agree completely.

To compact what I'm saying; For K8 at least DDR2 offers no big boost - no Conroe beating hopes. However it lays the path for quad-core, and K8L. We will have to wait and see if major core enhancements on K8L can make DDR2 a winner for AMD in the long run...

 

[XooM]

DDR2 is a win no matter what; better bandwidth, better latencies; its just not a conroe-beating win on K8... what would drastically help AMD vs conroe for now is if they'd bin more aggressively; AMD could easily bin chips at 3 ghz, plausibly even at 3.2, which would greatly close the gap with conroe in many areas.