• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

Phenom/Barcelona Reviews/Previews and Pre-release Discussion

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.
Assuming the preview scaling is correct (I believe the jump from K8->K10 will be more around 20% on average on desktop once a proper platform is setup),
That's doubtful, while the K10 is penalized by the server board, so is the simulated K8.

then looking here and calculating for a Phenom, it would actually put a K10 cpu at a virtual tie with a Core2 cpu in Oblivion and about a 4% advantage in HL2 episode 1.

From here:

http://anandtech.com/cpuchipsets/intel/showdoc.aspx?i=3038&p=10

the scaling going from the Q6600 to the QX6800 is better than the scaling of Barcelona from 2GHz to 2.5GHz in terms of % increase in performance/% increase in clockspeed. Also, for the same set of applications used to calculate the 15% advantage of the K10 vs the K8, the QX6850 is 30% faster than the FX-74, and that's with Oblivion and HL2 running at 1600x1200. Based on this, plus the Techreport review showing the 2.33GHz Clovertown winning as many benches as it loses with the 2.5GHz Opteron, I fully expect Phenom to be slower clock for clock than Conroe/Kentsfield.
 
I feel that this is quite an ambigous start for "the next big thing". Okay, an ES isnt the product we will get, and there may well be tweaks. But can I see this, or the phenom equivilant in my rig, having paid money for it. No. Unless phenom really stomps C2D, which looks unlikley, Intel can hold with what they have and beat AMD down with its next generation of products.
 
i would like to take a moment and laugh due to the fact that the intel guys are running scared so they have to sit there and complain and say it aint nothing. always got a complainer, 10 times out of 10 he owns a C2D.

AMD 4 life

yeah thats right i am a amd fan boy, u got a problem with that! then byte my flops

(made no sense i know :D )
 
i would like to take a moment and laugh due to the fact that the intel guys are running scared so they have to sit there and complain and say it aint nothing. always got a complainer, 10 times out of 10 he owns a C2D.

AMD 4 life

yeah thats right i am a amd fan boy, u got a problem with that! then byte my flops

(made no sense i know :D )

I think you are still chewing on your 30% improvement over Conroe FUD and it is hard to swallow the truth of only 10% - 15% improvement over K8. :beer:
 
wasent looking at that, but to me performance i really dont care about, a quad core for cheap is what i do care about. so what yeah intel (the man) has a fast chip who cares, (well obviously u do :D )
fight the man get AMD !!!

now let us gloat and u just run back to the intel section ^_^
 
Actually, I think this short paragraph in Anandtech's summary on their Barcelona article sums up my feeling so far:

Anandtech article said:
The future looks very interesting with the 45nm Xeon Harpertown and a 2.5GHz AMD quad-core in the next quarter. AMD hasn't clearly hit a homerun this time, but at least they're playing in the same ballpark.

Hopefully the upcoming Phenom procs can also be in the same ballpark as the Intel Quads. I love to see both companies with product that is competitive in performance from both companies because it keeps both companies advancing their designs at a good pace and it also helps to keep the prices lower for us buyers. :D
 
i think when we see phenom running at a higher clock speed with higher speed ram they will show to be faster than c2d than it appears from the server info. if i understand correctly higher clocked parts will also have a a higher clocked mem controller (on the new boards with the different power planes). seems like i read that the L3 cache runs at the same speed as the mem controller. if that's true then the higher speeds should reduce L3 latency. this combined with faster desktop ram (1066?) should make for much higher memory bandwidth. maybe this is what the whole supercharger thing was referring to. i think phenom at 3.0 with some high speed ram will be a monster.

i think this chip has got more in it than the server #'s will show. this is coming form someone who is running an e6600 right now, so these aren't the words of a fanboy.

just one man's opinion. i'm certainly no expert on these things.
 
Last edited:
That's doubtful, while the K10 is penalized by the server board, so is the simulated K8.

It is true that both would be bottlenecked, but that doesn't matter. If your graphics system is your bottleneck, it will mask cpu performance gains if one or both of the systems is bottlenecked.

From here:

http://anandtech.com/cpuchipsets/intel/showdoc.aspx?i=3038&p=10

the scaling going from the Q6600 to the QX6800 is better than the scaling of Barcelona from 2GHz to 2.5GHz in terms of % increase in performance/% increase in clockspeed.

According to my calculations I have the K10 system with a 70% efficiency comparing clock speed to performance gains and the Woodcrest (sorry if this is wrong, not sure of the code names) example you gave with a 62.4% efficiency. Either way though it is apples to oranges as my example was scaling in games whereas this is an entirely different type of benchmark.

Also, for the same set of applications used to calculate the 15% advantage of the K10 vs the K8, the QX6850 is 30% faster than the FX-74, and that's with Oblivion and HL2 running at 1600x1200.

Again, the 15% is an approximation based off an improper platform for those benchmarks. Could that end up the final result? Sure, but more than likely the disparity between K8 and K10 will grow given the proper platform. Also, the K10 improvements over K8 in those two games was shown to be greater than the average 15% in anandtechs preview. As far as the 30% goes, the FX-74 in those reviews was again hindered by the platform in the gaming applications. All you need to do is look at the A64 6000+ which is clocked at the same speed (3ghz) but on a proper desktop platform and see how a K8 really performs in that game. If you take the result of the 6000+ and calculate the Phenom preview % approximations, then you can see that atleast in those two games, a Phenom platform will most likely exceed a comprable Core2 system if only by small gains. That's why it will most likely come down to price at a given performance level.

Based on this, plus the Techreport review showing the 2.33GHz Clovertown winning as many benches as it loses with the 2.5GHz Opteron, I fully expect Phenom to be slower clock for clock than Conroe/Kentsfield.

I haven't gone through and tallied the wins/losses in the techreport review, but obviously the suite of benches anandtech has tested on as well as the german site show different results. Personally, I am not fully satisfied with all the results we have thus far and will wait for some more reviews to show up before drawing a full conclusion.
 
Last edited:
Again, the 15% is an approximation based off an improper platform for those benchmarks. Could that end up the final result?
The improper platform affects both the K10 and K8. The K10 will be faster on a desktop board, but so would a K8.

Sure, but more than likely the disparity between K8 and K10 will grow given the proper platform. As far as the 30% goes, the FX-74 in those reviews was again hindered by the platform in the gaming applications. All you need to do is look at the A64 6000+ which is clocked at the same speed (3ghz) but on a proper desktop platform and see how a K8 really performs in that game.
The FX-74 runs with fast, unregistered memory but it is hindered by the second socket hit on latency. But this also affects the simulated K8 vs the single quad K10. A lot of the "performance advantage" shown by the K10 is in fact caused by the second socket.

approximations, then you can see that atleast in those two games, a Phenom platform will most likely exceed a comprable Core2 system if only by small gains. That's why it will most likely come down to price at a given performance level.
Using Anandtech numbers, a 3GHz QX6850 beats the 2 socket 3GHz FX-74, both running on enthusiast boards with fast unregistered memory, by 34% in Oblivion and 25% in Half Life 2 at 1600 x 1200 with a single 8800 GTX. By comparison a 2GHz K10 beats a 2 socket 2GHz Opteron system, both running with server type memory, only by 16% in Oblivion and 21% in Half Life 2 at 1024 x 768 with a single 8800 GTX.
 
That's why it will most likely come down to price at a given performance level.

Yet again as Anand said big 65nm native quad vs stick together duals on 45nm, which one is cheaper to make...

The problem is I do not see the killing, "Intel being pushed out of the market" as savage predicted, and these are only clock for clock comparisons.
If the performance doesn't improve gonna be all the same again like last 18 months.
 
The improper platform affects both the K10 and K8. The K10 will be faster on a desktop board, but so would a K8.

Again, the conecept of bottlenecks. True both systems would improve, but the more powerful cpu will benefit greater from an unhindered system. Even anandtech prefaced their results with this same guidance,
Anandtech Phenom Preview said:
Keep this in mind as you're looking at these results, at best all we're offering is an idea of, at a minimum, how much faster Phenom will be over an identically clocked Athlon 64 X2. As Phenom is a more data hungry CPU than its predecessor, it will rely more on having a faster memory subsystem so the performance improvement could be even greater when we measure it on the desktop.

*obviously my emphasis


The FX-74 runs with fast, unregistered memory but it is hindered by the second socket hit on latency. But this also affects the simulated K8 vs the single quad K10. A lot of the "performance advantage" shown by the K10 is in fact caused by the second socket.[/q]


Using Anandtech numbers, a 3GHz QX6850 beats the 2 socket 3GHz FX-74, both running on enthusiast boards with fast unregistered memory, by 34% in Oblivion and 25% in Half Life 2 at 1600 x 1200 with a single 8800 GTX. By comparison a 2GHz K10 beats a 2 socket 2GHz Opteron system, both running with server type memory, only by 16% in Oblivion and 21% in Half Life 2 at 1024 x 768 with a single 8800 GTX.

HOWEVER, if you use the 6000+ results which show an a64 on a proper gaming platform unhindered by the poor graphics performance, which is what we should expect to see from a proper Phenom platform, the difference between the highest Core2 and A64 scores are only 14.5% and 16.7% respectfully, which is obviously less than the improvement from K8 to K10.

Also, if you notice, the K10 was also using 2 sockets, so that point is moot.
 
I have to say I've been very optimistic about this architecture, but I have to say I'm very impressed so far! WOW! I can't wait to see how fast those Phenom's will be!
 
Add +5% for penryn and +30% clockspeed and you understand anand's conclusion. ;)

That is why I added the stipulation in my comments that AMD needs to get the yields up in order to compete in the top end, all the rest of my comments have merely been about clock to clock performance.

Yet again as Anand said big 65nm native quad vs stick together duals on 45nm, which one is cheaper to make...

Although I would argue that intel has pretty much always been ahead on manufacturing, this is true. However, there is a lot more to take into account when talking financials. . . just for a small example, AMD is a much smaller company with less overhead. I'm certainly not going to go into a financial discussion as this isn't the forum nor do I have any desire to go that deep, but basically with Barcelona, AMD just needs to get back into the black until the next refresh which Barcelona very well may allow them to do if they play it right. Here's to hoping :)

The problem is I do not see the killing, "Intel being pushed out of the market" as savage predicted, and these are only clock for clock comparisons.
If the performance doesn't improve gonna be all the same again like last 18 months.

Intel is far too big a company to be "pushed out of the market". Even when they were behind with the Pentium4, they were still pulling insane amounts of revenue. Intel's around to stay, that's all there is to it. With this move AMD needs to become an attractive alternative, and initial results show they have done that. Absolute performance lead is not really needed as not much revenue is gained from that segment, especially in servers where a performance/power balance is usually wanted and AMD may compete very well in that area.
 
CPU Type Linpack score
Barcelona 2347 (1.9Ghz) 37.5 Gflop/s
Intel Xeon 5150(2.6Ghz) 35.3 Gflop/s

http://www.techwarelabs.com/reviews/processors/barcelona/index_3.shtml

Here is the score I'm interested in (notice the 700MHz difference). I run F@H 24/7 on my system, play a light amount of games and do some 3D design. Flops is where it is. And this thing is only 1.9GHz!!! There is plenty of room to grow.
I saw that as well. Outside of folding, I don't think I'd ever use 4 cores, unless of course software I use is designed for it. If the desktop processors can fold with a better performance/watt than my current AM2, and they do it at an affordable price, that's all I'm concerned with. Hell, for me, a 2.1ghz dual core K8 is just fine, I think I'll just sit back and wait for the low power X4's to hit retailers.
 
The poor thing, it's starved of memory, DDR2-667-5-6-5-12 just can't cut it... for a single core, 4 is out of the question.

I bet the thing will scale very well if they reduce the L3's latency and give the thing some data to process.

I have one to put in my workstation, the 1.9GHz EE, but I think I'll wait for the next revision to clean things up.
 
CPU Type Linpack score
Barcelona 2347 (1.9Ghz) 37.5 Gflop/s
Intel Xeon 5150(2.6Ghz) 35.3 Gflop/s

http://www.techwarelabs.com/reviews/processors/barcelona/index_3.shtml

Here is the score I'm interested in (notice the 700MHz difference). I run F@H 24/7 on my system, play a light amount of games and do some 3D design. Flops is where it is. And this thing is only 1.9GHz!!! There is plenty of room to grow.

This is the kind of shoddy comparisons that makes me laugh. Comparing an 8 core Barcelona system (look at the screenshot from your link with taskmanager clearly showing 8 cores) vs a dual core 5150, that they did not even provide a taskmanager screenshot for so it was probably running with just one dual core. So an 8 core vs 2 core hahahahahah, get some real credible benches.
 
The HKEPC Penryn preview popped into my mind and it shows VERY surprising results...

HL2 1024X768 HQ
E6550 (2.33GHz/4MB L2/1333MHz FSB) 132FPS
Intel "Wolfdate" ES Sample(2.33GHz/6MB L2/1333MHz FSB) 181FPS

+31.12%

Not bad from a die shrink...
 
I like how the editorial on the front page of this site from today supports my view that this is just a stepping stone and is nowhere near dethroning Intel.

No offense to you, but the front page is not a source to be referenced. As you said it is an editorial, meaning one man's opinion, that's all. Just because he has a website does not validate his opinion. No offense to Ed, I respect him, but anyone could make a website and quote themselves whenever they wanted to prove a point. As far as dethroning intel, I never mentioned this nor do I expect AMD to overthrow the "empire" anytime soon, even if the next 2 product cycles are a landslide performance win for AMD (won't happen either).

The high voltage he points out to reach 2.5Ghz is an eye opener. Also the die size is more than the clovertown and the Penryn based tigerton is to be even smaller, so no price per chip advantage. I see no competition here.

As far as the voltage, we have no confirmation that cpu-z is even reading the voltage right, that is probably one of the most common errors related to cpu-z and given these tested results, I'd have to say that either the leakage problem with anandtech's cpu was fixed very quickly or cpu-z was wrong,

http://techreport.com/articles.x/13176/10

With price per chip, there is a lot more than wafer costs to calculate when calculating price per cpu. As I said in another thread, AMD is a much smaller company with less overhead, just for an example. Without going into a deep financial discussion it basically boils down to AMD needing to get back into the black, which Barcelona very well may allow them to do.
 
Last edited:
Back