K10 Scores starting to surface

[SEA]

Capabilities
	HyperTransport Capability
		Offset			A0h
		Revision		3.00
		Interface type		Host/Secondary
		Device number		0
		...

Does Revision 3.00 mean the same as HT3?
If so, where is that data from, i mean what chip shows it?

[lukija]

Does Revision 3.00 mean the same as HT3?
If so, where is that data from, i mean what chip shows it?

From PCI device. For example you have memory controller as PCI device and it has some configuration and timings registers which you can access or change some parameters (like A64Tweaker,A64Info ....). The same goes for HT Link controller.

[JumpingJack]

Does Revision 3.00 mean the same as HT3?
If so, where is that data from, i mean what chip shows it?

HT3.0 has been moved in the roadmap from 2007 to 2008, the chipset will support HT3.0, but it appears the CPU will not:
http://www.amd.com/us-en/assets/cont...ck_Bergman.pdf
See slide 19.

HT3.0 support in the CPU appears to have been moved to the 45 nm sandtiger product.

[Mad_Man]

You're right.
The dump I got reveals several problems with the HT link of the CPU. According to the config regs, the CPU HT link width is set to "Link physically not connected".
I don't exactly know if this is possible that the system works in such conditions, however, there's definitely something odd.

Capabilities
	HyperTransport Capability
		Offset			A0h
		Revision		3.00
		Interface type		Host/Secondary
		Device number		0
		Link 0 frequency	200MHz
	HyperTransport Capability
		Offset			C0h
		Revision		3.00
		Interface type		Host/Secondary
		Device number		0
		Link 0 frequency	200MHz
	HyperTransport Capability
		Offset			E0h
		Revision		3.00
		Interface type		Host/Secondary
		Device number		0
		Link 0 frequency	200MHz

so now, this is interesting...
so it means that its running only at 200htt? that gives roughly 1.6GBps bndwidth...
pair this with non-tweaked bios and 667mhz DDR2 Cl5 and you have current K8 performance.
i can imagine that phenom at full 1000mhz htt, good bios and ddr2 1066 Cl4 should fly...

[Lightman]

HT3.0 has been moved in the roadmap from 2007 to 2008, the chipset will support HT3.0, but it appears the CPU will not:
http://www.amd.com/us-en/assets/cont...ck_Bergman.pdf
See slide 19.

HT3.0 support in the CPU appears to have been moved to the 45 nm sandtiger product.

You right! But if you follow that presentation then you would find on page 20 that Pinwheel platform based around Athlon X2 AM2+ and AMD690G is supporting HT3.0. Wired isn't it???

Either person responsible for these slides was drunk or heavily misinformed
At the moment I'm still thinking Phenom will have HT3.0 but not Barcelona....

[hawkeyefan]

You right! But if you follow that presentation then you would find on page 20 that Pinwheel platform based around Athlon X2 AM2+ and AMD690G is supporting HT3.0. Wired isn't it???

Either person responsible for these slides was drunk or heavily misinformed
At the moment I'm still thinking Phenom will have HT3.0 but not Barcelona....

Wait...wouldn't it be the other way around? Ideally anyway?

[[XC] leviathan18]

hope to see the real deal soon.... im sick of all the speculation going around barcelona

[JumpingJack]

You right! But if you follow that presentation then you would find on page 20 that Pinwheel platform based around Athlon X2 AM2+ and AMD690G is supporting HT3.0. Wired isn't it???

Either person responsible for these slides was drunk or heavily misinformed
At the moment I'm still thinking Phenom will have HT3.0 but not Barcelona....

I noticed this shortly after I posted and went through more of the presentation, it may be a typo or they are one way or the other.... so it may be that they mistakenly put HT1.0 when they meant HT3.0.

[red]

http://forum.coolaler.com/showpost.p...&postcount=125
Harpertown 2GHz numbers for comparison!

23.3s vs 39.7s for Pi
20.48 vs 17.13 in "relative speed" for Fritz
2454 vs 1896 in single cpu cinebench
15334 vs 13295 in mutli cpu cinebench
6.25x vs 7.01 cinebench scaling (78 vs 88 percent)
10.4s vs 10.6s for wprime 32m
310.9s vs 327.4s for wprime 1024m

Any near direct 2GHz Clovertown/dual socket F for comparison? All else I have is http://www.xtremesystems.org/forums/...3&postcount=34

[Donnie27]

If he means in 3Dmark, that doesn't contradict Coolaner's site's numbers. We have seen that K10 improves over K8 by about 7%. Per Core and Clockspeed.

So how does that compare to a 3.0 GHz Core2 if you consider two eight-cores? We have seen (in cinebench) that Barcelona's disadvatage per clock is less the more processors you count, which is hardly surprising given Hypertransport, integrated memory controller and NUMA. It has been the same way with K8 in both multiplprocessor sockets.

But the Intel Core2 eight-core is a socket 771 system, so no SLI.

So even if the per-core per-clock performance of Barcelona is still lower (which it is, face it), a benchmark involving both multi-processor code and graphics code, with Barcelona using SLI, will easily see K10 beat Core2.

That's a good thing Computers have always been about the whole system. If K10 isn't faster Processor vs. Processor but the whole system kicks Intel's a$$ then I'd want AMD's best. This was the same argument I made when I tried to tell nn that it's more to great PC than a speedy processor alone. There shouldn't be any *if* excuses for AMD or Intel. Processor is just one part of a system and the whole system should be judge as such!

Please don't forget, Intel does have V8 that was supposed to have bu CF and SLI. It seems that AMD-ATI and nVidia is fighting Intel on support of both techs.

[terrace215]

http://img.coolaler.com.tw/images/jj...mckdzozmt1.jpg

15 cycle L2 is vs. 12 cycle Penryn L2 (for 6 TIMES the L2 cache per core)

45 cycle L3.

[Donnie27]

So do M23 NFDDs (Noise and Flash Diversionary Devices aka flashbangs)

but eventually the truth is reveled

What, rubbed a sore spot on ya'?

Sure, so does the sun, flash bulbs and something like me punching in the eye LOL! Even after K10 hits the market, there will still be plenty of lies told. See the AMDZone after C2D? See [H]'s game review of C2D. Folks who lied on C2D reviews will NOT be trusted. They'll more than likely lie for K10 if it doesn't do so well.

No nn, I'll wait for Fugger, Victor, MovieMan and others here who I trust and not Ad-Whorrrresss who'll say anything for hits and Ad-Dollars.

[JumpingJack]

http://img.coolaler.com.tw/images/jj...mckdzozmt1.jpg

15 cycle L2 is vs. 12 cycle Penryn L2 (for 6 TIMES the L2 cache per core)

45 cycle L3.

The L2 latency has gone up or down for Barcey relative to Brisbane depending on how you look at it... I think CPUI latency measurements put Brisbane as high as 20 cylces, but Lostciruits did an analysis that put L2 latency on Brisbane at 14 cycles...

AMD added 2 more metal layers to Barcey, so equivalent to slightly better is expected. No wonder AMD/IBM are pushing for ultralow-K for 45 nm. Nice dig.

[JumpingJack]

http://forum.coolaler.com/showpost.p...&postcount=125
Harpertown 2GHz numbers for comparison!

23.3s vs 39.7s for Pi
20.48 vs 17.13 in "relative speed" for Fritz
2454 vs 1896 in single cpu cinebench
15334 vs 13295 in mutli cpu cinebench
6.25x vs 7.01 cinebench scaling (78 vs 88 percent)
10.4s vs 10.6s for wprime 32m
310.9s vs 327.4s for wprime 1024m

Any near direct 2GHz Clovertown/dual socket F for comparison? All else I have is http://www.xtremesystems.org/forums/...3&postcount=34

The thread scaling here is to be expected, this is also spanning over two sockets... so in a single socket comparision (4 core v 4 core), the scaling will difference between the two will be closer, but Barcey should be overall better at that..... (again, proceeding from the assumption that this is production silicon and no older ES samples).... the problem is that even with slightly better thread scaling performance, absolute clock for clock will not surpass....

The best hope is that there is something wrong with this particular sample/testing.

[Jimmer411]

I totally agree, people tend to particiapte more and enjoy the threads that dont involve flaming...

But the nature of this beast topic is hard not to show your true fanboy colors.

6 months from now there will be no arguing.

Oh trust me there will be argueing

Just wait till the upgrade threads, and someone wants to spec something out that isnt the fastest but is in their price range. Happened in the AXP and A64 days, as well as the C2D days.


Im still seeing the "Dont waste your money, buy a conroe!" replies in AMD threads, tho not nearly as much as 6 months ago. Same goes for AMD benching, some 4ghz conroe guy always comes along and bashes it.

[informal]

The best hope is that there is something wrong with this particular sample/testing.

I think it was a board problem(with possible earlier EVT).IPC on K10 won't be lower than on Merom parts,on the contrary.

[cpuz]

@cpuz: Has this any impact on the performance of the k10?

I don't know. And I'm not sure this result is not due to a BIOS or a driver issue.

[JumpingJack]

I think it was a board problem(with possible earlier EVT).IPC on K10 won't be lower than on Merom parts,on the contrary.

Hard to say at this point... there is no real data to compare, I mean unless we accept this data set as true.... which is subject to a deal of speculation. There is simply too much conflicting statements one way or the other to make such a conclusion.

[cpuz]

Does Revision 3.00 mean the same as HT3?
If so, where is that data from, i mean what chip shows it?

It's a Phenom ES.

[informal]

And to the one dude who said that HT3.0 was pushed to 2009,he is wrong.All Phenoms are HT3.0 ready,the boards (RD790) support HT3.0(16 bit link width,5.2GT/s).The chips CAN and will work(100&#37 on AM2 motherboards and the HT scales back to HT1.0(What this does to performance is not clear ,yet).

[cpuz]

http://img.coolaler.com.tw/images/jj...mckdzozmt1.jpg

15 cycle L2 is vs. 12 cycle Penryn L2 (for 6 TIMES the L2 cache per core)

45 cycle L3.

At least that shows that all cache levels are enabled.
(latency gets the size by measuring the access time, unlike cpuz that relies on the cpuid outputs)

[zakelwe]

All this talk of K10 needing MHz to really fly as such is really baloney.

No cpu in history has managed to achieve better than linear scaling. All have shown linear scaling low down and then drop off from this when other system bottlenecks start to appear. True, a good CPU will not drop off as much but it will never improve with speed. To say it scales better than linear you are sort of getting a perpetual motion machine scenario here, or more power out than in.


Gary is no mug but a K10 at 3GHz will be 50% or less quicker than one at 2Ghz, not 60% or 100% or 100000000%.

Maybe he is saying it has a high IPC and so any increase in speed gives a good rate of return on total cpu power due to the IPC level. That is a reasonable arguement. Non-linear scaling is definitely not.

As for the actual scores, you get this all the time with video card releases, somebody posts some high score and somebody posts some low score and it is only when the proper reviews come out with "relevant" cpu's to what you and I can buy, should you form your opinion.

We just have to be patient and see.

Regards

Andy

[JumpingJack]

All this talk of K10 needing MHz to really fly as such is really baloney.

No cpu in history has managed to achieve better than linear scaling. All have shown linear scaling low down and then drop off from this when other system bottlenecks start to appear. True, a good CPU will not drop off as much but it will never improve with speed. To say it scales better than linear you are sort of getting a perpetual motion machine scenario here, or more power out than in.

Makes one wonder doesn't it ... how they come up with this logic

[JumpingJack]

And to the one dude who said that HT3.0 was pushed to 2009,he is wrong.All Phenoms are HT3.0 ready,the boards (RD790) support HT3.0(16 bit link width,5.2GT/s).The chips CAN and will work(100&#37 on AM2 motherboards and the HT scales back to HT1.0(What this does to performance is not clear ,yet).

I am just going by what was in AMD's latest roadmap:
http://pc.watch.impress.co.jp/docs/2.../kaigai_02.pdf
(Pg 42)
http://www.amd.com/us-en/assets/cont...ck_Bergman.pdf
(Pg 19)

Two different presenters, same conlcusion.... Phenom X2/X4 will have HT1.0 specification. It is speculation as to why, if AMD is indeed accurate, perhaps they do not understand their own plans.... I could guess, purely guessing, that HT3.0 is having issues and holding back clocks, so they may drop to 1.0 specs in order to get higher CPU clockspeed.

Some seem to think budapest will transistion to HT3.0.... but, unless AMD botched their presentation, it appears that it has been pushed.

[informal]

I am just going by what was in AMD's latest roadmap:
http://pc.watch.impress.co.jp/docs/2.../kaigai_02.pdf
(Pg 42)
http://www.amd.com/us-en/assets/cont...ck_Bergman.pdf
(Pg 19)

Two different presenters, same conlcusion.... Phenom X2/X4 will have HT1.0 specification. It is speculation as to why, if AMD is indeed accurate, perhaps they do not understand their own plans.... I could guess, purely guessing, that HT3.0 is having issues and holding back clocks, so they may drop to 1.0 specs in order to get higher CPU clockspeed.

Some seem to think budapest will transistion to HT3.0.... but, unless AMD botched their presentation, it appears that it has been pushed.

It's botched,Phenom is HT3.0