AMD K8L: Next generation CPU

ownage said:
It will be good until July 23rd, I gurantee it. K8 is getting old. It used to be good, but not anymore. Core 2 will destroy it. I dont see AMD setting any records in 3D anymore. Even less in Super Pi 1M AND 32M. Conroe has better clock for clock performance than Windsor or Toledo. K8s are getting extremely hot too. 125W TDP, that is pushing the envelope toward Prescotts. Is K8L a new type of architecture? No, its only a tweaked K8. 4 core? give me a break, we barely use two right now. Intel automatically diverts to 2T, so what? It makes up for it in the higher FSB. The only thing that K8 have is high memory bandwidth. Why? cuz of IMC. Even with recommended memory, K8 will still go to 2T with 4 DIMMs populated. It is very difficult to achieve 1T with 4 DIMMs. AMDs K8 is getting old and it needs to be retired. They can try to tweak it even more, but the performance isnt going to be a drastic improvement. While Core 2 has a whole new road in front of it. There is more room for expansion. I would really like to see a 9sec super pi from ANY AMD processor.

I'm really not wanting to argue with you, but you seem to be throwing up so many unvalid points it's not funny. 1st, AMD is still leagues ahead of Intel in the 3D gaming arena-
http://tomshardware.co.uk/2006/05/23/amd_reinvents_itself_uk/page47.html
http://tomshardware.co.uk/2006/05/23/amd_reinvents_itself_uk/page48.html
So that shatters that point of them not making any new '3D records'.

Your point about the 125W thermal output is only for AMD's over the top FX-62. If you look at the following page, with the charts, you'll actually notice that all but the FX62 AMD CPUs thermal output has been lower, quite dramatically in comparison with there older models.
http://tomshardware.co.uk/2006/05/23/amd_reinvents_itself_uk/page16.html
Look how it still compares with Intels offerings. Can I also point out that your comparing it with heat dissapation from a currently released Intel CPU for the Laptop market. Your forgetting also, this is only when the CPU's on full pelt, cool'n'quite keeps it cool when not being fully utilised.

Your better performance, clock for clock rant, look through ALL the benchmarks here.
http://tomshardware.co.uk/2006/04/24/dual_core_intel_processors_for_low_uk/index.html
You'll notice that the 2.16Ghz Coreduo and the 2GHz AMD Athlon64x2 3800 are probably the most similar in benchmarks. The coreduo slaughtering it in some benchmarks (Audio encoding), the Athlon slaughtering the Core duo in others (Gaming). Other benchmarks there roughtly the same. It should be noted this is one of the 'older' 3800+, not the AM2 with DDR2 etc etc etc. So, if you were to imagine this as an AM2 AMDx2 5000+, with DDR2 etc, I think, the coreduo would have had it's pounding. Clock for clock, impressive, but not a huge deal greater than AMD, or if any better at all.

Is K8L a new type of architecture? No, its only a tweaked K8. 4 core? give me a break, we barely use two right now.
Hmm, well, the architecture of the AMD allows 4 cores easily to be implemented using one of the 3 Hypertransport buses actually designed for the Athlon 64's (that breathing room that was originall designed into it for upgrading i future). AMD cant use intels Hyperthreading so it will go 4 core to compete. It's on it's road map. Give you a break? lol

Intel automatically diverts to 2T, so what? It makes up for it in the higher FSB.
Hmm, so it's ok for Intel, but you stab at AMD for the same problem? Fanboy behavour??? What I mean is, no system currently out can run 4 1GB DIMMS at 1T, it's too unstable (i.e. not perfected) on any platform. As for the higher FSB, clever. Is that the same FSB that Intel has to use to communicate to the entire system? You know, PCIexpress through the northbridge and such? In an AMD system, the IMC means the CPU has the full bandwidth of the memory at the same time and using the full bandwidth of 2 SLI PCIexpress graphics cards, the SATA Raid HDDs, the DVD running there too. All this congest the OLD FSB architecture when the CPU wants to use the full bandwidth of the memory at the same time. Hint, thats why individual synthetic benchmarks (the one's you seem to cherish) do good on an Intel, you higher FSB. In the real world, when everything wants to use your lovely high FSB at the same time, it suffers. Besides, where is the huge difference in FSB?

Right, taking this all into consideration, and the fact this is all about current AMD K8 CPU's against Intels latest, with the K8 tweaked with bolted on 4 cores, with 2GHz Hypertransport buses, lower manufacturing process etc etc etc, there is still no problem with the K8 series. What would you replace it's architecure with? Intel still needs to let go of the FSB only setup and the lack of a IMC on a coreduo limits it's potential to the northbridge already. It's all great having a low wattage CPU, but not if your northbridge is having to make up for it. Architectly wise, even the K8 has better potential than the Coreduo! - if you think I'm wrong, give me a few examples that matter where the K8 is not. Show me where that room for expansion is. Just because the K8 is currently using it's 'room for expansion' doesn't mean it's to be retired just yet. And, it's still got the lower manufacturing processes to start on that Intel are already on, not to mention AMD's SOI patented technology.

The K8 plant been growing nicely for the last couple of years, I think it's just about to flower.
 
First off, I dont like Tom's Hardware, for me, they seem a little biased. And that review was a joke, just like the Core Duo vs X2 3800+ at anandtech. That is comparing apples to oranges. A mobile processor to a desktop processor. How about comparing Conroe and AM2? How about an ORB comparison? http://service.futuremark.com/compare?2k1=8982813 (Intel)
http://service.futuremark.com/compare?2k1=8928671 (AMD) It is 3DMark01SE because that is more cpu dependant than 03, 05, and 06. So it will show you who has the better cpu rather than video cards. As you can see Intel has the whole first page. Populated by Conroes and Meroms and P-Ms, yes Pentium M.

Lets see a AM2 do this:
21k0um.jpg

And this:
188k8jg.jpg

And this:
x6800499810s3129pe.gif


The first two are FUGGAR's runs. All credit go to him. The second one is TAM and Team Japan. All credit go to them too.

FUGGAR's setup: Intel 975XBX-304
Intel Conroe 2.4Ghz
Astek Vapochill LS modded
Vcore 1.50v
4GB OCZ PC8000 @ SPD 1:1
or
2GB Super Talent @ SPD 1:1
X1900XTX Crossfire
4x RaptorX 150GB Raid 0
Gigabyte I-Ram

TAM's setup: Clock: 4997.9MHz (fsb416.5MHz x12)
CPU: Core 2 Extreme X6800 : (2.93G/266x11/FSB 1066/L2 4MB)
M/B: Intel D975XBXLKR-rev302 mod (bios:1073)
CPU Cooling: LN2 (Pro4 + V-TEC ARC Bed Rev.3.4)
Memory: Corsair PC2-5400UL(v1.3) 512MB x2
VGA: ATI rageXL (PCI)
HDD: IBM DTLA-307020
Power Supply:Zippy-460w
Vcore: 1.71V, Vdimm: 2.55V, Vio: 3.65V
OS: Windows Server 2003
MEM: 4:5, CL4-3-3-2

I just found this too: http://service.futuremark.com/compare?3dm05=1901399
It takes an AMD FX-60 @ 3.7Ghz, UTT-BH @ 2-2-2-5, CF X1900XTs at 890/850 AND DICE to just get 20k. While Conroe at 3.6 gets 21k and still has more room to go.
 
I love how (connchri) you use Tom's Hardware, after all, they are one of the most biased review sites around. Try Anandtech or X-bit Labs instead.
 
I think AMD may eventually switch to co-processing technology to help speed up PC performance. They want to initiate this using the new Opteron server CPUs. Can you imagine having 3 different coprocessors together? With each one dedicated to specialized tasks. All three processors would communicate directly with Opteron processors and the system chipset synchronously. What do you think about that idea? Looks good.

Torrenza is the new platform that AMD want to initiate that will utilize next generation multi core 64 bit processor that have the capability to work alongside specialized coprocessors.

 
You know, I find it funny how you both claim that Tomshardware is biased. The same was said when I used to in a discussion supporting a P4 630. Tell me, are they biased to Intel or AMD, or perhaps they change from month to month, let me know the pair of you because I'm pretty confused about all this biased nonsense. Oh course, Andantech must also be biased because you don't agree with them (apart from Alvino, you seem to think Anandtech is cool, shame they say the same as Tomshardware, talk about Irony). Silly me. I forgot, you have all the hardware and experience of both tomshardware and anandtech to know better.

Right, enough of the bitching.....

I'll work through your post backwards, it's easier.
It takes an AMD FX-60 @ 3.7Ghz, UTT-BH @ 2-2-2-5, CF X1900XTs at 890/850 AND DICE to just get 20k. While Conroe at 3.6 gets 21k and still has more room to go.
Right, lets cut out all the crap. So your saying that a performance difference of 8% per clock cycle (I've just done the math) between the Conroe and current AM2's (Current ones, not the ones AMD are planning on releasing with 4 cores and DDR3 - you know, the ones the Conroe will eventually compete against) validates the retirement of the K8 architecture - and thats using your figures of 3.6GHz Conroe getting 21k, so I'm taking a leap of faith her with you (even though your Cronoe @5GHz takes 10.3sec for Pi, and not the 9 you claimed a standard Cronoe could in your last post). If thats what you think it takes to rid the K8, then I'm glad you ain't working for AMD.

Your two links you posted with the 3D mark 2001 scores. I'll point out now I suppose, that the Cronoe is clocked over 200MHz faster than the Athlon setup. Oh, and in the description, it seems that the Cronoe has a X1900XTX, while the Athlon has a X1900. Not 100% sure what the difference is, but I'm sure the XTX is 'go faster' somehow. Ontop of that, and perhaps more to the point, Why 3Dmark01???? You say it's more CPU limited? then why on scimp on 3Dmark06 with identical hardware except the Mobo and CPU, there's bound to be someone out there done it and it tests the latest CPU's to the fullest, not with 2001 code and 2001 intruction sets! The P4's beat the Athlons in 3Dmark01, hardly a reallife comparison is it. Again, even if the Intel setup is faster then the AMD, with these differences in the test setup, it's not enough to dismiss a future K8 CPU is it?!?!? PS, course the graphics card is tested in the 3DMark01 benchmarks, geeze!

Where's the real game framerates here? the ones that count (for the reasons in my last post), not benchmarks (even though they say the same as me). Real life situations would be good.

Just to clarify too, by the way. Before you think I am dissing the whole Intel Conroe thing and think that it is a bucket of balls, I am not. I think it is a gonna be a gobsmacking CPU. I just completely dissagree that the K8 architecture needs rid off. Again, for the same reasons in my last post. It is still a worthy CPU, and a worthy competitor. The only difference now is, Intel for once has a decent competitor.

I'll repost the end of my last post too as you never asnwered the question(s) in it -

Right, taking this all into consideration, and the fact this is all about current AMD K8 CPU's against Intels latest, with the K8 tweaked with bolted on 4 cores, with 2GHz Hypertransport buses, lower manufacturing process etc etc etc, there is still no problem with the K8 series. What would you replace it's architecure with? Intel still needs to let go of the FSB only setup and the lack of a IMC on a coreduo limits it's potential to the northbridge already. It's all great having a low wattage CPU, but not if your northbridge is having to make up for it. Architectly wise, even the K8 has better potential than the Coreduo! - if you think I'm wrong, give me a few examples that matter where the K8 is not. Show me where that room for expansion is. Just because the K8 is currently using it's 'room for expansion' doesn't mean it's to be retired just yet. And, it's still got the lower manufacturing processes to start on that Intel are already on, not to mention AMD's SOI patented technology.


Reply with a technical argument why the K8 sucks, not a bunch of pro-cronoe benchmarks and a 'every review that says the 3800+ beats a cronoe is biased' attitude.
 
TRDCorolla said:
I think AMD may eventually switch to co-processing technology to help speed up PC performance. They want to initiate this using the new Opteron server CPUs. Can you imagine having 3 different coprocessors together? With each one dedicated to specialized tasks. All three processors would communicate directly with Opteron processors and the system chipset synchronously. What do you think about that idea? Looks good.

Torrenza is the new platform that AMD want to initiate that will utilize next generation multi core 64 bit processor that have the capability to work alongside specialized coprocessors.


Sounds interesting. Bet it'll be a while before it works down to the desktop CPU's though.
 
I never did say that Conroe can finish super pi 1M in 9 seconds. I said i want to see an AMD processor that can do it. K8 needs to be retired. All they are doing is upping the clock speeds and die shrinks. Thats about it. It went from the Clawhammers/New Castle to Winchester then to Venice/SD. 130nm to 90nm to a more refined 90nm. The Windors are pushing the thermal envelope of cpus. There is no head room left in K8. Sure you can just slap on 2 more cores, but that will only raise the TDP even more. The only article that i disagreed with Anandtech was the 3800+x2 against a Core Duo. It wasnt right to compare desktop to mobile. They should have done it with a Turion X2 but alas, there was no Turion X2 out at that time. So there you go. K8 are getting too hot, the IMC is good, but there needs to be better watt to peformance ratio. AM2 are basically turning into Intels Prescotts if they keep going on the same road. Anandtech is a very trustworthy source, Toms Hardware not so much. As i dont like their reviews and how sometimes they get different scores than other major reviews with similar setups. And how is having a northbridge making up for the low TDP processor? I dont get your logic. Even without an IMC the Core 2 Duo spanks the K8. Now with a IMC, which Intel plans to implant in 2008, how will the K8 fair against it? It simply wont keep up even if you do slap on 8 cores on a single die.
 
ownage said:
I never did say that Conroe can finish super pi 1M in 9 seconds. I said i want to see an AMD processor that can do it. K8 needs to be retired. All they are doing is upping the clock speeds and die shrinks. Thats about it. It went from the Clawhammers/New Castle to Winchester then to Venice/SD. 130nm to 90nm to a more refined 90nm. The Windors are pushing the thermal envelope of cpus. There is no head room left in K8. Sure you can just slap on 2 more cores, but that will only raise the TDP even more. The only article that i disagreed with Anandtech was the 3800+x2 against a Core Duo. It wasnt right to compare desktop to mobile. They should have done it with a Turion X2 but alas, there was no Turion X2 out at that time. So there you go. K8 are getting too hot, the IMC is good, but there needs to be better watt to peformance ratio. AM2 are basically turning into Intels Prescotts if they keep going on the same road. Anandtech is a very trustworthy source, Toms Hardware not so much. As i dont like their reviews and how sometimes they get different scores than other major reviews with similar setups. And how is having a northbridge making up for the low TDP processor? I dont get your logic. Even without an IMC the Core 2 Duo spanks the K8. Now with a IMC, which Intel plans to implant in 2008, how will the K8 fair against it? It simply wont keep up even if you do slap on 8 cores on a single die.

Most of what you've said I'm mentioned already, so I'll keep it short DDR3 is not a simple refinment yet it's in the works. Additional cores, was planned and catered for with the potential 3 Hypertransport buses. The latest AM2s have a TDP30% lower than there SKT939 counterparts - there getting cooler, not hotter, all except the FX60 as mentioned before. Agreed, you didn't say a Intel CPU could do superpi in 9s, but you wanted a AMD to - bit of a wasted point. A Intel chipset will have a higher TDP than a AMD one because it also caters for additional and faster buses such as for the memory. By 2008, who knows what either side will have in the works. AMD, as TRDCorolla mentioned has Torrenza in the works as well as other stuff Here (Kudos to TRDCorolla again).

Lets call it the end now. We obviously have completely different views on the K8 and were not gonna change.
 
Whatever. :rolleyes: I'm not going to even argue with you. I didn't even know Ownage posted or what he even posted until I posted. Either way, well all have different viewpoints and opinions so whatever.
 
By 2008 Intel will be implimenting CSI into all of their processors. And K8 is old, K8 is pretty much K7 with an IMC and higher clock speeds and longer pipelines. 10.3s is pretty damn close to 9 seconds. Which is quite a feat for Intel. Im not bashing AMD, they have dominated for 4 years straight. And those 4 years were because of K8, but dont you think that after 4 years, AMD will need a new architecture? AM2 cant keep up with conroe. Personally, I think that AM2 should have been the S939. But AMD has been truely remarkable. But i really think its time for Intel to take the crown for a couple years. And have AMD whip out something new to keep up.
 
Back
Top Bottom