Barcelona Benchmarks according to AMD pricing.
http://freshscoop.com/modules....&mode=&order=0&thold=0
http://freshscoop.com/modules....&mode=&order=0&thold=0
Barcelona Benchmarks according to AMD pricing.
- Thread starter Mitch101
- Start date
- Status
Yup it is. More of a DIY website than anything else but we have 2 other tech's possibly looking for more.
We wont be doing motherboards, cpu, ram, video card reviews leave that for anantech and the other sites out there. Once and a while I might talk about it but its not really what we want to do. We think there are plenty of sites that do that stuff well there is no need for another.
Were going to focus on other items that make computers more interesting. Like we review training material, build our own projectors, Build arcade cabinets, redesign the computer desk, X-10 devices and programming X-10. Were looking at the PC's surrounding and looking at resolving common problems that are difficult to find answers to.
We may touch on repairing HDTV's on the cheap as most of the time its a $10.00 IC that goes bad but for some reason everyone thinks they can take a screwdriver to a few knobs and fix a bad chip. That kind of different.
We wont be doing motherboards, cpu, ram, video card reviews leave that for anantech and the other sites out there. Once and a while I might talk about it but its not really what we want to do. We think there are plenty of sites that do that stuff well there is no need for another.
Were going to focus on other items that make computers more interesting. Like we review training material, build our own projectors, Build arcade cabinets, redesign the computer desk, X-10 devices and programming X-10. Were looking at the PC's surrounding and looking at resolving common problems that are difficult to find answers to.
We may touch on repairing HDTV's on the cheap as most of the time its a $10.00 IC that goes bad but for some reason everyone thinks they can take a screwdriver to a few knobs and fix a bad chip. That kind of different.
confusing as heck to read....what the heck does it have to do with july 22nd intel prices? huh? this is talking about chips not to be released until near 4th quarter.....
bottom line on AMD barcelona......
1) too low of clock speed...so they damn well better have a much better IPC then intel
2) cache per core is TOOOOOOOOO LOOOOOOOOOOW and the overall pool is pathetic.....
I am sensing a big disappointment on my end.....
Here I have the money waiting for one of these too....hoping to get 2.5ghz bins out????? menas not much of an OC as well.....I can get 533mhz out of my INtel 65nm quad while still using stock vcore....with a little juice another 300mhz....problem is the boards cant handle much more due to power issues and cooling rears its ugly head at 3.2ghz.....
I am thinking at best a lateral move for me....so I will stay and wait for the 45nm quads and see if I can get 3.4-3.5ghz with modest cooling needs and with my water cooling 3.7+....we will see, but right now I am not seeing much of anything from AMD but a confusing naming system
bottom line on AMD barcelona......
1) too low of clock speed...so they damn well better have a much better IPC then intel
2) cache per core is TOOOOOOOOO LOOOOOOOOOOW and the overall pool is pathetic.....
I am sensing a big disappointment on my end.....
Here I have the money waiting for one of these too....hoping to get 2.5ghz bins out????? menas not much of an OC as well.....I can get 533mhz out of my INtel 65nm quad while still using stock vcore....with a little juice another 300mhz....problem is the boards cant handle much more due to power issues and cooling rears its ugly head at 3.2ghz.....
I am thinking at best a lateral move for me....so I will stay and wait for the 45nm quads and see if I can get 3.4-3.5ghz with modest cooling needs and with my water cooling 3.7+....we will see, but right now I am not seeing much of anything from AMD but a confusing naming system
We could have used the current prices for Intel but since the barcelona chip isnt coming for several months we have to use the Intel price drop to draw our conclusions. We assume AMD would not price the chips with existing prices as everyone knows about the July 22nd price drop.
I dont think anyone can go wrong with either an Intel or AMD basically people buy within their price range or buy for the performance of something they wish to do. Since both chips are comparably priced its sort of moot. A lot of us look for the best overclocker in our price range. Right now I feel AMD has the bottom end covered and a bit of the middle but when the price drops occur AMD has to follow however Core 2 is a pretty good overclocker. Intel has great manufacturing which is tough for AMD. In the end we can only hope AMD remains competitive and is profitable a world without AMD would be very pricey. As much as people love or hate AMD or Intel for whatever reason we cant afford to lose AMD. Intel would love it but your wallet wont.
We should also remember its not just the CPU price that makes the machine you generally need a coresponding motherboard and ram for that system. If the CPU's are priced the same but it costs you $50.00 more for one motherboard and $100 more because your an early adopter of DDR3 then there is a cost difference between them. I certainly feel too many people stress the cost of thier CPU instead of the cost of the CPU, MOBO, RAM to determine true cost.
I dont think anyone can go wrong with either an Intel or AMD basically people buy within their price range or buy for the performance of something they wish to do. Since both chips are comparably priced its sort of moot. A lot of us look for the best overclocker in our price range. Right now I feel AMD has the bottom end covered and a bit of the middle but when the price drops occur AMD has to follow however Core 2 is a pretty good overclocker. Intel has great manufacturing which is tough for AMD. In the end we can only hope AMD remains competitive and is profitable a world without AMD would be very pricey. As much as people love or hate AMD or Intel for whatever reason we cant afford to lose AMD. Intel would love it but your wallet wont.
We should also remember its not just the CPU price that makes the machine you generally need a coresponding motherboard and ram for that system. If the CPU's are priced the same but it costs you $50.00 more for one motherboard and $100 more because your an early adopter of DDR3 then there is a cost difference between them. I certainly feel too many people stress the cost of thier CPU instead of the cost of the CPU, MOBO, RAM to determine true cost.
:roll: Uh chief, ...AMD does not need the crutch of more cache because their design is much more advanced than Intel which relies heavily on it.
More cache = crutch to make up more performance due to the use of antiquated designs.
<div class="FTQUOTE"><begin quote>Originally posted by: WAWA
<div class="FTQUOTE"><begin quote>Originally posted by: Duvie
bottom line on AMD barcelona......
2) cache per core is TOOOOOOOOO LOOOOOOOOOOW and the overall pool is pathetic.....</end></div>
:roll: Uh chief, ...AMD does not need the crutch of more cache because their design is much more advanced than Intel which relies heavily on it.
More cache = crutch to make up more performance due to the use of antiquated designs.
</end></div>
really? Give me some documented proof or benches that show this...on the "more advanced then intel" claim....I haven't seen any proof or benches to show this.
and while you are pulling that out of your arse....
show me where cache is considered a cructh and part of an antiquated design....
the reason for cache is ability to use faster lower latency memory to store bits of applications being ran to run faster and avoid the bootleneck of physical ram capabilities. Now design of how cores use cache and whether it is shared or not is more of a design issue and reason for debate on which is better...
fact is...I have done research and have shown some application would like bigger cache pools and will run faster with such...That is about size and availability and not architecture. Those apps would not be able to load fully up into the cache on the AMD models and thus would run slower by using physical ram or having to repeatedly clear and reload parts into the cache...or use higher latency physical ram.
I think you need to add a bit more to your statement....
<div class="FTQUOTE"><begin quote>Originally posted by: Duvie
bottom line on AMD barcelona......
2) cache per core is TOOOOOOOOO LOOOOOOOOOOW and the overall pool is pathetic.....</end></div>
:roll: Uh chief, ...AMD does not need the crutch of more cache because their design is much more advanced than Intel which relies heavily on it.
More cache = crutch to make up more performance due to the use of antiquated designs.
</end></div>
really? Give me some documented proof or benches that show this...on the "more advanced then intel" claim....I haven't seen any proof or benches to show this.
and while you are pulling that out of your arse....
show me where cache is considered a cructh and part of an antiquated design....
the reason for cache is ability to use faster lower latency memory to store bits of applications being ran to run faster and avoid the bootleneck of physical ram capabilities. Now design of how cores use cache and whether it is shared or not is more of a design issue and reason for debate on which is better...
fact is...I have done research and have shown some application would like bigger cache pools and will run faster with such...That is about size and availability and not architecture. Those apps would not be able to load fully up into the cache on the AMD models and thus would run slower by using physical ram or having to repeatedly clear and reload parts into the cache...or use higher latency physical ram.
I think you need to add a bit more to your statement....
zsdersw
Lifer
- Oct 29, 2003
- 10,505
- 2
- 0
Oh yes.. because the jump in L2 cache from 2 to 4MB *really* made a huge difference, across the board, for Core 2. :roll:
Why do you suppose the Pentium Exxxx processors perform so relatively well with their 1MB of L2 cache? Because L2 cache isn't all there is to tell in the CMA performance story.
Be cocky all you want.. but in the interests of your own reputation, it's best to know what you're talking about beforehand.
I think he's talking about the ODMC that A64s sport, which is more advanced, and provides high memory performance.
perhaps but even with AMD On Die Memory Controller (ODMC)...it is no substitute to the higher density and lower latency SRAM.....not even a comparison....The ODMC for intel would help in situation where the cpu needs to access physical memory without the latency hits, but this is not the same thing.
He still was talking out of his arse...like that commercial for ESPN....
the amazing thing is how Intel has been able to place as much cache on the chip they do and be able to keep their die size relatively compacted....
He still was talking out of his arse...like that commercial for ESPN....
the amazing thing is how Intel has been able to place as much cache on the chip they do and be able to keep their die size relatively compacted....
That is quite a crude prediction on Barcelona performance based purely on pricing. Whats to say Intel won't cut prices in the coming months (like they are doing with C2D) which throws your argument right out the window?
Finally, to suggest that Barcelona scales so well that at 1.9GHz it matches a 2GHz Xeon while at 2.4GHz it matches a 3GHz Xeon is just absurd. The Xeon still scales pretty well despite still using a FSB, you are suggesting that Barcelona scales twice as much as Xeon per clock, which would be a physical impossibility considering Xeon already scales on average 60 - 70% over dual core on highly multithreaded tasks.
Finally, to suggest that Barcelona scales so well that at 1.9GHz it matches a 2GHz Xeon while at 2.4GHz it matches a 3GHz Xeon is just absurd. The Xeon still scales pretty well despite still using a FSB, you are suggesting that Barcelona scales twice as much as Xeon per clock, which would be a physical impossibility considering Xeon already scales on average 60 - 70% over dual core on highly multithreaded tasks.
VirtualLarry
No Lifer
- Aug 25, 2001
- 56,585
- 10,225
- 126
It makes a noticable performance difference, up to 10% in some cases.
Actually, they don't perform all that well. Read the XBitLabs review on the E2140/2160. Gaming performance was very low, relative to the other chips.
Actually, the jump from 2MB -> 4MB doesn't really impact performance very much. Yes, 'up to' 10% but that is the exception rather than the rule. Most applications see less than a 5% boost.
In fact the average speedup is only 2.4% according to Anandtech. http://anandtech.com/cpuchipse...howdoc.aspx?i=2795&p=4
The performance hit from 1MB -> 2MB is somewhat bigger, in some cases up to 15%.
http://xbitlabs.com/articles/c...ium-e2160_4.html#sect0
That said, the Pentium Es are still a fair bit faster than the X2s clock for clock.
Markfw
Moderator Emeritus, Elite Member
- May 16, 2002
- 27,250
- 16,108
- 136
<div class="FTQUOTE"><begin quote>Originally posted by: harpoon84
<div class="FTQUOTE"><begin quote>Originally posted by: VirtualLarry
<div class="FTQUOTE"><begin quote>Originally posted by: zsdersw
Oh yes.. because the jump in L2 cache from 2 to 4MB *really* made a huge difference, across the board, for Core 2. :roll:</end quote></div>
It makes a noticable performance difference, up to 10% in some cases.
<div class="FTQUOTE"><begin quote>Originally posted by: zsdersw
Why do you suppose the Pentium Exxxx processors perform so relatively well with their 1MB of L2 cache? Because L2 cache isn't all there is to tell in the CMA performance story.</end quote></div>
Actually, they don't perform all that well. Read the XBitLabs review on the E2140/2160. Gaming performance was very low, relative to the other chips.
</end quote></div>
Actually, the jump from 2MB -> 4MB doesn't really impact performance very much. Yes, 'up to' 10% but that is the exception rather than the rule. Most applications see less than a 5% boost.
In fact the average speedup is only 2.4% according to Anandtech. http://anandtech.com/cpuchipse...howdoc.aspx?i=2795&p=4
The performance hit from 1MB -> 2MB is somewhat bigger, in some cases up to 15%.
http://xbitlabs.com/articles/c...ium-e2160_4.html#sect0
That said, the Pentium Es are still a fair bit faster than the X2s clock for clock.</end quote></div>
Well, there is at least one exception to that...U folding@home'ers a while back had some 1495 units. If you had an E6400@3.2 and an E6600@3.2 with two instances running these units, the 6600 would do them in HALF the time, or a 100% improvement.
I realize that is just ONE example, but Duvie and I both (and others) know it to be true.
As to this thread, first this guy should be banned for spamming, constantly making threads, and then linking to his own site. Also, its all crap, since its all speculation. I would contact a mod if I could.... Anybody want to help me out here ?
<div class="FTQUOTE"><begin quote>Originally posted by: VirtualLarry
<div class="FTQUOTE"><begin quote>Originally posted by: zsdersw
Oh yes.. because the jump in L2 cache from 2 to 4MB *really* made a huge difference, across the board, for Core 2. :roll:</end quote></div>
It makes a noticable performance difference, up to 10% in some cases.
<div class="FTQUOTE"><begin quote>Originally posted by: zsdersw
Why do you suppose the Pentium Exxxx processors perform so relatively well with their 1MB of L2 cache? Because L2 cache isn't all there is to tell in the CMA performance story.</end quote></div>
Actually, they don't perform all that well. Read the XBitLabs review on the E2140/2160. Gaming performance was very low, relative to the other chips.
</end quote></div>
Actually, the jump from 2MB -> 4MB doesn't really impact performance very much. Yes, 'up to' 10% but that is the exception rather than the rule. Most applications see less than a 5% boost.
In fact the average speedup is only 2.4% according to Anandtech. http://anandtech.com/cpuchipse...howdoc.aspx?i=2795&p=4
The performance hit from 1MB -> 2MB is somewhat bigger, in some cases up to 15%.
http://xbitlabs.com/articles/c...ium-e2160_4.html#sect0
That said, the Pentium Es are still a fair bit faster than the X2s clock for clock.</end quote></div>
Well, there is at least one exception to that...U folding@home'ers a while back had some 1495 units. If you had an E6400@3.2 and an E6600@3.2 with two instances running these units, the 6600 would do them in HALF the time, or a 100% improvement.
I realize that is just ONE example, but Duvie and I both (and others) know it to be true.
As to this thread, first this guy should be banned for spamming, constantly making threads, and then linking to his own site. Also, its all crap, since its all speculation. I would contact a mod if I could.... Anybody want to help me out here ?
zsdersw
Lifer
- Oct 29, 2003
- 10,505
- 2
- 0
The fact that there are specific instances where extra cache provides a noticeable performance benefit doesn't mean that Core 2 performance entirely *depends* on L2 cache (and that Intel uses it as a "crutch" for their "antiquated designs", which is what WAWA was asserting.
This is the exception rather then the rule. Cache obviously can be very beneficial in certain circumstances.
DrMrLordX
Lifer
- Apr 27, 2000
- 22,912
- 12,983
- 136
Well . . . I hate to appear to be agreeing with a clueless poster who joined the forum yesterday and refuses to back up his assertions, but an on-die memory controller can actually be a substitute for high density/low latency SRAM. Sort of.
If your processor is already memory bandwidth-saturated, then, in theory, you could eliminate the need for l2 cache altogether if you could get your system memory running at latencies as low as your l2. Easier said than done, of course.
An on-die memory controller comes into play by significantly reducing communication latency between the processor and the memory controller. This means the only obstacle between you and system RAM with 10-20 cycle latency is your RAM and your memory bus since, at least in AMD's on-die memory controller, the memory controller's performance scales upwards linearly with the CPU, and the latency between the CPU and memory controller seems to remain constant no matter what the CPU's frequency.
My lowly Brisbane can achieve a RAM latency of around 122 cycles when using the 9.5x multiplier and the 1:2 memory ratio. No matter what the CPU frequency I run, so long as I maintain that CPU multiplier and that memory ratio, the memory latency remains fixed at 122 cycles. Of course, using the 1:2 memory ratio meant achieving some pretty high memory clocks which in turn took some serious vdimm on my DDR2, but it was doable. However, that's only RAM latency of 122 cycles. Assuming that memory twice as fast (with the same timings) would generate half the latency, it would take something like DDR2-6384 5-5-5-15 2T to get 20.3 cycle memory latency on an X2-3600+ running at 2.66 ghz with the 9.5x multiplier. That would require a 1:12 memory ratio which I have never seen supported before. Anyone seen DDR2-6384 for that matter? I haven't, and I doubt anyone else has either.
DDR3 shows some promise, since it seems to produce nearly identical (or better) performance at the same speed rating with the same latencies. That is, DDR3-1333 5-5-5-15 seems to be about the same as DDR2-1333 5-5-5-15. Thanks to higher potential clock speeds and lower stock voltages, we should see anywhere from DDR3-2000 to DDR3-2500, at least from enthusiast parts. DDR3-2500 5-5-5-15 would be fast enough for 52 cycle memory latency when paired with a 2.66 ghz X2-3600+ using something like a 1:8.929 ratio (pfft).
Someday, either Intel or AMD will figure out how to eliminate the need for l2 by giving people absurdly fast system RAM. At that point, an HTX or CSI-connected GPU would be lookin pretty sweet.
Well, they have to do that. As with P4s, Core 2s need more cache at higher clock speeds to scale properly. Still, it's a well-executed strategy.
That's not a typo right and you realyl are talking about DDR2-6000???
DDR3-2500 if it ever comes to be is still quite some time in the future, DDR2 was introduced back in 2004, and were only now seeing DDR3 in 2007.
As well, DDR3 is inherently higher latency and only scales from CAS 5 to 9, and I am not sure where your seeing CAS 5 DDR3-1333, CAS 5-5-5 isn't even achieveable on DDR3-1066 currently let alone DDR3-1333.
But at any rate both large LV2 cache and ODMC try to achieve the general purpose of getting data to the processor quicker. They just use different approaches to achieve it. It doesn't matter in the end how you reach this end, as long as you achieve the results is what matters.
- Status
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 25K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes + WCL Discussion Threads
- Started by Tigerick
- Replies: 22K
-
News NVIDIA and Intel to Develop AI Infrastructure and Personal Computing Products- Started by poke01
- Replies: 384
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 23K
Facebook
Twitter