The Phenom II X4 810 & X3 720: AMD Gets DDR3 But Doesn't Need It

[Cookie Monster]

Link

Great read.

 

[soonerproud]

Tom's Hardware PhII AM 3 Review


FiringSquad PhII AM 3 Review


PcPerspective PhII AM 3 Review


Hexus PhII AM 3 Review


Hot Hardware PhII AM 3 Review


Xibit Labs PhII AM 3 Review


ExtremeTech PhII AM 3 Review

 

[fusion238]

Anandtech gives a good review but doesn't provide a comprehensive sense of how easily the Phenom II X3 720 BE
can be overclocked to compete against dramatically higher priced Intel chips such as the new i7 series.

For a comparison, check out the new overclocked review and charts at overclockersclub.com

http://www.overclockersclub.co...omii_720_810_am3/9.htm

The overclocked Phenom II X3 720 BE either exceeds or is competitive with the Intel i7 920 in a number
of tests and benchmarks.

 

[OCGuy]

Originally posted by: fusion238

Anandtech gives a good review but doesn't provide a comprehensive sense of how easily the Phenom II X3 720 BE
can be overclocked to compete against dramatically higher priced Intel chips such as the new i7 series.

For a comparison, check out the new overclocked review and charts at overclockersclub.com

http://www.overclockersclub.co...omii_720_810_am3/9.htm

The overclocked Phenom II X3 720 BE either exceeds or is competitive with the Intel i7 920 in a number
of tests and benchmarks.

I think you are reading it wrong...

 

[beat mania]

Aren't these savaged parts coming a bit too soon? I was hoping that it takes AMD a few more months before they collect enough of these to start selling them.

 

[Nemesis 1]

Very nice . Well done review. As always. Nice job AMD. But I still believe AMd is in for a rough ride. Once Intel has to cut pricies because of performance. It will be a cut AMD won't beable to match and stay in business. Yep great for those who care only about $$$

Ya could have thrown in the C2D power consumption numbers. It really wasn't right not to include them . Since you compared X3 to C2D everwhere else. Its kinda remissive not to include those. After all people who buy lowend are also interested in power consumption. Because of Cost over time.

But The big picture sure isn't looking good. Best move intel could make is. Release the C2D 8600 for around $125. Than Bring the Q9550 down to $175. Than leave it there if AMD wants to price lower let them. Its still way cheaper for intel to produce Good 2 core chips. Where as AMD needs to use defective chips. Which is fine . Except for the cost of processing those wasted transitors. I was afraid AMD might get itself between a rock(Nehalem) and a hard place(Penryn). Intel can do alot with penryn pricing to keep the low end market. Nehalem has Server and Highend covered. The Middle is wide open . This is exactly were AMD didn't want to be. If AMD fanbois wouldn't have hyped PHI so much and made fools out of themselves. AMD would be in better position. But the Fanbois once again do more damage than good. Coming up short of hype is very disappointing and it hurts companies badly.

 

[perdomot]

Only complaint I can see from the review is that they didn't show how high the 720 can OC on default voltage by simply raising the multi. They did with the base clock but when they did the multi, they raised the vcore. Hope they provide this info as not all of us want to pump 1.55v through our cpus.

 

[soccerballtux]

Originally posted by: fusion238

Anandtech gives a good review but doesn't provide a comprehensive sense of how easily the Phenom II X3 720 BE
can be overclocked to compete against dramatically higher priced Intel chips such as the new i7 series.

For a comparison, check out the new overclocked review and charts at overclockersclub.com

http://www.overclockersclub.co...omii_720_810_am3/9.htm

The overclocked Phenom II X3 720 BE either exceeds or is competitive with the Intel i7 920 in a number
of tests and benchmarks.

Wow that's a terrible benchmark. At 10x7 they're graphics-limited.

 

[heyheybooboo]

I kinda wish Anand had used something other than the MSI DKA790GX.

In the Tech Report mobo review hard drive read/writes (might have been HD Tune) were similar in speed to other 790GXs but cpu utilization was more than 3 times that of the Gigabyte board.

It's one of those odd stats that makes you think it could be a performance limiter when pegging cpu utilization - even with a SSD - as I would think that would reduce read/write speeds ...

 

[Idontcare]

Originally posted by: beat mania
Aren't these savaged parts coming a bit too soon? I was hoping that it takes AMD a few more months before they collect enough of these to start selling them.

It depends on many things, obviously, but the primary inputs should be (1) yields on fully-functional 9xx dies, (2) total functional chip-outs (9xx, 8xx, 7xx), i.e. magnitude of volume that they are pushing out the fab, and (3) demand for the 9xx chips at current pricepoints.

If yields of 9xx are low but harvesting 8xx and 7xx results in large number of sellable chips then they have a business model to work with.

If wafer-volume is high and 9xx yields are high but even still the net number of 8xx and 7xx harvestable chips is still large enough to satiate a market segment's demand for them then they have a business model to work with.

If wafer-volume is high and 9xx yields are high but 9xx demand is low (and opteron demand is low, same die/mask-set from what I hear) then it makes sense to create intentionally crippled chips to satiate a lower ASP market segment for the time being rather than just keep dropping ASP's on the 9xx chips for market perception reasons.

Personally I doubt demand for 9xx chips is low, so my belief is that AMD's 45nm volume ramp is on a very steep slope and they simply have plenty of harvestable 8xx and 7xx chips exiting the fab already.

 

[Idontcare]

Originally posted by: beat mania
Aren't these savaged parts coming a bit too soon? I was hoping that it takes AMD a few more months before they collect enough of these to start selling them.

Personally I am rather disappointed in the AT review.

No data showing us the performance difference between running these AM3 chips in an AM3 setup versus and AM2+ setup. Sure they tell us there is no difference, but what's wrong with actually giving us the numbers in a nice graph to see for ourselves?

They also don't answer the question of whether the higher clocked uncore (and higher HT) of the AM3 chips reverts back to the lower/slower clocks of the existing AM2+ chips when the AM3 chips are put into an AM2+ board. (if they did tell us then someone please point me to it cause I looked, honestly)

And what's with no heads-up comparisons between benches of overclocked 7xx and 8xx with that of stock or OC'ed Intel chips? Weak. Very weak. You got the chips, you got the benches, you got the overclocks, you are a review site after all, and yet total fail when it comes to putting all that together in some graphs in the review?

I cannot rationalize why this decision (to exclude OC bench comparisons) was deemed to be in the best interest of the reader of the review. Surely it is one everone's minds when they read the review..."so how well does an OC'ed 7xx compare with an OC'ed Q8300?" etc. Guess they just ran out of time to do the tests and add in the results?

 

[Martimus]

Originally posted by: Idontcare

Originally posted by: beat mania
Aren't these savaged parts coming a bit too soon? I was hoping that it takes AMD a few more months before they collect enough of these to start selling them.

Personally I am rather disappointed in the AT review.

No data showing us the performance difference between running these AM3 chips in an AM3 setup versus and AM2+ setup. Sure they tell us there is no difference, but what's wrong with actually giving us the numbers in a nice graph to see for ourselves?

They also don't answer the question of whether the higher clocked uncore (and higher HT) of the AM3 chips reverts back to the lower/slower clocks of the existing AM2+ chips when the AM3 chips are put into an AM2+ board. (if they did tell us then someone please point me to it cause I looked, honestly)

And what's with no heads-up comparisons between benches of overclocked 7xx and 8xx with that of stock or OC'ed Intel chips? Weak. Very weak. You got the chips, you got the benches, you got the overclocks, you are a review site after all, and yet total fail when it comes to putting all that together in some graphs in the review?

I cannot rationalize why this decision (to exclude OC bench comparisons) was deemed to be in the best interest of the reader of the review. Surely it is one everone's minds when they read the review..."so how well does an OC'ed 7xx compare with an OC'ed Q8300?" etc. Guess they just ran out of time to do the tests and add in the results?

I wholeheartedly agree with you on all of your points. I was also rather disappointed with the review. I expected them to improve on their mistakes they made in the first review, but I was disappointed instead.

 

[Idontcare]

Originally posted by: soonerproud
Tom's Hardware PhII AM 3 Review

Lol toms actually adds some value to us enthusiasts knowledge-base: (I love them for doing this test BTW)

Modding And Overclocking?Doable?

Our first thought after hearing that AMD wouldn't be launching a Phenom II X4 940/920 equivalent for the AM3 platform was, "what about all of the enthusiasts who've been eying those high-end models and still want to experiment with DDR3 memory?"

We knew we had a Phenom II X4 940 in the house that didn't scale very well and would likely be replaced soon by another chip that would ideally have more headroom built into it, so we decided to try "creating" a 938-pin AM3 chip out of our 940-pin AM2+ sample. After all, the silicon under its proverbial hood was the same?the only difference was its interface.

So, using a mechanical pencil, we bent the two offending pins back and forth until they snapped off. The chip now fit into our AM3 test platform, albeit not flush due to the metal nubs where each pin broke. Unfortunately, the modded processor would not POST at all, forcing us to conclude that the task wouldn't be as easy as popping off pins. It's truly a shame that enthusiasts can't get access to AMD's AM2 pinout, which would describe the exact role of each pin rather than force us to guess. The most recent tech doc publicly-available relates to the old-school Socket 940 interface.

Wondering if we'd just nuked a perfectly good CPU, we moved the Phenom II X4 940 back to its AM2+ board. Lo and behold, it still ran fine, without any immediately apparent issues.

Thank you tomshardware, not exactly a question I would have asked but now that we know the answer to the question (can you snap pins of an AM2+ PhII and magically turn it into an AM3 chip?) we are all the more educated for it :laugh:

 

[Idontcare]

Originally posted by: soonerproud
PcPerspective PhII AM 3 Review

OK cool, finally, confirmation that putting an AM3 processor into an AM2+ mobo will NOT cause the uncore to downclock from 2GHz stock to 1.8GHz or some such.

The only other major change is that the north bridge portion of the chip (the ?un-core?) now runs at 2 GHz vs. the 1.8 GHz that the AM2+ compatible X4 940 runs at. This should speed up memory performance, as well as allow faster access to the L3 cache (plus slightly lower latencies

Checkout the CPU-z screenshot, AM2+ socket and 2GHz HT.

 

[taltamir]

I distinctly recall saying that I predict that DDR3 will do nothing for phenom2, i distinctly recall many people vehemently disagreeing... I guess they were wrong.
It is a shame though, I WANTED p2 to get faster and be more competative, all intel has to do is slash prices on their LAST gen parts to bury the p2.

 

[Martimus]

Originally posted by: taltamir
I distinctly recall saying that I predict that DDR3 will do nothing for phenom2, i distinctly recall many people vehemently disagreeing... I guess they were wrong.
It is a shame though, I WANTED p2 to get faster and be more competative, all intel has to do is slash prices on their LAST gen parts to bury the p2.

Which review did you see the comparison between DDR2 Ph2 performance and DDR3 Ph2 performance? I didn't see that in the Anandtech article (even though the title would have suggested it). I would definitely be interested in reading the performance difference.

 

[Idontcare]

Originally posted by: taltamir
I distinctly recall saying that I predict that DDR3 will do nothing for phenom2, i distinctly recall many people vehemently disagreeing... I guess they were wrong.
It is a shame though, I WANTED p2 to get faster and be more competative, all intel has to do is slash prices on their LAST gen parts to bury the p2.

You appear to be vindicated. I don't recall people vehemently disagreeing with you, but if it was me then I still disagree with you despite the data to the contrary.

Something just compels me to never agree with a taltamir, I blame my mother for her storytime nursery rhymes about the taltamir's that pillaged my ancestors village...or was that the visigoths?

Anyways, all ancient history so who cares, except me. No way DDR3 is only as good as DDR2 on AMD platform taltamir, you have been vehemently disagreed with.

Originally posted by: soonerproud
Hot Hardware PhII AM 3 Review

Finally! Hot Hardware ran the heads-up tests we wanted to see: X4 810 on DDR3 vs. X4 810 on DDR2

I like the fact they included the X4 810 in these DDR2 vs DDR3 tests as the 810 is the L3$ deprived quad-core meaning it is more likely to highlight any benefits of a faster DDR3 sub-system relative to that of a DDR2 sub-system versus say the benefits that might be shown by a larger L3$ X4 945.

Of course the AM3 boards are not fully tweaked/optimized as I am sure the PhII's DDR3 IMC is not as well, but at time-zero we aren't seeing a 5% performance benefit that could/might grow to be 7% or 8% with maturity...we really are basically seeing next-to-zero benefit at this time which to me means someday we may see a 3-4% benefit of AM3 over AM2+ once the AM3 situation matures. (just my opinion obviously, would love to be surprised and find out I had it all wrong)

(taltamir if you are still reading this my statement above does not extend to you, for you I still vehemently disagree that DDR3 fails to provide a benefit, I will never concede this point, ever. Except in conversations with any and all posters who are not named taltamir )

Originally posted by: Martimus

Originally posted by: taltamir
I distinctly recall saying that I predict that DDR3 will do nothing for phenom2, i distinctly recall many people vehemently disagreeing... I guess they were wrong.
It is a shame though, I WANTED p2 to get faster and be more competative, all intel has to do is slash prices on their LAST gen parts to bury the p2.

Which review did you see the comparison between DDR2 Ph2 performance and DDR3 Ph2 performance? I didn't see that in the Anandtech article (even though the title would have suggested it). I would definitely be interested in reading the performance difference.

The Hot Hardware explicitly shows the results, no difference with DDR3. There is some (<1%), it is not zero, but it is as Anand said "no difference". At least the data are there.

Pssst - don't let talta know about it though, I'm trying to give him a complex just for the fun of it. Ix-nay on the ack-lay of the erformance-pay...eh

 

[Martimus]

Originally posted by: Idontcare

Originally posted by: Martimus

Originally posted by: taltamir
I distinctly recall saying that I predict that DDR3 will do nothing for phenom2, i distinctly recall many people vehemently disagreeing... I guess they were wrong.
It is a shame though, I WANTED p2 to get faster and be more competative, all intel has to do is slash prices on their LAST gen parts to bury the p2.

Which review did you see the comparison between DDR2 Ph2 performance and DDR3 Ph2 performance? I didn't see that in the Anandtech article (even though the title would have suggested it). I would definitely be interested in reading the performance difference.

The Hot Hardware explicitly shows the results, no difference with DDR3. There is some (<1%), it is not zero, but it is as Anand said "no difference". At least the data are there.

Pssst - don't let talta know about it though, I'm trying to give him a complex just for the fun of it. Ix-nay on the ack-lay of the erformance-pay...eh

Thanks! That was exactly what I was looking for.

 

[heyheybooboo]

I think it's essentially the same as it was with DDR --> DDR2.

Because of reduced timings step up one speed grade for similar performance ... so DDR2-1066 with 5-5-5-15 timings is generally equal to DDR3-1333 with 7-7-7-20 timings.

Isn't that how that worked with DDR --> DDR2 ?? (Or have I let a brain stinky?)

I imagine the 'performance' debate will be settled soon enough - both Asus and Gigabyte have qualified their AMDs with DDR3 1600MHz+

 

[cusideabelincoln]

In most of AT's benchmarks the 810 performs the same, and sometimes outperforming, the 910. I wonder if they got some results mixed up.

 

[Idontcare]

Originally posted by: heyheybooboo
I think it's essentially the same as it was with DDR --> DDR2.

Because of reduced timings step up one speed grade for similar performance ... so DDR2-1066 with 5-5-5-15 timings is generally equal to DDR3-1333 with 7-7-7-20 timings.

Isn't that how that worked with DDR --> DDR2 ?? (Or have I let a brain stinky?)

I imagine the 'performance' debate will be settled soon enough - both Asus and Gigabyte have qualified their AMDs with DDR3 1600MHz+

Yeah, pretty much the same thing as what happened with DDR2->DDR3 for Intel Core CPU's.

I suspect the same is true of Nehalem as well, that DDR3 adds little value to the platform's performance above and beyond what and equally tuned DDR2 IMC would have provided, but we'll never know whether this is/was the case since we only have DDR3 Nehalem platforms.

Only speaking of desktop platforms here, not speaking about the benefits of higher bandwidth provided to multi-socket systems in enterprise environments. One would hope DDR3 bandwidth adds value in that environment. (as opposed to power savings, that part of DDR3 is obviously value-add in enterprise situations) If it doesn't then there is no chance of it ever adding value to the desktop environment where bandwidth needs are drastically lower.

 

[cusideabelincoln]

Well DDR3 prices will continue to drop, so its value compared to DDR2 should equal. About this time last year 4GB of DDR2 RAM cost the same as 4GB of DDR3 does now.

 

[Denithor]

Xbitlabs: DDR2 vs DDR3

3-4% difference on average between DDR2-800 CAS5 and DDR3-1600 CAS7.

DDR2-1066 will remain the best performance value for the moment.

Really makes me wonder why Intel chose to ram DDR3 down our throats with i7 (pun fully intended). With the adjustable memory multiplier combined with triple-channel bandwidth I just cannot understand their logic behind the decision.

 

[kknd1967]

THG review made my day !
they dare to bend the pins, trying to make AM2+ chip compatible with AM3 socket, thinking others (particularly technical staff in AMD) do not know how to break 2 pins.
Then they published 3DMark score with a 20% difference in GPU score, and start making wishful performance claims on i7 that are pretty well known.

quite a funny thing. All the hardware enthusiasts can finally have a min of good laugh on Monday morning even if this is still a gloomy day for IT industry.

Originally posted by: soonerproud
Tom's Hardware PhII AM 3 Review


FiringSquad PhII AM 3 Review


PcPerspective PhII AM 3 Review


Hexus PhII AM 3 Review


Hot Hardware PhII AM 3 Review

 

[Idontcare]

Originally posted by: Denithor
Really makes me wonder why Intel chose to ram DDR3 down our throats with i7 (pun fully intended). With the adjustable memory multiplier combined with triple-channel bandwidth I just cannot understand their logic behind the decision.

Given the length of the design cycle (~4yrs) the DDR3 thing is basically the result of Intel decision makers assuming the odds were high that DDR3 would be mainstream by the time Nehalem hit the stores. They missed it by a year, but at the pace things change in this industry I'd say it was pretty a darn decent dart throw.

AMD had it a little easier because they already had DDR2 IMC in their hands and extending to DDR3 compatibility was less of a challenge. So less guessing was needed on their behalf when it came to 45nm Phenom II architecture that got underway 4 yrs ago. Integrate both and let the markets decide which part of the hardware would be used.

Intel could have gone the safe route and engineered in both DDR2 and DDR3 IMC's but for whatever reason based on data they had and projections they made they felt electing for DDR3 only was an acceptably safe strategy.

If we assume the real target market for Nehalem was the enterprise/server markets (multi-socket) then we can see why the designers went for the real high-performance ram solution. If you are dropping $2k-$3k per Nehalem Xeon cpu on a four-socket rig then what do you care if the highest bandwidth DDR3 dimms are going to set you back $25/GB versus $12GB for lower bandwidth DDR2?

What really impresses me is just how much better/superior Intel's first modern stab at IMC turned out to be relative to AMD's considering AMD has the benefit of much deeper experience with modern IMC's.

Think about it, Intel came out of the IMC gate with triple channel DDR3 and crazy low-latency and supporting high bandwidth modes in 3-dimms per channel and three channels. Compare this to the current state-of-the-art implementation of AMD's IMC where you get DDR2-800 speeds if you populate four dimms in dual-channel mode and the top-speed for DDR3 is 1333 but that is only validated for 1 dimm per channel (versus 3-dimms per channel for Intels).

For Intel to come from so far behind on IMC technology and jump so far ahead of AMD from the get go really makes me have high expectations for what they are going to do to the GPU world when they release Larrabee.