Overclocking
After the problems AMD has had in getting higher speed grades of the Phenom to market, I was initially worried about the overclockability of the Phenom 9900. It is less than three weeks since the release of the Phenom 9600, and I could only overclock our Phenom 9600 to 2.645GHz!
Fortunately it looks like the Phenom 9900 AMD sent us is a newer stepping - B2 to be precise - and even though it is not the fabled B3 stepping that will fix TLB erratum, it does seem to have opened up some extra "elbow" room.
How much?
How does a stable 3.0GHz strike you?
Mind you, I had to replace the stock AMD cooler with its sticky heatsink compound with a Noctua 12 and Arctic Silver - but that dropped the CPU temperature by a whopping 15'C when idling at 3GHz!
In order to run at 3GHz, I did the following:
I also tried to overclock with the HT speed, however I could not get even 233x13 stable on this Asus M3A32-MVP.
Conclusion
I was actually quite pleased with the performance of the Phenom 9900.
Given AMD's public commitment to fixing the TLB errata, and their move to 45nm with the "K10.5" architecture (Barcelona with some additional improvements) I suspect that AMD will be more competitive on the high end sometime next year.
But even until then... looking at the benchmarks, it did not seem like the Phenom 9900 was that far behind - if it was in fact behind.
The table below shows the relative rank of the three processors in the results for each benchmark - one means best result, two means second best, and third means last. These rankings are based purely on stock performance, as that is how the vast majority of people run their systems.
| Phenom 9900 | QX6700 | Xeon X3210 | |
| Business Winstone | 1 | 2 | 3 |
| Content Creation | 1 | 2 | 3 |
| Sandra CPU | 3 | 1 | 2 |
| Sandra Memory | 1 | 2 | 3 |
| WinRAR MT | 1 | 2 | 3 |
| RM Read | 1 | 2 | 3 |
| RM Write | 1 | 3 | 2 |
| RM Latency | 1 | 2 | 3 |
| RM Bandwidth | 1 | 3 | 2 |
| LAME MT | 2 | 1 | 3 |
| TMPGEnc | 3 | 1 | 2 |
| CineBench | 2 | 1 | 3 |
| POVRay | 3 | 1 | 2 |
| COD | 2 | 1 | 3 |
| Comanche | 2 | 1 | 3 |
| Doom 3 | 2 | 1 | 3 |
| Halo | 2 | 1 | 3 |
| Jedi | 2 | 1 | 3 |
| UT2004 | 2 | 1 | 3 |
| TOTAL 1st place results | 8 | 11 | 0 |
That is a very interesting result.
The Phenom 9900 generally took first place in business applications and memory oriented benchmarks - sometime with quite a margin.
The Phenom consistently beat the Xeon X3210 for the gaming tests, and was generally trounced by the much more expensive QX6700 (DUH).
Now if we end up the rankings... we see that over the whole set of benchmarks the Phenom 9900 was not far behind the QX6700 in the number of first place finishes!
Moral of the story:
The Phenom 9900 running at 2.66GHz DEFINITELY outperforms a Xeon X3210 running at 2.13GHz - as it should.
For gaming, at unrealistically low resolutions for todays gamers, the QX6700 badly beats the Phenom 9900 - but no one really games at those resolutions, and as World In Conflict shows, modern games are GPU not CPU bound.
Frankly, the Phenom 9900 did much better against the QX6700 than I had expected. Now if the BIOS on the motherboard allowed finer grained control over memory speed settings, I might have been able to squeeze even more performance out of the chip at both stock and overclocked settings; but a future BIOS revision may well take care of that.
AMD needs the 45nm shrink badly. It will address the power consumption issue - there is no question that Intel quad core processors are "greener" right now - and when that is combined with the purported architectural enhancements, and the ability to ramp up the clock speed due to lower voltage 45nm designs generating less heat, Intel may again have some real competition for the high end next year.
But you know what?
The consumer wins already. If aggressively priced, the Phenom 9900 - and upcoming lower end "Black Edition" Phenom's - will present an excellent value proposition to consumers, and provide a very viable alternative in the low and mid-range markets.
Overclocking
After the problems AMD has had in getting higher speed grades of the Phenom to market, I was initially worried about the overclockability of the Phenom 9900. It is less than three weeks since the release of the Phenom 9600, and I could only overclock our Phenom 9600 to 2.645GHz!
Fortunately it looks like the Phenom 9900 AMD sent us is a newer stepping - B2 to be precise - and even though it is not the fabled B3 stepping that will fix TLB erratum, it does seem to have opened up some extra "elbow" room.
How much?
How does a stable 3.0GHz strike you?
Mind you, I had to replace the stock AMD cooler with its sticky heatsink compound with a Noctua 12 and Arctic Silver - but that dropped the CPU temperature by a whopping 15'C when idling at 3GHz!
In order to run at 3GHz, I did the following:
I also tried to overclock with the HT speed, however I could not get even 233x13 stable on this Asus M3A32-MVP.
Conclusion
I was actually quite pleased with the performance of the Phenom 9900.
Given AMD's public commitment to fixing the TLB errata, and their move to 45nm with the "K10.5" architecture (Barcelona with some additional improvements) I suspect that AMD will be more competitive on the high end sometime next year.
But even until then... looking at the benchmarks, it did not seem like the Phenom 9900 was that far behind - if it was in fact behind.
The table below shows the relative rank of the three processors in the results for each benchmark - one means best result, two means second best, and third means last. These rankings are based purely on stock performance, as that is how the vast majority of people run their systems.
| Phenom 9900 | QX6700 | Xeon X3210 | |
| Business Winstone | 1 | 2 | 3 |
| Content Creation | 1 | 2 | 3 |
| Sandra CPU | 3 | 1 | 2 |
| Sandra Memory | 1 | 2 | 3 |
| WinRAR MT | 1 | 2 | 3 |
| RM Read | 1 | 2 | 3 |
| RM Write | 1 | 3 | 2 |
| RM Latency | 1 | 2 | 3 |
| RM Bandwidth | 1 | 3 | 2 |
| LAME MT | 2 | 1 | 3 |
| TMPGEnc | 3 | 1 | 2 |
| CineBench | 2 | 1 | 3 |
| POVRay | 3 | 1 | 2 |
| COD | 2 | 1 | 3 |
| Comanche | 2 | 1 | 3 |
| Doom 3 | 2 | 1 | 3 |
| Halo | 2 | 1 | 3 |
| Jedi | 2 | 1 | 3 |
| UT2004 | 2 | 1 | 3 |
| TOTAL 1st place results | 8 | 11 | 0 |
That is a very interesting result.
The Phenom 9900 generally took first place in business applications and memory oriented benchmarks - sometime with quite a margin.
The Phenom consistently beat the Xeon X3210 for the gaming tests, and was generally trounced by the much more expensive QX6700 (DUH).
Now if we end up the rankings... we see that over the whole set of benchmarks the Phenom 9900 was not far behind the QX6700 in the number of first place finishes!
Moral of the story:
The Phenom 9900 running at 2.66GHz DEFINITELY outperforms a Xeon X3210 running at 2.13GHz - as it should.
For gaming, at unrealistically low resolutions for todays gamers, the QX6700 badly beats the Phenom 9900 - but no one really games at those resolutions, and as World In Conflict shows, modern games are GPU not CPU bound.
Frankly, the Phenom 9900 did much better against the QX6700 than I had expected. Now if the BIOS on the motherboard allowed finer grained control over memory speed settings, I might have been able to squeeze even more performance out of the chip at both stock and overclocked settings; but a future BIOS revision may well take care of that.
AMD needs the 45nm shrink badly. It will address the power consumption issue - there is no question that Intel quad core processors are "greener" right now - and when that is combined with the purported architectural enhancements, and the ability to ramp up the clock speed due to lower voltage 45nm designs generating less heat, Intel may again have some real competition for the high end next year.
But you know what?
The consumer wins already. If aggressively priced, the Phenom 9900 - and upcoming lower end "Black Edition" Phenom's - will present an excellent value proposition to consumers, and provide a very viable alternative in the low and mid-range markets.
Overclocking
After the problems AMD has had in getting higher speed grades of the Phenom to market, I was initially worried about the overclockability of the Phenom 9900. It is less than three weeks since the release of the Phenom 9600, and I could only overclock our Phenom 9600 to 2.645GHz!
Fortunately it looks like the Phenom 9900 AMD sent us is a newer stepping - B2 to be precise - and even though it is not the fabled B3 stepping that will fix TLB erratum, it does seem to have opened up some extra "elbow" room.
How much?
How does a stable 3.0GHz strike you?
Mind you, I had to replace the stock AMD cooler with its sticky heatsink compound with a Noctua 12 and Arctic Silver - but that dropped the CPU temperature by a whopping 15'C when idling at 3GHz!
In order to run at 3GHz, I did the following:
I also tried to overclock with the HT speed, however I could not get even 233x13 stable on this Asus M3A32-MVP.
Conclusion
I was actually quite pleased with the performance of the Phenom 9900.
Given AMD's public commitment to fixing the TLB errata, and their move to 45nm with the "K10.5" architecture (Barcelona with some additional improvements) I suspect that AMD will be more competitive on the high end sometime next year.
But even until then... looking at the benchmarks, it did not seem like the Phenom 9900 was that far behind - if it was in fact behind.
The table below shows the relative rank of the three processors in the results for each benchmark - one means best result, two means second best, and third means last. These rankings are based purely on stock performance, as that is how the vast majority of people run their systems.
| Phenom 9900 | QX6700 | Xeon X3210 | |
| Business Winstone | 1 | 2 | 3 |
| Content Creation | 1 | 2 | 3 |
| Sandra CPU | 3 | 1 | 2 |
| Sandra Memory | 1 | 2 | 3 |
| WinRAR MT | 1 | 2 | 3 |
| RM Read | 1 | 2 | 3 |
| RM Write | 1 | 3 | 2 |
| RM Latency | 1 | 2 | 3 |
| RM Bandwidth | 1 | 3 | 2 |
| LAME MT | 2 | 1 | 3 |
| TMPGEnc | 3 | 1 | 2 |
| CineBench | 2 | 1 | 3 |
| POVRay | 3 | 1 | 2 |
| COD | 2 | 1 | 3 |
| Comanche | 2 | 1 | 3 |
| Doom 3 | 2 | 1 | 3 |
| Halo | 2 | 1 | 3 |
| Jedi | 2 | 1 | 3 |
| UT2004 | 2 | 1 | 3 |
| TOTAL 1st place results | 8 | 11 | 0 |
That is a very interesting result.
The Phenom 9900 generally took first place in business applications and memory oriented benchmarks - sometime with quite a margin.
The Phenom consistently beat the Xeon X3210 for the gaming tests, and was generally trounced by the much more expensive QX6700 (DUH).
Now if we end up the rankings... we see that over the whole set of benchmarks the Phenom 9900 was not far behind the QX6700 in the number of first place finishes!
Moral of the story:
The Phenom 9900 running at 2.66GHz DEFINITELY outperforms a Xeon X3210 running at 2.13GHz - as it should.
For gaming, at unrealistically low resolutions for todays gamers, the QX6700 badly beats the Phenom 9900 - but no one really games at those resolutions, and as World In Conflict shows, modern games are GPU not CPU bound.
Frankly, the Phenom 9900 did much better against the QX6700 than I had expected. Now if the BIOS on the motherboard allowed finer grained control over memory speed settings, I might have been able to squeeze even more performance out of the chip at both stock and overclocked settings; but a future BIOS revision may well take care of that.
AMD needs the 45nm shrink badly. It will address the power consumption issue - there is no question that Intel quad core processors are "greener" right now - and when that is combined with the purported architectural enhancements, and the ability to ramp up the clock speed due to lower voltage 45nm designs generating less heat, Intel may again have some real competition for the high end next year.
But you know what?
The consumer wins already. If aggressively priced, the Phenom 9900 - and upcoming lower end "Black Edition" Phenom's - will present an excellent value proposition to consumers, and provide a very viable alternative in the low and mid-range markets.
Overclocking
After the problems AMD has had in getting higher speed grades of the Phenom to market, I was initially worried about the overclockability of the Phenom 9900. It is less than three weeks since the release of the Phenom 9600, and I could only overclock our Phenom 9600 to 2.645GHz!
Fortunately it looks like the Phenom 9900 AMD sent us is a newer stepping - B2 to be precise - and even though it is not the fabled B3 stepping that will fix TLB erratum, it does seem to have opened up some extra "elbow" room.
How much?
How does a stable 3.0GHz strike you?
Mind you, I had to replace the stock AMD cooler with its sticky heatsink compound with a Noctua 12 and Arctic Silver - but that dropped the CPU temperature by a whopping 15'C when idling at 3GHz!
In order to run at 3GHz, I did the following:
I also tried to overclock with the HT speed, however I could not get even 233x13 stable on this Asus M3A32-MVP.
Conclusion
I was actually quite pleased with the performance of the Phenom 9900.
Given AMD's public commitment to fixing the TLB errata, and their move to 45nm with the "K10.5" architecture (Barcelona with some additional improvements) I suspect that AMD will be more competitive on the high end sometime next year.
But even until then... looking at the benchmarks, it did not seem like the Phenom 9900 was that far behind - if it was in fact behind.
The table below shows the relative rank of the three processors in the results for each benchmark - one means best result, two means second best, and third means last. These rankings are based purely on stock performance, as that is how the vast majority of people run their systems.
| Phenom 9900 | QX6700 | Xeon X3210 | |
| Business Winstone | 1 | 2 | 3 |
| Content Creation | 1 | 2 | 3 |
| Sandra CPU | 3 | 1 | 2 |
| Sandra Memory | 1 | 2 | 3 |
| WinRAR MT | 1 | 2 | 3 |
| RM Read | 1 | 2 | 3 |
| RM Write | 1 | 3 | 2 |
| RM Latency | 1 | 2 | 3 |
| RM Bandwidth | 1 | 3 | 2 |
| LAME MT | 2 | 1 | 3 |
| TMPGEnc | 3 | 1 | 2 |
| CineBench | 2 | 1 | 3 |
| POVRay | 3 | 1 | 2 |
| COD | 2 | 1 | 3 |
| Comanche | 2 | 1 | 3 |
| Doom 3 | 2 | 1 | 3 |
| Halo | 2 | 1 | 3 |
| Jedi | 2 | 1 | 3 |
| UT2004 | 2 | 1 | 3 |
| TOTAL 1st place results | 8 | 11 | 0 |
That is a very interesting result.
The Phenom 9900 generally took first place in business applications and memory oriented benchmarks - sometime with quite a margin.
The Phenom consistently beat the Xeon X3210 for the gaming tests, and was generally trounced by the much more expensive QX6700 (DUH).
Now if we end up the rankings... we see that over the whole set of benchmarks the Phenom 9900 was not far behind the QX6700 in the number of first place finishes!
Moral of the story:
The Phenom 9900 running at 2.66GHz DEFINITELY outperforms a Xeon X3210 running at 2.13GHz - as it should.
For gaming, at unrealistically low resolutions for todays gamers, the QX6700 badly beats the Phenom 9900 - but no one really games at those resolutions, and as World In Conflict shows, modern games are GPU not CPU bound.
Frankly, the Phenom 9900 did much better against the QX6700 than I had expected. Now if the BIOS on the motherboard allowed finer grained control over memory speed settings, I might have been able to squeeze even more performance out of the chip at both stock and overclocked settings; but a future BIOS revision may well take care of that.
AMD needs the 45nm shrink badly. It will address the power consumption issue - there is no question that Intel quad core processors are "greener" right now - and when that is combined with the purported architectural enhancements, and the ability to ramp up the clock speed due to lower voltage 45nm designs generating less heat, Intel may again have some real competition for the high end next year.
But you know what?
The consumer wins already. If aggressively priced, the Phenom 9900 - and upcoming lower end "Black Edition" Phenom's - will present an excellent value proposition to consumers, and provide a very viable alternative in the low and mid-range markets.
Overclocking
After the problems AMD has had in getting higher speed grades of the Phenom to market, I was initially worried about the overclockability of the Phenom 9900. It is less than three weeks since the release of the Phenom 9600, and I could only overclock our Phenom 9600 to 2.645GHz!
Fortunately it looks like the Phenom 9900 AMD sent us is a newer stepping - B2 to be precise - and even though it is not the fabled B3 stepping that will fix TLB erratum, it does seem to have opened up some extra "elbow" room.
How much?
How does a stable 3.0GHz strike you?
Mind you, I had to replace the stock AMD cooler with its sticky heatsink compound with a Noctua 12 and Arctic Silver - but that dropped the CPU temperature by a whopping 15'C when idling at 3GHz!
In order to run at 3GHz, I did the following:
I also tried to overclock with the HT speed, however I could not get even 233x13 stable on this Asus M3A32-MVP.
Conclusion
I was actually quite pleased with the performance of the Phenom 9900.
Given AMD's public commitment to fixing the TLB errata, and their move to 45nm with the "K10.5" architecture (Barcelona with some additional improvements) I suspect that AMD will be more competitive on the high end sometime next year.
But even until then... looking at the benchmarks, it did not seem like the Phenom 9900 was that far behind - if it was in fact behind.
The table below shows the relative rank of the three processors in the results for each benchmark - one means best result, two means second best, and third means last. These rankings are based purely on stock performance, as that is how the vast majority of people run their systems.
| Phenom 9900 | QX6700 | Xeon X3210 | |
| Business Winstone | 1 | 2 | 3 |
| Content Creation | 1 | 2 | 3 |
| Sandra CPU | 3 | 1 | 2 |
| Sandra Memory | 1 | 2 | 3 |
| WinRAR MT | 1 | 2 | 3 |
| RM Read | 1 | 2 | 3 |
| RM Write | 1 | 3 | 2 |
| RM Latency | 1 | 2 | 3 |
| RM Bandwidth | 1 | 3 | 2 |
| LAME MT | 2 | 1 | 3 |
| TMPGEnc | 3 | 1 | 2 |
| CineBench | 2 | 1 | 3 |
| POVRay | 3 | 1 | 2 |
| COD | 2 | 1 | 3 |
| Comanche | 2 | 1 | 3 |
| Doom 3 | 2 | 1 | 3 |
| Halo | 2 | 1 | 3 |
| Jedi | 2 | 1 | 3 |
| UT2004 | 2 | 1 | 3 |
| TOTAL 1st place results | 8 | 11 | 0 |
That is a very interesting result.
The Phenom 9900 generally took first place in business applications and memory oriented benchmarks - sometime with quite a margin.
The Phenom consistently beat the Xeon X3210 for the gaming tests, and was generally trounced by the much more expensive QX6700 (DUH).
Now if we end up the rankings... we see that over the whole set of benchmarks the Phenom 9900 was not far behind the QX6700 in the number of first place finishes!
Moral of the story:
The Phenom 9900 running at 2.66GHz DEFINITELY outperforms a Xeon X3210 running at 2.13GHz - as it should.
For gaming, at unrealistically low resolutions for todays gamers, the QX6700 badly beats the Phenom 9900 - but no one really games at those resolutions, and as World In Conflict shows, modern games are GPU not CPU bound.
Frankly, the Phenom 9900 did much better against the QX6700 than I had expected. Now if the BIOS on the motherboard allowed finer grained control over memory speed settings, I might have been able to squeeze even more performance out of the chip at both stock and overclocked settings; but a future BIOS revision may well take care of that.
AMD needs the 45nm shrink badly. It will address the power consumption issue - there is no question that Intel quad core processors are "greener" right now - and when that is combined with the purported architectural enhancements, and the ability to ramp up the clock speed due to lower voltage 45nm designs generating less heat, Intel may again have some real competition for the high end next year.
But you know what?
The consumer wins already. If aggressively priced, the Phenom 9900 - and upcoming lower end "Black Edition" Phenom's - will present an excellent value proposition to consumers, and provide a very viable alternative in the low and mid-range markets.
Overclocking
After the problems AMD has had in getting higher speed grades of the Phenom to market, I was initially worried about the overclockability of the Phenom 9900. It is less than three weeks since the release of the Phenom 9600, and I could only overclock our Phenom 9600 to 2.645GHz!
Fortunately it looks like the Phenom 9900 AMD sent us is a newer stepping - B2 to be precise - and even though it is not the fabled B3 stepping that will fix TLB erratum, it does seem to have opened up some extra "elbow" room.
How much?
How does a stable 3.0GHz strike you?
Mind you, I had to replace the stock AMD cooler with its sticky heatsink compound with a Noctua 12 and Arctic Silver - but that dropped the CPU temperature by a whopping 15'C when idling at 3GHz!
In order to run at 3GHz, I did the following:
I also tried to overclock with the HT speed, however I could not get even 233x13 stable on this Asus M3A32-MVP.
Conclusion
I was actually quite pleased with the performance of the Phenom 9900.
Given AMD's public commitment to fixing the TLB errata, and their move to 45nm with the "K10.5" architecture (Barcelona with some additional improvements) I suspect that AMD will be more competitive on the high end sometime next year.
But even until then... looking at the benchmarks, it did not seem like the Phenom 9900 was that far behind - if it was in fact behind.
The table below shows the relative rank of the three processors in the results for each benchmark - one means best result, two means second best, and third means last. These rankings are based purely on stock performance, as that is how the vast majority of people run their systems.
| Phenom 9900 | QX6700 | Xeon X3210 | |
| Business Winstone | 1 | 2 | 3 |
| Content Creation | 1 | 2 | 3 |
| Sandra CPU | 3 | 1 | 2 |
| Sandra Memory | 1 | 2 | 3 |
| WinRAR MT | 1 | 2 | 3 |
| RM Read | 1 | 2 | 3 |
| RM Write | 1 | 3 | 2 |
| RM Latency | 1 | 2 | 3 |
| RM Bandwidth | 1 | 3 | 2 |
| LAME MT | 2 | 1 | 3 |
| TMPGEnc | 3 | 1 | 2 |
| CineBench | 2 | 1 | 3 |
| POVRay | 3 | 1 | 2 |
| COD | 2 | 1 | 3 |
| Comanche | 2 | 1 | 3 |
| Doom 3 | 2 | 1 | 3 |
| Halo | 2 | 1 | 3 |
| Jedi | 2 | 1 | 3 |
| UT2004 | 2 | 1 | 3 |
| TOTAL 1st place results | 8 | 11 | 0 |
That is a very interesting result.
The Phenom 9900 generally took first place in business applications and memory oriented benchmarks - sometime with quite a margin.
The Phenom consistently beat the Xeon X3210 for the gaming tests, and was generally trounced by the much more expensive QX6700 (DUH).
Now if we end up the rankings... we see that over the whole set of benchmarks the Phenom 9900 was not far behind the QX6700 in the number of first place finishes!
Moral of the story:
The Phenom 9900 running at 2.66GHz DEFINITELY outperforms a Xeon X3210 running at 2.13GHz - as it should.
For gaming, at unrealistically low resolutions for todays gamers, the QX6700 badly beats the Phenom 9900 - but no one really games at those resolutions, and as World In Conflict shows, modern games are GPU not CPU bound.
Frankly, the Phenom 9900 did much better against the QX6700 than I had expected. Now if the BIOS on the motherboard allowed finer grained control over memory speed settings, I might have been able to squeeze even more performance out of the chip at both stock and overclocked settings; but a future BIOS revision may well take care of that.
AMD needs the 45nm shrink badly. It will address the power consumption issue - there is no question that Intel quad core processors are "greener" right now - and when that is combined with the purported architectural enhancements, and the ability to ramp up the clock speed due to lower voltage 45nm designs generating less heat, Intel may again have some real competition for the high end next year.
But you know what?
The consumer wins already. If aggressively priced, the Phenom 9900 - and upcoming lower end "Black Edition" Phenom's - will present an excellent value proposition to consumers, and provide a very viable alternative in the low and mid-range markets.
Overclocking
After the problems AMD has had in getting higher speed grades of the Phenom to market, I was initially worried about the overclockability of the Phenom 9900. It is less than three weeks since the release of the Phenom 9600, and I could only overclock our Phenom 9600 to 2.645GHz!
Fortunately it looks like the Phenom 9900 AMD sent us is a newer stepping - B2 to be precise - and even though it is not the fabled B3 stepping that will fix TLB erratum, it does seem to have opened up some extra "elbow" room.
How much?
How does a stable 3.0GHz strike you?
Mind you, I had to replace the stock AMD cooler with its sticky heatsink compound with a Noctua 12 and Arctic Silver - but that dropped the CPU temperature by a whopping 15'C when idling at 3GHz!
In order to run at 3GHz, I did the following:
I also tried to overclock with the HT speed, however I could not get even 233x13 stable on this Asus M3A32-MVP.
Conclusion
I was actually quite pleased with the performance of the Phenom 9900.
Given AMD's public commitment to fixing the TLB errata, and their move to 45nm with the "K10.5" architecture (Barcelona with some additional improvements) I suspect that AMD will be more competitive on the high end sometime next year.
But even until then... looking at the benchmarks, it did not seem like the Phenom 9900 was that far behind - if it was in fact behind.
The table below shows the relative rank of the three processors in the results for each benchmark - one means best result, two means second best, and third means last. These rankings are based purely on stock performance, as that is how the vast majority of people run their systems.
| Phenom 9900 | QX6700 | Xeon X3210 | |
| Business Winstone | 1 | 2 | 3 |
| Content Creation | 1 | 2 | 3 |
| Sandra CPU | 3 | 1 | 2 |
| Sandra Memory | 1 | 2 | 3 |
| WinRAR MT | 1 | 2 | 3 |
| RM Read | 1 | 2 | 3 |
| RM Write | 1 | 3 | 2 |
| RM Latency | 1 | 2 | 3 |
| RM Bandwidth | 1 | 3 | 2 |
| LAME MT | 2 | 1 | 3 |
| TMPGEnc | 3 | 1 | 2 |
| CineBench | 2 | 1 | 3 |
| POVRay | 3 | 1 | 2 |
| COD | 2 | 1 | 3 |
| Comanche | 2 | 1 | 3 |
| Doom 3 | 2 | 1 | 3 |
| Halo | 2 | 1 | 3 |
| Jedi | 2 | 1 | 3 |
| UT2004 | 2 | 1 | 3 |
| TOTAL 1st place results | 8 | 11 | 0 |
That is a very interesting result.
The Phenom 9900 generally took first place in business applications and memory oriented benchmarks - sometime with quite a margin.
The Phenom consistently beat the Xeon X3210 for the gaming tests, and was generally trounced by the much more expensive QX6700 (DUH).
Now if we end up the rankings... we see that over the whole set of benchmarks the Phenom 9900 was not far behind the QX6700 in the number of first place finishes!
Moral of the story:
The Phenom 9900 running at 2.66GHz DEFINITELY outperforms a Xeon X3210 running at 2.13GHz - as it should.
For gaming, at unrealistically low resolutions for todays gamers, the QX6700 badly beats the Phenom 9900 - but no one really games at those resolutions, and as World In Conflict shows, modern games are GPU not CPU bound.
Frankly, the Phenom 9900 did much better against the QX6700 than I had expected. Now if the BIOS on the motherboard allowed finer grained control over memory speed settings, I might have been able to squeeze even more performance out of the chip at both stock and overclocked settings; but a future BIOS revision may well take care of that.
AMD needs the 45nm shrink badly. It will address the power consumption issue - there is no question that Intel quad core processors are "greener" right now - and when that is combined with the purported architectural enhancements, and the ability to ramp up the clock speed due to lower voltage 45nm designs generating less heat, Intel may again have some real competition for the high end next year.
But you know what?
The consumer wins already. If aggressively priced, the Phenom 9900 - and upcoming lower end "Black Edition" Phenom's - will present an excellent value proposition to consumers, and provide a very viable alternative in the low and mid-range markets.
Overclocking
After the problems AMD has had in getting higher speed grades of the Phenom to market, I was initially worried about the overclockability of the Phenom 9900. It is less than three weeks since the release of the Phenom 9600, and I could only overclock our Phenom 9600 to 2.645GHz!
Fortunately it looks like the Phenom 9900 AMD sent us is a newer stepping - B2 to be precise - and even though it is not the fabled B3 stepping that will fix TLB erratum, it does seem to have opened up some extra "elbow" room.
How much?
How does a stable 3.0GHz strike you?
Mind you, I had to replace the stock AMD cooler with its sticky heatsink compound with a Noctua 12 and Arctic Silver - but that dropped the CPU temperature by a whopping 15'C when idling at 3GHz!
In order to run at 3GHz, I did the following:
I also tried to overclock with the HT speed, however I could not get even 233x13 stable on this Asus M3A32-MVP.
Conclusion
I was actually quite pleased with the performance of the Phenom 9900.
Given AMD's public commitment to fixing the TLB errata, and their move to 45nm with the "K10.5" architecture (Barcelona with some additional improvements) I suspect that AMD will be more competitive on the high end sometime next year.
But even until then... looking at the benchmarks, it did not seem like the Phenom 9900 was that far behind - if it was in fact behind.
The table below shows the relative rank of the three processors in the results for each benchmark - one means best result, two means second best, and third means last. These rankings are based purely on stock performance, as that is how the vast majority of people run their systems.
| Phenom 9900 | QX6700 | Xeon X3210 | |
| Business Winstone | 1 | 2 | 3 |
| Content Creation | 1 | 2 | 3 |
| Sandra CPU | 3 | 1 | 2 |
| Sandra Memory | 1 | 2 | 3 |
| WinRAR MT | 1 | 2 | 3 |
| RM Read | 1 | 2 | 3 |
| RM Write | 1 | 3 | 2 |
| RM Latency | 1 | 2 | 3 |
| RM Bandwidth | 1 | 3 | 2 |
| LAME MT | 2 | 1 | 3 |
| TMPGEnc | 3 | 1 | 2 |
| CineBench | 2 | 1 | 3 |
| POVRay | 3 | 1 | 2 |
| COD | 2 | 1 | 3 |
| Comanche | 2 | 1 | 3 |
| Doom 3 | 2 | 1 | 3 |
| Halo | 2 | 1 | 3 |
| Jedi | 2 | 1 | 3 |
| UT2004 | 2 | 1 | 3 |
| TOTAL 1st place results | 8 | 11 | 0 |
That is a very interesting result.
The Phenom 9900 generally took first place in business applications and memory oriented benchmarks - sometime with quite a margin.
The Phenom consistently beat the Xeon X3210 for the gaming tests, and was generally trounced by the much more expensive QX6700 (DUH).
Now if we end up the rankings... we see that over the whole set of benchmarks the Phenom 9900 was not far behind the QX6700 in the number of first place finishes!
Moral of the story:
The Phenom 9900 running at 2.66GHz DEFINITELY outperforms a Xeon X3210 running at 2.13GHz - as it should.
For gaming, at unrealistically low resolutions for todays gamers, the QX6700 badly beats the Phenom 9900 - but no one really games at those resolutions, and as World In Conflict shows, modern games are GPU not CPU bound.
Frankly, the Phenom 9900 did much better against the QX6700 than I had expected. Now if the BIOS on the motherboard allowed finer grained control over memory speed settings, I might have been able to squeeze even more performance out of the chip at both stock and overclocked settings; but a future BIOS revision may well take care of that.
AMD needs the 45nm shrink badly. It will address the power consumption issue - there is no question that Intel quad core processors are "greener" right now - and when that is combined with the purported architectural enhancements, and the ability to ramp up the clock speed due to lower voltage 45nm designs generating less heat, Intel may again have some real competition for the high end next year.
But you know what?
The consumer wins already. If aggressively priced, the Phenom 9900 - and upcoming lower end "Black Edition" Phenom's - will present an excellent value proposition to consumers, and provide a very viable alternative in the low and mid-range markets.
Overclocking
After the problems AMD has had in getting higher speed grades of the Phenom to market, I was initially worried about the overclockability of the Phenom 9900. It is less than three weeks since the release of the Phenom 9600, and I could only overclock our Phenom 9600 to 2.645GHz!
Fortunately it looks like the Phenom 9900 AMD sent us is a newer stepping - B2 to be precise - and even though it is not the fabled B3 stepping that will fix TLB erratum, it does seem to have opened up some extra "elbow" room.
How much?
How does a stable 3.0GHz strike you?
Mind you, I had to replace the stock AMD cooler with its sticky heatsink compound with a Noctua 12 and Arctic Silver - but that dropped the CPU temperature by a whopping 15'C when idling at 3GHz!
In order to run at 3GHz, I did the following:
I also tried to overclock with the HT speed, however I could not get even 233x13 stable on this Asus M3A32-MVP.
Conclusion
I was actually quite pleased with the performance of the Phenom 9900.
Given AMD's public commitment to fixing the TLB errata, and their move to 45nm with the "K10.5" architecture (Barcelona with some additional improvements) I suspect that AMD will be more competitive on the high end sometime next year.
But even until then... looking at the benchmarks, it did not seem like the Phenom 9900 was that far behind - if it was in fact behind.
The table below shows the relative rank of the three processors in the results for each benchmark - one means best result, two means second best, and third means last. These rankings are based purely on stock performance, as that is how the vast majority of people run their systems.
| Phenom 9900 | QX6700 | Xeon X3210 | |
| Business Winstone | 1 | 2 | 3 |
| Content Creation | 1 | 2 | 3 |
| Sandra CPU | 3 | 1 | 2 |
| Sandra Memory | 1 | 2 | 3 |
| WinRAR MT | 1 | 2 | 3 |
| RM Read | 1 | 2 | 3 |
| RM Write | 1 | 3 | 2 |
| RM Latency | 1 | 2 | 3 |
| RM Bandwidth | 1 | 3 | 2 |
| LAME MT | 2 | 1 | 3 |
| TMPGEnc | 3 | 1 | 2 |
| CineBench | 2 | 1 | 3 |
| POVRay | 3 | 1 | 2 |
| COD | 2 | 1 | 3 |
| Comanche | 2 | 1 | 3 |
| Doom 3 | 2 | 1 | 3 |
| Halo | 2 | 1 | 3 |
| Jedi | 2 | 1 | 3 |
| UT2004 | 2 | 1 | 3 |
| TOTAL 1st place results | 8 | 11 | 0 |
That is a very interesting result.
The Phenom 9900 generally took first place in business applications and memory oriented benchmarks - sometime with quite a margin.
The Phenom consistently beat the Xeon X3210 for the gaming tests, and was generally trounced by the much more expensive QX6700 (DUH).
Now if we end up the rankings... we see that over the whole set of benchmarks the Phenom 9900 was not far behind the QX6700 in the number of first place finishes!
Moral of the story:
The Phenom 9900 running at 2.66GHz DEFINITELY outperforms a Xeon X3210 running at 2.13GHz - as it should.
For gaming, at unrealistically low resolutions for todays gamers, the QX6700 badly beats the Phenom 9900 - but no one really games at those resolutions, and as World In Conflict shows, modern games are GPU not CPU bound.
Frankly, the Phenom 9900 did much better against the QX6700 than I had expected. Now if the BIOS on the motherboard allowed finer grained control over memory speed settings, I might have been able to squeeze even more performance out of the chip at both stock and overclocked settings; but a future BIOS revision may well take care of that.
AMD needs the 45nm shrink badly. It will address the power consumption issue - there is no question that Intel quad core processors are "greener" right now - and when that is combined with the purported architectural enhancements, and the ability to ramp up the clock speed due to lower voltage 45nm designs generating less heat, Intel may again have some real competition for the high end next year.
But you know what?
The consumer wins already. If aggressively priced, the Phenom 9900 - and upcoming lower end "Black Edition" Phenom's - will present an excellent value proposition to consumers, and provide a very viable alternative in the low and mid-range markets.
Overclocking
After the problems AMD has had in getting higher speed grades of the Phenom to market, I was initially worried about the overclockability of the Phenom 9900. It is less than three weeks since the release of the Phenom 9600, and I could only overclock our Phenom 9600 to 2.645GHz!
Fortunately it looks like the Phenom 9900 AMD sent us is a newer stepping - B2 to be precise - and even though it is not the fabled B3 stepping that will fix TLB erratum, it does seem to have opened up some extra "elbow" room.
How much?
How does a stable 3.0GHz strike you?
Mind you, I had to replace the stock AMD cooler with its sticky heatsink compound with a Noctua 12 and Arctic Silver - but that dropped the CPU temperature by a whopping 15'C when idling at 3GHz!
In order to run at 3GHz, I did the following:
I also tried to overclock with the HT speed, however I could not get even 233x13 stable on this Asus M3A32-MVP.
Conclusion
I was actually quite pleased with the performance of the Phenom 9900.
Given AMD's public commitment to fixing the TLB errata, and their move to 45nm with the "K10.5" architecture (Barcelona with some additional improvements) I suspect that AMD will be more competitive on the high end sometime next year.
But even until then... looking at the benchmarks, it did not seem like the Phenom 9900 was that far behind - if it was in fact behind.
The table below shows the relative rank of the three processors in the results for each benchmark - one means best result, two means second best, and third means last. These rankings are based purely on stock performance, as that is how the vast majority of people run their systems.
| Phenom 9900 | QX6700 | Xeon X3210 | |
| Business Winstone | 1 | 2 | 3 |
| Content Creation | 1 | 2 | 3 |
| Sandra CPU | 3 | 1 | 2 |
| Sandra Memory | 1 | 2 | 3 |
| WinRAR MT | 1 | 2 | 3 |
| RM Read | 1 | 2 | 3 |
| RM Write | 1 | 3 | 2 |
| RM Latency | 1 | 2 | 3 |
| RM Bandwidth | 1 | 3 | 2 |
| LAME MT | 2 | 1 | 3 |
| TMPGEnc | 3 | 1 | 2 |
| CineBench | 2 | 1 | 3 |
| POVRay | 3 | 1 | 2 |
| COD | 2 | 1 | 3 |
| Comanche | 2 | 1 | 3 |
| Doom 3 | 2 | 1 | 3 |
| Halo | 2 | 1 | 3 |
| Jedi | 2 | 1 | 3 |
| UT2004 | 2 | 1 | 3 |
| TOTAL 1st place results | 8 | 11 | 0 |
That is a very interesting result.
The Phenom 9900 generally took first place in business applications and memory oriented benchmarks - sometime with quite a margin.
The Phenom consistently beat the Xeon X3210 for the gaming tests, and was generally trounced by the much more expensive QX6700 (DUH).
Now if we end up the rankings... we see that over the whole set of benchmarks the Phenom 9900 was not far behind the QX6700 in the number of first place finishes!
Moral of the story:
The Phenom 9900 running at 2.66GHz DEFINITELY outperforms a Xeon X3210 running at 2.13GHz - as it should.
For gaming, at unrealistically low resolutions for todays gamers, the QX6700 badly beats the Phenom 9900 - but no one really games at those resolutions, and as World In Conflict shows, modern games are GPU not CPU bound.
Frankly, the Phenom 9900 did much better against the QX6700 than I had expected. Now if the BIOS on the motherboard allowed finer grained control over memory speed settings, I might have been able to squeeze even more performance out of the chip at both stock and overclocked settings; but a future BIOS revision may well take care of that.
AMD needs the 45nm shrink badly. It will address the power consumption issue - there is no question that Intel quad core processors are "greener" right now - and when that is combined with the purported architectural enhancements, and the ability to ramp up the clock speed due to lower voltage 45nm designs generating less heat, Intel may again have some real competition for the high end next year.
But you know what?
The consumer wins already. If aggressively priced, the Phenom 9900 - and upcoming lower end "Black Edition" Phenom's - will present an excellent value proposition to consumers, and provide a very viable alternative in the low and mid-range markets.
Overclocking
After the problems AMD has had in getting higher speed grades of the Phenom to market, I was initially worried about the overclockability of the Phenom 9900. It is less than three weeks since the release of the Phenom 9600, and I could only overclock our Phenom 9600 to 2.645GHz!
Fortunately it looks like the Phenom 9900 AMD sent us is a newer stepping - B2 to be precise - and even though it is not the fabled B3 stepping that will fix TLB erratum, it does seem to have opened up some extra "elbow" room.
How much?
How does a stable 3.0GHz strike you?
Mind you, I had to replace the stock AMD cooler with its sticky heatsink compound with a Noctua 12 and Arctic Silver - but that dropped the CPU temperature by a whopping 15'C when idling at 3GHz!
In order to run at 3GHz, I did the following:
I also tried to overclock with the HT speed, however I could not get even 233x13 stable on this Asus M3A32-MVP.
Conclusion
I was actually quite pleased with the performance of the Phenom 9900.
Given AMD's public commitment to fixing the TLB errata, and their move to 45nm with the "K10.5" architecture (Barcelona with some additional improvements) I suspect that AMD will be more competitive on the high end sometime next year.
But even until then... looking at the benchmarks, it did not seem like the Phenom 9900 was that far behind - if it was in fact behind.
The table below shows the relative rank of the three processors in the results for each benchmark - one means best result, two means second best, and third means last. These rankings are based purely on stock performance, as that is how the vast majority of people run their systems.
| Phenom 9900 | QX6700 | Xeon X3210 | |
| Business Winstone | 1 | 2 | 3 |
| Content Creation | 1 | 2 | 3 |
| Sandra CPU | 3 | 1 | 2 |
| Sandra Memory | 1 | 2 | 3 |
| WinRAR MT | 1 | 2 | 3 |
| RM Read | 1 | 2 | 3 |
| RM Write | 1 | 3 | 2 |
| RM Latency | 1 | 2 | 3 |
| RM Bandwidth | 1 | 3 | 2 |
| LAME MT | 2 | 1 | 3 |
| TMPGEnc | 3 | 1 | 2 |
| CineBench | 2 | 1 | 3 |
| POVRay | 3 | 1 | 2 |
| COD | 2 | 1 | 3 |
| Comanche | 2 | 1 | 3 |
| Doom 3 | 2 | 1 | 3 |
| Halo | 2 | 1 | 3 |
| Jedi | 2 | 1 | 3 |
| UT2004 | 2 | 1 | 3 |
| TOTAL 1st place results | 8 | 11 | 0 |
That is a very interesting result.
The Phenom 9900 generally took first place in business applications and memory oriented benchmarks - sometime with quite a margin.
The Phenom consistently beat the Xeon X3210 for the gaming tests, and was generally trounced by the much more expensive QX6700 (DUH).
Now if we end up the rankings... we see that over the whole set of benchmarks the Phenom 9900 was not far behind the QX6700 in the number of first place finishes!
Moral of the story:
The Phenom 9900 running at 2.66GHz DEFINITELY outperforms a Xeon X3210 running at 2.13GHz - as it should.
For gaming, at unrealistically low resolutions for todays gamers, the QX6700 badly beats the Phenom 9900 - but no one really games at those resolutions, and as World In Conflict shows, modern games are GPU not CPU bound.
Frankly, the Phenom 9900 did much better against the QX6700 than I had expected. Now if the BIOS on the motherboard allowed finer grained control over memory speed settings, I might have been able to squeeze even more performance out of the chip at both stock and overclocked settings; but a future BIOS revision may well take care of that.
AMD needs the 45nm shrink badly. It will address the power consumption issue - there is no question that Intel quad core processors are "greener" right now - and when that is combined with the purported architectural enhancements, and the ability to ramp up the clock speed due to lower voltage 45nm designs generating less heat, Intel may again have some real competition for the high end next year.
But you know what?
The consumer wins already. If aggressively priced, the Phenom 9900 - and upcoming lower end "Black Edition" Phenom's - will present an excellent value proposition to consumers, and provide a very viable alternative in the low and mid-range markets.
Overclocking
After the problems AMD has had in getting higher speed grades of the Phenom to market, I was initially worried about the overclockability of the Phenom 9900. It is less than three weeks since the release of the Phenom 9600, and I could only overclock our Phenom 9600 to 2.645GHz!
Fortunately it looks like the Phenom 9900 AMD sent us is a newer stepping - B2 to be precise - and even though it is not the fabled B3 stepping that will fix TLB erratum, it does seem to have opened up some extra "elbow" room.
How much?
How does a stable 3.0GHz strike you?
Mind you, I had to replace the stock AMD cooler with its sticky heatsink compound with a Noctua 12 and Arctic Silver - but that dropped the CPU temperature by a whopping 15'C when idling at 3GHz!
In order to run at 3GHz, I did the following:
I also tried to overclock with the HT speed, however I could not get even 233x13 stable on this Asus M3A32-MVP.
Conclusion
I was actually quite pleased with the performance of the Phenom 9900.
Given AMD's public commitment to fixing the TLB errata, and their move to 45nm with the "K10.5" architecture (Barcelona with some additional improvements) I suspect that AMD will be more competitive on the high end sometime next year.
But even until then... looking at the benchmarks, it did not seem like the Phenom 9900 was that far behind - if it was in fact behind.
The table below shows the relative rank of the three processors in the results for each benchmark - one means best result, two means second best, and third means last. These rankings are based purely on stock performance, as that is how the vast majority of people run their systems.
| Phenom 9900 | QX6700 | Xeon X3210 | |
| Business Winstone | 1 | 2 | 3 |
| Content Creation | 1 | 2 | 3 |
| Sandra CPU | 3 | 1 | 2 |
| Sandra Memory | 1 | 2 | 3 |
| WinRAR MT | 1 | 2 | 3 |
| RM Read | 1 | 2 | 3 |
| RM Write | 1 | 3 | 2 |
| RM Latency | 1 | 2 | 3 |
| RM Bandwidth | 1 | 3 | 2 |
| LAME MT | 2 | 1 | 3 |
| TMPGEnc | 3 | 1 | 2 |
| CineBench | 2 | 1 | 3 |
| POVRay | 3 | 1 | 2 |
| COD | 2 | 1 | 3 |
| Comanche | 2 | 1 | 3 |
| Doom 3 | 2 | 1 | 3 |
| Halo | 2 | 1 | 3 |
| Jedi | 2 | 1 | 3 |
| UT2004 | 2 | 1 | 3 |
| TOTAL 1st place results | 8 | 11 | 0 |
That is a very interesting result.
The Phenom 9900 generally took first place in business applications and memory oriented benchmarks - sometime with quite a margin.
The Phenom consistently beat the Xeon X3210 for the gaming tests, and was generally trounced by the much more expensive QX6700 (DUH).
Now if we end up the rankings... we see that over the whole set of benchmarks the Phenom 9900 was not far behind the QX6700 in the number of first place finishes!
Moral of the story:
The Phenom 9900 running at 2.66GHz DEFINITELY outperforms a Xeon X3210 running at 2.13GHz - as it should.
For gaming, at unrealistically low resolutions for todays gamers, the QX6700 badly beats the Phenom 9900 - but no one really games at those resolutions, and as World In Conflict shows, modern games are GPU not CPU bound.
Frankly, the Phenom 9900 did much better against the QX6700 than I had expected. Now if the BIOS on the motherboard allowed finer grained control over memory speed settings, I might have been able to squeeze even more performance out of the chip at both stock and overclocked settings; but a future BIOS revision may well take care of that.
AMD needs the 45nm shrink badly. It will address the power consumption issue - there is no question that Intel quad core processors are "greener" right now - and when that is combined with the purported architectural enhancements, and the ability to ramp up the clock speed due to lower voltage 45nm designs generating less heat, Intel may again have some real competition for the high end next year.
But you know what?
The consumer wins already. If aggressively priced, the Phenom 9900 - and upcoming lower end "Black Edition" Phenom's - will present an excellent value proposition to consumers, and provide a very viable alternative in the low and mid-range markets.
Overclocking
After the problems AMD has had in getting higher speed grades of the Phenom to market, I was initially worried about the overclockability of the Phenom 9900. It is less than three weeks since the release of the Phenom 9600, and I could only overclock our Phenom 9600 to 2.645GHz!
Fortunately it looks like the Phenom 9900 AMD sent us is a newer stepping - B2 to be precise - and even though it is not the fabled B3 stepping that will fix TLB erratum, it does seem to have opened up some extra "elbow" room.
How much?
How does a stable 3.0GHz strike you?
Mind you, I had to replace the stock AMD cooler with its sticky heatsink compound with a Noctua 12 and Arctic Silver - but that dropped the CPU temperature by a whopping 15'C when idling at 3GHz!
In order to run at 3GHz, I did the following:
I also tried to overclock with the HT speed, however I could not get even 233x13 stable on this Asus M3A32-MVP.
Conclusion
I was actually quite pleased with the performance of the Phenom 9900.
Given AMD's public commitment to fixing the TLB errata, and their move to 45nm with the "K10.5" architecture (Barcelona with some additional improvements) I suspect that AMD will be more competitive on the high end sometime next year.
But even until then... looking at the benchmarks, it did not seem like the Phenom 9900 was that far behind - if it was in fact behind.
The table below shows the relative rank of the three processors in the results for each benchmark - one means best result, two means second best, and third means last. These rankings are based purely on stock performance, as that is how the vast majority of people run their systems.
| Phenom 9900 | QX6700 | Xeon X3210 | |
| Business Winstone | 1 | 2 | 3 |
| Content Creation | 1 | 2 | 3 |
| Sandra CPU | 3 | 1 | 2 |
| Sandra Memory | 1 | 2 | 3 |
| WinRAR MT | 1 | 2 | 3 |
| RM Read | 1 | 2 | 3 |
| RM Write | 1 | 3 | 2 |
| RM Latency | 1 | 2 | 3 |
| RM Bandwidth | 1 | 3 | 2 |
| LAME MT | 2 | 1 | 3 |
| TMPGEnc | 3 | 1 | 2 |
| CineBench | 2 | 1 | 3 |
| POVRay | 3 | 1 | 2 |
| COD | 2 | 1 | 3 |
| Comanche | 2 | 1 | 3 |
| Doom 3 | 2 | 1 | 3 |
| Halo | 2 | 1 | 3 |
| Jedi | 2 | 1 | 3 |
| UT2004 | 2 | 1 | 3 |
| TOTAL 1st place results | 8 | 11 | 0 |
That is a very interesting result.
The Phenom 9900 generally took first place in business applications and memory oriented benchmarks - sometime with quite a margin.
The Phenom consistently beat the Xeon X3210 for the gaming tests, and was generally trounced by the much more expensive QX6700 (DUH).
Now if we end up the rankings... we see that over the whole set of benchmarks the Phenom 9900 was not far behind the QX6700 in the number of first place finishes!
Moral of the story:
The Phenom 9900 running at 2.66GHz DEFINITELY outperforms a Xeon X3210 running at 2.13GHz - as it should.
For gaming, at unrealistically low resolutions for todays gamers, the QX6700 badly beats the Phenom 9900 - but no one really games at those resolutions, and as World In Conflict shows, modern games are GPU not CPU bound.
Frankly, the Phenom 9900 did much better against the QX6700 than I had expected. Now if the BIOS on the motherboard allowed finer grained control over memory speed settings, I might have been able to squeeze even more performance out of the chip at both stock and overclocked settings; but a future BIOS revision may well take care of that.
AMD needs the 45nm shrink badly. It will address the power consumption issue - there is no question that Intel quad core processors are "greener" right now - and when that is combined with the purported architectural enhancements, and the ability to ramp up the clock speed due to lower voltage 45nm designs generating less heat, Intel may again have some real competition for the high end next year.
But you know what?
The consumer wins already. If aggressively priced, the Phenom 9900 - and upcoming lower end "Black Edition" Phenom's - will present an excellent value proposition to consumers, and provide a very viable alternative in the low and mid-range markets.
Overclocking
After the problems AMD has had in getting higher speed grades of the Phenom to market, I was initially worried about the overclockability of the Phenom 9900. It is less than three weeks since the release of the Phenom 9600, and I could only overclock our Phenom 9600 to 2.645GHz!
Fortunately it looks like the Phenom 9900 AMD sent us is a newer stepping - B2 to be precise - and even though it is not the fabled B3 stepping that will fix TLB erratum, it does seem to have opened up some extra "elbow" room.
How much?
How does a stable 3.0GHz strike you?
Mind you, I had to replace the stock AMD cooler with its sticky heatsink compound with a Noctua 12 and Arctic Silver - but that dropped the CPU temperature by a whopping 15'C when idling at 3GHz!
In order to run at 3GHz, I did the following:
I also tried to overclock with the HT speed, however I could not get even 233x13 stable on this Asus M3A32-MVP.
Conclusion
I was actually quite pleased with the performance of the Phenom 9900.
Given AMD's public commitment to fixing the TLB errata, and their move to 45nm with the "K10.5" architecture (Barcelona with some additional improvements) I suspect that AMD will be more competitive on the high end sometime next year.
But even until then... looking at the benchmarks, it did not seem like the Phenom 9900 was that far behind - if it was in fact behind.
The table below shows the relative rank of the three processors in the results for each benchmark - one means best result, two means second best, and third means last. These rankings are based purely on stock performance, as that is how the vast majority of people run their systems.
| Phenom 9900 | QX6700 | Xeon X3210 | |
| Business Winstone | 1 | 2 | 3 |
| Content Creation | 1 | 2 | 3 |
| Sandra CPU | 3 | 1 | 2 |
| Sandra Memory | 1 | 2 | 3 |
| WinRAR MT | 1 | 2 | 3 |
| RM Read | 1 | 2 | 3 |
| RM Write | 1 | 3 | 2 |
| RM Latency | 1 | 2 | 3 |
| RM Bandwidth | 1 | 3 | 2 |
| LAME MT | 2 | 1 | 3 |
| TMPGEnc | 3 | 1 | 2 |
| CineBench | 2 | 1 | 3 |
| POVRay | 3 | 1 | 2 |
| COD | 2 | 1 | 3 |
| Comanche | 2 | 1 | 3 |
| Doom 3 | 2 | 1 | 3 |
| Halo | 2 | 1 | 3 |
| Jedi | 2 | 1 | 3 |
| UT2004 | 2 | 1 | 3 |
| TOTAL 1st place results | 8 | 11 | 0 |
That is a very interesting result.
The Phenom 9900 generally took first place in business applications and memory oriented benchmarks - sometime with quite a margin.
The Phenom consistently beat the Xeon X3210 for the gaming tests, and was generally trounced by the much more expensive QX6700 (DUH).
Now if we end up the rankings... we see that over the whole set of benchmarks the Phenom 9900 was not far behind the QX6700 in the number of first place finishes!
Moral of the story:
The Phenom 9900 running at 2.66GHz DEFINITELY outperforms a Xeon X3210 running at 2.13GHz - as it should.
For gaming, at unrealistically low resolutions for todays gamers, the QX6700 badly beats the Phenom 9900 - but no one really games at those resolutions, and as World In Conflict shows, modern games are GPU not CPU bound.
Frankly, the Phenom 9900 did much better against the QX6700 than I had expected. Now if the BIOS on the motherboard allowed finer grained control over memory speed settings, I might have been able to squeeze even more performance out of the chip at both stock and overclocked settings; but a future BIOS revision may well take care of that.
AMD needs the 45nm shrink badly. It will address the power consumption issue - there is no question that Intel quad core processors are "greener" right now - and when that is combined with the purported architectural enhancements, and the ability to ramp up the clock speed due to lower voltage 45nm designs generating less heat, Intel may again have some real competition for the high end next year.
But you know what?
The consumer wins already. If aggressively priced, the Phenom 9900 - and upcoming lower end "Black Edition" Phenom's - will present an excellent value proposition to consumers, and provide a very viable alternative in the low and mid-range markets.
Please do not redistribute or use this article in whole, or in part, for commercial purposes.