Where are the 4*4 against KENTSFIELD tests?

[NightlySputnik]

Hi everybody.

Did I miss something or what? 4*4 came out only 1-2 days later than Kentsfield if I'm not wrong. Why is it then that nobody has posted test of these against older A64FX62 and new Kentsfield system?

I know that only Asus will build the board for it with NVidia chipset. Is it the reason or is AMD pushing back as possible these because of lower performances? Or was it posponed and I didn't read about it? That would be a surpirse, but I haven't had much time lately so it's possible.

Let me know please and/or link if these test do exist. Thanks

 

[Heyyou27]

So far there has been no publishing of any comparison between 4x4 and Kentsfield.

 

[digitaldoc]

No numbers anywhere yet, but the AMD faithful are eagerly awaiting some data to see if 4x4 changes anything or not.

 

[epsilon84]

No numbers anywhere yet, but the AMD faithful are eagerly awaiting some data to see if 4x4 changes anything or not.

Yeah, some are very eager.... unfortunately, I believe they will be disappointed, not in the raw performance, but I think they are hoping (regardless of price) it shows stuff to out do C2D....


You mean C2Q.

If 4x4 can't beat C2D (in multithreading apps of course ) then AMD's got some serious problems.

 

[epsilon84]

4x4 was delayed, rumor is November 30th.

Thats what the inquerer says (check out amdzone for more info)

But I wouldn't get a 4x4 if your into speed (yes, I am an amd fan). People are actually better with an x6800 unless if they extreme multitask, and thats where 4x4 kills kentsfield because the cores don't share stuff like ram, allowing it to scale up faster

Wow, so you've compared Kentsfield and 4x4 yourself to come to this conclusion? What constitutes 'extreme multitasking' anyway? Oh, yeah, AMD's marketing term 'megatasking' LOL.

Reviews of 4x4 aren't even out (heck there aren't even any leaked benches) and you're jumping to conclusions already.

I guess it perfectly illustrates Jumping Jacks's point about AMD fans being 'very eager' to see 4x4 deliver some saving face for AMD.

 

[MoNeY3865]

Relax a bit Epsilon, how many months of fanboy speculation and bickering did the readers of this forum have to endure before the Core 2's came out. Let's not start that again.

At least there is a little server data (4 and 8 Opteron cores) that seems to suggest what I Love Tacos said. Saying the 4x4 Kills is speculating a bit too much, but he didn't sound like he was foaming at the mouth like so many Intel fanboys and their Core 2s.

Intel is the undisputed champ and there's no need to raise blood preassure defending them. Wait for the 4x4 data and be content with the fact that Intel will probably still rule the desktop and gaming domains.

 

[Synthetickiller]

4x4 was delayed, rumor is November 30th.

Thats what the inquerer says (check out amdzone for more info)

But I wouldn't get a 4x4 if your into speed (yes, I am an amd fan). People are actually better with an x6800 unless if they extreme multitask, and thats where 4x4 kills kentsfield because the cores don't share stuff like ram, allowing it to scale up faster

Wow, so you've compared Kentsfield and 4x4 yourself to come to this conclusion? What constitutes 'extreme multitasking' anyway? Oh, yeah, AMD's marketing term 'megatasking' LOL.

Reviews of 4x4 aren't even out (heck there aren't even any leaked benches) and you're jumping to conclusions already.

I guess it perfectly illustrates Jumping Jacks's point about AMD fans being 'very eager' to see 4x4 deliver some saving face for AMD.

I'm an AMD fan, at least of the Thunderbird way back when it was the new kid on the block for fast CPUs. I run dual opterons now and am looking at another CPU to buy.

Now, with that background, you'd expect me to say "the 4x4 will kill the 2cQ".

Well, no. I'm disappointed in AMD actually. They need to step up. If the 4x4 does NOT deliver the power / price point it needs to, then I'm going intel, no questions asked.

Until I, or anyone else for that matter, sees bench marks, I can't make any comparison or even speculation about which company has a better product. If Bench Marks come out on 11/30/06, then yes, we'll compare and figure it out.

Until then, no more speculation. its just boring and well, I don't have time to read more BS.

 

[quantumsheep]

Relax a bit Epsilon, how many months of fanboy speculation and bickering did the readers of this forum have to endure before the Core 2's came out. Let's not start that again.

Wouldn't say endure, i actually quite enjoyed watching all the squabbling. Was like watching a soap on TV really.

 

[TabrisDarkPeace]

Will the new K8 cores (latest revisions) used on 4x4 have Load before Store ?

Similar to what the Core 2 Duo / Xeon 5100 - 5300 have (which is why they scale so well with 4 pipelines).

...and if so, with these K8 4x4 chips have 3 pipelines, or 4 ?, and will they be able to execute SSE instructions in (near or under) one clock cycle at peak throughput ?

 

[Stimpy]

I don't have any comparisons of 4x4 with a single socket intel quad core, but I do have some good comparisons (in a server environent) of a dual socket quad core against a 4 socket dual core opteron at
http://www50.sap.com/benchmarkdata/sd2tier.asp

Taking the same HP server model

Intel.
HP ProLiant BL480c, 2 processors / 8 cores / 8 threads, Quad-Core Intel Xeon Processor X5355 2.66 GHz, 64 KB L1 cache per core, 4 MB L2 cache per 2 cores

Result: 9050

AMD.
HP ProLiant BL45p G2, 4 processors / 8 cores / 8 threads, AMD Opteron processor Model 8218 2.6 GHz, 128 KB L1 cache and 1 MB L2 cache per core

Result: 9570

Higher is better. Note this is a heavy multiuser/multitasking application. Note average response time is the same (1.98s).

My conclusion? Although per core the new Core2s are faster than anything AMD has at the moment as soon as you start scaling out with number of cores and sockets the advantage disappears very quickly

The funny thing is both x86 architectures beat the crap out of Itanium2 dual core

i.e

Itanium 2
HP Integrity rx4640, 4 processors / 8 cores / 16 threads, Dual-Core Intel Itanium 2 9050 1.6 GHz, 32 KB(I) + 32 KB(D) L1 cache, 2 MB(I) + 512 KB(D) L2 cache, 24 MB L3 cache

Result:7070

 

[dlmacline]

4x4 will probably won't suck.. (cause if they do what is the use of releasing and wasting resources)

but it will probably be a hedgehog on your money like feeding aircraft gasoline in your automobile

 

[NightlySputnik]

But I wouldn't get a 4x4 if your into speed (yes, I am an amd fan). People are actually better with an x6800 unless if they extreme multitask, and thats where 4x4 kills kentsfield because the cores don't share stuff like ram, allowing it to scale up faster

Like I said, I'll wait for real test as oppose to theory. That's what the different tells us, Kentsfielg sharing through FSB and 4*4 sharing through HT link. But again, if Jumping Jack isn't wrong,

This is not quite entirely true. Kentsfield, while not having an on-die memory controller, does access memory uniformly. 4x4 will basically the same as a dual core opteron with two channels of DDR2 wired up to each core. This arrangement is called NUMA, or non-uniform memory access. You are mistaken that memory is not shared between all cores (main, not cache) -- in a multi-socket system, at boot the CPUs setup the address table and physical location of all memory installed and all CPUs have access to all installed memory.

The conglomerate bandwidth will increase, no doubt, but on typical destktop applications bandwidth is not the premium (evidence -- AM2 increased bandwidth 30% but no performance improvement at all). Where 4x4 will show it's fault compared to the uniform access model of kentsfield is in latency.

To make this explanation simple, call the memory for socket 1 DDR2-A and CPU-A, and on socket 2 call it DDR2-B and CPUB. If a thread is running on CPUB, has a cache miss and requires data from DDR2-B no big deal, it resides on the memory connected to that CPU; however, if that data resided in DDR2-A, then the data needs to interrupt the IMC on CPUA, fetch the data, send the data over the cHT, and load the traceline into CPUB. This creates latency, in fact, it has been shown to double it.

Kentsfield will not suffer from this.

it might not be such a problem anyway. I also remember poeple here saying CONROE wasn't gonna perform because of that same FSB and look at the number.

Saying the 4x4 Kills is speculating a bit too much, but he didn't sound like he was foaming at the mouth like so many Intel fanboys and their Core 2s.

Intel is the undisputed champ and there's no need to raise blood preassure defending them. Wait for the 4x4 data and be content with the fact that Intel will probably still rule the desktop and gaming domains.

For me, everything AMD does look like when Intel was gonna release first PentiumD. They knew it was gonna make poeple talk but also that it would be an underperformer and consume as much electricity as an old 440 block street racer consume gas. So they talked about it but took their time to show us anything concrete. I only hope I'm wrong.

Until I, or anyone else for that matter, sees bench marks, I can't make any comparison or even speculation about which company has a better product. If Bench Marks come out on 11/30/06, then yes, we'll compare and figure it out.

Until then, no more speculation. its just boring and well, I don't have time to read more BS.

No offence anybody, but exactly what I ask. Some real number and no "I think that..." So basicly, like all of you, I'll keep on looking around.

Thanks for your input everybody, it's much apprecieted.

 

[NightlySputnik]

Agree partially. I didn't know it was pushed back to 11/30. So I agree about that date for benchmark. Usually tough, site have access to these system before released to have the test ready for that date. Unless it's a paper launch to steal some of Intel headline with Kentsafield released. I would understand, but we all blamed Intel not too long ago for doing the same, so it's not fair in that case.

Anyway, 1-2 more week to wait isn't a problem anyway

 

[NightlySputnik]

its not like a site will do a full analysis of the parts in one day.

In case I wasn't clear enough, that's why I said that magazine and review site that have signed NDA (non disclosure agreement) usually have access to the system 1-2 weeks before official released, so they can have their review ready on time. And those that don't respect these NDA have to pass their turn the next big released, which is usually a bid enough reason to listen to them.

Anyway, we'll see.

 

[niz]

Not "megatasking" platformance (what that is supposed to mean) is AMD's new slogan for 4x4

Megatasking platformance?

Holy crap that really takes the prize for meaningless technical-sounding salespeak.

I wonder how long it will be before intel come out with gigatasking multiformance?

 

[quantumsheep]

My conclusion? Although per core the new Core2s are faster than anything AMD has at the moment as soon as you start scaling out with number of cores and sockets the advantage disappears very quickly

The funny thing is both x86 architectures beat the crap out of Itanium2 dual core

i.e

Itanium 2
HP Integrity rx4640, 4 processors / 8 cores / 16 threads, Dual-Core Intel Itanium 2 9050 1.6 GHz, 32 KB(I) + 32 KB(D) L1 cache, 2 MB(I) + 512 KB(D) L2 cache, 24 MB L3 cache

Result:7070

Did you not realize that the Itanium 2 in question was running at 1.6ghz vs the 2.6/2.66ghz of the K8 and Conroe ?

 

[qcmadness]

Did you not realize that the Itanium 2 in question was running at 1.6ghz vs the 2.6/2.66ghz of the K8 and Conroe ?

Do you know Itanium 2 architecture can do 6 instructions per clock?

 

[BaronMatrix]

Also 4x4 was designed to take on conroe, and I think their main fear is that it can't do that, because it can only multi task better, its not like doubling the frequency of the fx processors

QFX was designed to give an increase to FX. Everyone knows they can't double the clock. It would have to be 65nm to even get more than 3.6GHz.

AMD is gaining customers not losing them. WHen the quads come out they will need greater differentiation. Two sockets will do that. I never bought an FX. Now I will. I know it won't outdo C2Q per clock, but I like AMD and will be glad to have 8 cores next year.

 

[NightlySputnik]

Also 4x4 was designed to take on conroe, and I think their main fear is that it can't do that, because it can only multi task better, its not like doubling the frequency of the fx processors

QFX was designed to give an increase to FX. Everyone knows they can't double the clock. It would have to be 65nm to even get more than 3.6GHz.

AMD is gaining customers not losing them. WHen the quads come out they will need greater differentiation. Two sockets will do that. I never bought an FX. Now I will. I know it won't outdo C2Q per clock, but I like AMD and will be glad to have 8 cores next year.

I don't want to blow your fanboy bubble Baron (no offence), but I really don't think that these 4x4 board will last very long. Otherwise AMD will have ask more than NVidia AND Asus to do these board. 1 chipset + 1 manufacturer = restricted availability.

On my side of the fence, I don't like neither Intel or AMD more than the other. As a matter of fact, AMD is now loosing some of it's better looking side by going Intel way. You know, reacting instead of acting, keeping all the good stuff for close friend, shitting on the others head, etc... Don't get me wrong, Intel is doing the same. It's only that they seem to get less selfish while AMD seem to go the other way. We'll know for sure in about a year I guess.

And I forget, AMD is gaining market just like Intel did 2 years ago with inferior product. It takes time for the average Joe to know what's happening in the tech world. My point is, superior market share doesn't mean superior product. Otherwise AMD would have had more than 50% within the last 2 years, or sell whatever they could for much higher price than Intel to compensate their performances advantages. But it wasn't the case.

Truth is: right now Intel dominate AMD in pure performances and nobody here knows about when 4x4 will come out if the situation will change. This was the original point of that tread by the way

 

[orsino]

Megatasking platformance?

Holy crap that really takes the prize for meaningless technical-sounding salespeak.

I wonder how long it will be before intel come out with gigatasking multiformance?

Virtualizing Speedstep L2 cache 775-land 64 cubed 65nm processor with 128-bit Streaming SIMD Extension, 1066 FSB and Execute Disable Bit.

In the immortal words of Jessica Simpson:

I totally don't know what that means but I want it.

:lol:

 

[orsino]

Well, AMD can try to get people with more disposable income than IQ points to buy something flashy but underperforming, however market dynamics usually indicate that the savvy enthusiast-level consumer keeps a close eye on the price/performance chart. I can't see AMD dropping their top end FX pricing to the point where two of them cost less than just one C2Q, so the question I want to ask AMD is the same that Anthony Bourdain did on yesterday's Top Chef: "What kind of crack house are you running here?" :lol:

 

[BaronMatrix]

I don't want to blow your fanboy bubble Baron (no offence), but I really don't think that these 4x4 board will last very long. Otherwise AMD will have ask more than NVidia AND Asus to do these board. 1 chipset + 1 manufacturer = restricted availability.

When you consider that this is FX and has a much lower volume and will only be OEM in systems, using one manuf for the launch isn't a bad thing. That way if it doesn't take off, there aren't a bunch of mobos collecting dust.

I'm sure that they are using this as their first "reference" platform and the next will be ATi and the other manufs. But again, how many of these do you anticipate being sold? I doubt if C2Q gets high volume either for the first six months.

 

[orsino]

I'm holding off on my C2Q/DX10 buy until early summer hoping that by then the prices will have come down considerably on both the CPU and the cards. I'm going that way since I want the next system to last me an absolute minimum of three years, four would be better, without cracking the case every month to update this that or the other thing. Now all I have to do is find a whole whack of hobos to rip off and I'm set! :lol:

 

[quantumsheep]

Did you not realize that the Itanium 2 in question was running at 1.6ghz vs the 2.6/2.66ghz of the K8 and Conroe ?

Do you know Itanium 2 architecture can do 6 instructions per clock?

No, actually i didn't! But surely that would not offset the 1ghz disadvantage that it is put at when competing with those 2 other CPUs.

 

[mcain591]

i don't think one topic on this forum doesn't get reacted to fanboys and flamers alike.